NFDK firmware supports 48-bit dma addressing and
parses 16 high bits of dma addresses.
In nfp_nfdk_tx_desc, dma related structure and tso
related structure are union. When "mss" be filled
with nonzero value due to enable tso, the memory used
by "padding" may be also filled. Then, firmware may
parse wrong dma addresses which causes TX watchdog
timeout problem.
This patch removes padding and unifies the dma_addr_hi
bits with the one in firmware. nfp_nfdk_tx_desc_set_dma_addr
is also added to match this change.
Fixes: c10d12e3dc ("nfp: add support for NFDK data path")
Signed-off-by: Fei Qin <fei.qin@corigine.com>
Signed-off-by: Yinjun Zhang <yinjun.zhang@corigine.com>
Signed-off-by: Louis Peens <louis.peens@corigine.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
Link: https://lore.kernel.org/r/20220601083449.50556-1-simon.horman@corigine.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Add new data path. The TX is completely different, each packet
has multiple descriptor entries (between 2 and 32). TX ring is
divided into blocks 32 descriptor, and descritors of one packet
can't cross block bounds. The RX side is the same for now.
ABI version 5 or later is required. There is no support for
VLAN insertion on TX. XDP_TX action and AF_XDP zero-copy is not
implemented in NFDK path.
Changes to Jakub's work:
* Move statistics of hw_csum_tx after jumbo packet's segmentation.
* Set L3_CSUM flag to enable recaculating of L3 header checksum
in ipv4 case.
* Mark the case of TSO a packet with metadata prepended as
unsupported.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Xingfeng Hu <xingfeng.hu@corigine.com>
Signed-off-by: Yinjun Zhang <yinjun.zhang@corigine.com>
Signed-off-by: Dianchao Wang <dianchao.wang@corigine.com>
Signed-off-by: Fei Qin <fei.qin@corigine.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Prepare for choosing data path based on the firmware version field.
Exploit one bit from the reserved byte in the firmware version field
as the data path type. We need the firmware version right after
vNIC is allocated, so it has to be read inside nfp_net_alloc(),
callers don't have to set it afterwards.
Following patches will bring the implementation of the second data
path.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Fei Qin <fei.qin@corigine.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Newer versions of the PCIe microcode support writing back the
position of the TX pointer back into host memory. This speeds
up TX completions, because we avoid a read from device memory
(replacing PCIe read with DMA coherent read).
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Fei Qin <fei.qin@corigine.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
New datapaths may use multiple descriptor units to describe
a single packet. Prepare for that by adding a descriptors
per simple frame constant into ring size calculations.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Fei Qin <fei.qin@corigine.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
To reduce the coupling of slow path ring implementations and their
callers, use callbacks instead.
Changes to Jakub's work:
* Also use callbacks for xmit functions
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Yinjun Zhang <yinjun.zhang@corigine.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In preparation for support for a new datapath format move all
ring and fast path logic into separate files. It is basically
a verbatim move with some wrapping functions, no new structures
and functions added.
The current data path is called NFD3 from the initial version
of the driver ABI it used. The non-fast path, but ring related
functions are moved to nfp_net_dp.c file.
Changes to Jakub's work:
* Rebase on xsk related code.
* Split the patch, move the callback changes to next commit.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Fei Qin <fei.qin@corigine.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
NFP3800 has slightly different queue controller range bounds.
Use the static chip data instead of defines. This commit
still assumes unchanged descriptor format. Later datapath
changes will allow adjusting for descriptor accounting.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Fei Qin <fei.qin@corigine.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The queue controller (QCP) is accessed based on a device specific
offset. The NFP3800 device also supports more queues.
Furthermore, the NFP3800 VFs also access the QCP differently to how the
NFP6000 VFs accesses it, though still indirectly. Fortunately, we can
remove the offset all together for both VF types. This is safe for
NFP6000 VFs since the offset was effectively a wrap around and only used
for convenience to have it set the same as the NFP6000 PF.
Use nfp_dev_info to store queue controller parameters.
Signed-off-by: Dirk van der Merwe <dirk.vandermerwe@netronome.com>
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Fei Qin <fei.qin@corigine.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
In preparation for supporting new chip add a driver data structure
which will hold per-chip-version information such as register
offsets.
