Age | Commit message (Collapse) | Author |
|
When CQE mode or DIM state is changed, gracefully reconfigure channels to
handle new configuration. Previously, would create new channels that would
reflect the changes rather than update the original channels.
Co-developed-by: Nabil S. Alramli <dev@nalramli.com>
Signed-off-by: Nabil S. Alramli <dev@nalramli.com>
Co-developed-by: Joe Damato <jdamato@fastly.com>
Signed-off-by: Joe Damato <jdamato@fastly.com>
Signed-off-by: Rahul Rameshbabu <rrameshbabu@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://lore.kernel.org/r/20240419080445.417574-5-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Each queue counter object counts some events (in hardware) for the RQs
that are attached to it, like events of packet drops due to no receive
WQE (rx_out_of_buffer).
Each RQ can be attached to a queue counter only within the same vhca. To
still cover all RQs with these counters, we create multiple instances,
one per vhca.
The result that's shown to the user is now the sum of all instances.
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
The cited commits missed passing frag_size to __xdp_rxq_info_reg, which
is required by bpf_xdp_adjust_tail to support growing the tail pointer
in fragmented packets. Pass the missing parameter when the current RQ
mode allows XDP multi buffer.
Fixes: ea5d49bdae8b ("net/mlx5e: Add XDP multi buffer support to the non-linear legacy RQ")
Fixes: 9cb9482ef10e ("net/mlx5e: Use fragments of the same size in non-linear legacy RQ with XDP")
Signed-off-by: Maxim Mikityanskiy <maxtram95@gmail.com>
Cc: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
Here we add support for multi-buffer XDP handling in Striding RQ, which
is our default out-of-the-box RQ type. Before this series, loading such
an XDP program would fail, until you switch to the legacy RQ (by
unsetting the rx_striding_rq priv-flag).
To overcome the lack of headroom and tailroom between the strides, we
allocate