Plumb it through to the relevant functions (nfpcore and nfp_net).
For now only a very simple member holding chip names is added,
following commits will add more.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Fei Qin <fei.qin@corigine.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Multiple writes cause intermediate pointer values that do not
end on complete TX descriptors.
The QCP peripheral on the NFP provides a number of access
modes. In some access modes, the maximum amount to add must
be restricted to a 6bit value. The particular access mode
used by _nfp_qcp_ptr_add() has no such restrictions, so the
"< NFP_QCP_MAX_ADD" test is unnecessary.
Note that trying to add more that the configured ring size
in a single add will cause a QCP overflow, caught and handled
by the QCP peripheral.
Signed-off-by: Christo du Toit <christo.du.toit@netronome.com>
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Fei Qin <fei.qin@corigine.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
This patch adds zero-copy Rx and Tx support for AF_XDP sockets. It do so
by adding a separate NAPI poll function that is attached to a each
channel when the XSK socket is attached with XDP_SETUP_XSK_POOL, and
restored when the XSK socket is terminated, this is done per channel.
Support for XDP_TX is implemented and the XDP buffer can safely be moved
from the Rx to the Tx queue and correctly freed and returned to the XSK
pool once it's transmitted.
Note that when AF_XDP zero-copy is enabled, the XDP action XDP_PASS
will allocate a new buffer and copy the zero-copy frame prior
passing it to the kernel stack.
This patch is based on previous work by Jakub Kicinski.
Signed-off-by: Niklas Söderlund <niklas.soderlund@corigine.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Each data path needs an array of xsk pools to track if an xsk socket is
in use. Add this array and make sure it's handled correctly when the
data path is duplicated.
Signed-off-by: Niklas Söderlund <niklas.soderlund@corigine.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
There are some common functionality that can be reused in the upcoming
AF_XDP support. Expose those functions in the header. While at it mark
some arguments of nfp_net_rx_csum() as const.
Signed-off-by: Niklas Söderlund <niklas.soderlund@corigine.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use nn->tlv_caps.me_freq_mhz instead of nn->me_freq_mhz to check whether
rx-usecs/tx-usecs is valid.
This is because nn->tlv_caps.me_freq_mhz represents the clock_freq (MHz) of
the flow processing cores (FPC) on the NIC. While nn->me_freq_mhz is not
be set.
Fixes: ce991ab666 ("nfp: read ME frequency from vNIC ctrl memory")
Signed-off-by: Diana Wang <na.wang@corigine.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
NFP conversion is pretty straightforward. We want to be able
to sleep, and only get callbacks when the device is open.
NFP did not ask for port replay when ports were removed, now
new infra will provide this feature for free.
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The simple RX resync strategy controlled by the kernel does not
guarantee as good results as if the device helps by detecting
the potential record boundaries and keeping track of them.
We've called this strategy stream scan in the tls-offload doc.
Implement this strategy for the NFP. The device sends a request
for record boundary confirmation, which is then recorded in
per-TLS socket state and responded to once record is reached.
Because the device keeps track of records passing after the
request was sent the response is not as latency sensitive as
when kernel just tries to tell the device the information
about the next record.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If control channel MTU is too low to support map operations a warning
will be printed. This is not enough, we want to make sure probe fails
in such scenario, as this would clearly be a faulty configuration.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Acked-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Connection 4 tuple reuse is slightly problematic - TLS socket
and context do not get destroyed until all the associated skbs
left the system and all references are released. This leads
to stale connection entry in the device preventing addition
of new one if the 4 tuple is reused quickly enough.
Instead of using read 4 tuple as the key use a unique ID.
Set the protocol to TCP and port to 0 to ensure no collisions
with real connections.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Dirk van der Merwe <dirk.vandermerwe@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Some control messages must be sent from atomic context. The mailbox
takes sleeping locks and uses a waitqueue so add a "posted" version
of communication.
Trylock the semaphore and if that's successful kick of the device
communication. The device communication will be completed from
a workqueue, which will also release the semaphore.
If locks are taken queue the message and return. Schedule a
different workqueue to take the semaphore and run the communication.
Note that the there are currently no atomic users which would actually
need the return value, so all replies to posted messages are just
freed.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Dirk van der Merwe <dirk.vandermerwe@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Firmware indicates when a packet has been decrypted by reusing the
currently unused BPF flag. Transfer this information into the skb
and provide a statistic of all decrypted segments.
Signed-off-by: Dirk van der Merwe <dirk.vandermerwe@netronome.com>
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Count TX TLS packets: successes, out of order, and dropped due to
missing record info. Make sure the RX and TX completion statistics
don't share cache lines with TX ones as much as possible. With TLS
stats they are no longer reasonably aligned.
Signed-off-by: Dirk van der Merwe <dirk.vandermerwe@netronome.com>
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch adds the functionality to add and delete TLS connections on
the NFP, received from the kernel TLS callbacks.
Make use of the common control message (CCM) infrastructure to propagate
the kernel state to firmware.
Signed-off-by: Dirk van der Merwe <dirk.vandermerwe@netronome.com>
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Prepend connection handle to each transmitted TLS packet.
For each connection, the driver tracks the next sequence number
expected. If an out of order packet is observed, the driver calls into
the TLS kernel code to reencrypt that particular skb.
Signed-off-by: Dirk van der Merwe <dirk.vandermerwe@netronome.com>
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
FW may prefer to handle some communication via a mailbox
or the vNIC may simply not have a control queue (VFs).
Add a way of exchanging ccm-compatible messages via a
mailbox.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Dirk van der Merwe <dirk.vandermerwe@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We will need to release the bar lock from a workqueue
so move from a mutex to a semaphore. This lock should
not be too hot. Unfortunately semaphores don't have
lockdep support.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Dirk van der Merwe <dirk.vandermerwe@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Soon we will try to write to the vNIC mailbox without RTNL held.
Add a new mutex to protect access to specific parts of the PCI
control BAR.
Move the mailbox size checking to the mailbox lock() helper, where
it can be more effective (happen prior to potential overwrite of
other data).
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Dirk van der Merwe <dirk.vandermerwe@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
FW reconfiguration timeouts are a common indicator of FW trouble.
To make debugging easier print requested update and control word
when reconfiguration fails.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Dirk van der Merwe <dirk.vandermerwe@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Chained descriptors for fragments need to duplicate all the descriptor
fields of the skb head, so we copy the descriptor and then modify the
relevant fields. This is wasteful, because the top half of the descriptor
will get overwritten entirely while the bottom half is not modified at all.
Copy only the bottom half. This saves us 0.3% of CPU in a GSO test.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Dirk van der Merwe <dirk.vandermerwe@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Learn how to set the DSCP map. FW uses a packed array which
geometry depends on the number of supported priorities and
virtual queues. Write code to assemble this map and to communicate
the setting to the FW via mailbox.
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>
Commit 7f1c684a89 ("nfp: setup xdp_rxq_info") mixed the cache
cold and cache hot data in the nfp_net_rx_ring structure (ignoring
the feedback), to try to fit the structure into 2 cache lines
after struct xdp_rxq_info was added. Now that we are about to add
a new field the structure will grow back to 3 cache lines, so
order the members correctly.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Dirk van der Merwe <dirk.vandermerwe@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Split handling of offloaded and driver programs completely. Since
offloaded programs always come with XDP_FLAGS_HW_MODE set in reality
there could be no sharing, anyway, programs would only be installed
in driver or in hardware. Splitting the handling allows us to install
programs in HW and in driver at the same time.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Basic operations drivers perform during xdp setup and query can
be moved to helpers in the core. Encapsulate program and flags
into a structure and add helpers. Note that the structure is
intended as the "main" program information source in the driver.
Most drivers will additionally place the program pointer in their
fast path or ring structures.
The helpers don't have a huge impact now, but they will
decrease the code duplication when programs can be installed
in HW and driver at the same time. Encapsulating the basic
operations in helpers will hopefully also reduce the number
of changes to drivers which adopt them.
Helpers could really be static inline, but they depend on
definition of struct netdev_bpf which means they'd have
to be placed in netdevice.h, an already 4500 line header.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
.ndo_get_phys_port_name was recently extended to support multi-vNIC
FWs. These are firmwares which can have more than one vNIC per PF
without associated port (e.g. Adaptive Buffer Management FW), therefore
we need a way of distinguishing the vNICs. Unfortunately, it's too
late to make flower use the same naming. Flower users may depend on
.ndo_get_phys_port_name returning -EOPNOTSUPP, for example the name
udev gave the PF vNIC was just the bare PCI device-based name before
the change, and will have 'nn0' appended after.
To ensure flower's vNIC doesn't have phys_port_name attribute, add
a flag to vNIC struct and set it in flower code. New projects will
not set the flag adhere to the naming scheme from the start.
Fixes: 51c1df83e3 ("nfp: assign vNIC id as phys_port_name of vNICs which are not ports")
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Dirk van der Merwe <dirk.vandermerwe@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When NFP is modelled as a switch we assign phys_port_name to respective
port(representor )s:
vNIC0 - | - PF port (pf%d) MAC/PHY (p%d[s%d]) - |E==
In most cases there is only one vNIC for communication with the switch.
If there is more than one we need to be able to identify them. Use %d
as phys_port_name of the vNICs.
We don't have to pass ID to nfp_net_debugfs_vnic_add() separately any
more.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Dirk van der Merwe <dirk.vandermerwe@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We are currently counting packets with CHECKSUM_COMPLETE as
"hw_rx_csum_ok". This is confusing. Add a new counter.
To make sure it fits in the same cacheline move the less used
error counter to a different location.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Dirk van der Merwe <dirk.vandermerwe@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Commit 84ce5b9877 ("scripts: kernel-doc: improve nested logic to
handle multiple identifiers") improved the handling of nested structure
definitions in scripts/kernel-doc, and changed the expected format of
documentation. This causes new warnings to appear on W=1 builds.
Only comment changes.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
NFP is entirely programmable, including the PCI data interface.
Using a fixed control BAR layout certainly makes implementations
easier, but require careful considerations when space is allocated.
Once BAR area is allocated to one feature nothing else can use it.
Allocating space statically also requires it to be sized upfront,
which leads to either unnecessary limitation or wastage.
We currently have a 32bit capability word defined which tells drivers
which application FW features are supported. Most of the bits
are exhausted. The same bits are also reused for enabling specific
features. Bulk of capabilities don't have a need for an enable bit,
however, leading to confusion and wastage.
TLVs seems like a better fit for expressing capabilities of applications
running on programmable hardware.
This patch leaves the front of the BAR as is, and declares a TLV
capability start at offset 0x58. Most of the space up to 0x0d90
is already allocated, but the used space can be wrapped with RESERVED
TLVs. E.g.:
Address Type Length
0x0058 RESERVED 0xe00 /* Wrap basic structures */
0x0e5c FEATURE_A 0x004
0x0e64 FEATURE_B 0x004
0x0e6c RESERVED 0x990 /* Wrap qeueue stats */
0x1800 FEATURE_C 0x100
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Dirk van der Merwe <dirk.vandermerwe@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
For map support we will need to send and receive control messages.
Add basic support for sending a message to FW, and waiting for a
reply.
Control messages are tagged with a 16 bit ID. Add a simple ID
allocator and make sure we don't allow too many messages in flight,
to avoid request <> reply mismatches.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Driver hook points for xdp_rxq_info:
* reg : nfp_net_rx_ring_alloc
* unreg: nfp_net_rx_ring_free
In struct nfp_net_rx_ring moved member @size into a hole on 64-bit.
Thus, the size remaines the same after adding member @xdp_rxq.
Cc: oss-drivers@netronome.com
Cc: Jakub Kicinski <jakub.kicinski@netronome.com>
Cc: Simon Horman <simon.horman@netronome.com>
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Since commit 3a025e1d1c ("Add optional check for bad kernel-doc
comments") when built with W=1 build will complain about kdoc errors.
Fix the kdoc issues we have. kdoc is still confused by defines in
nfp_net_ctrl.h but those are not really errors.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
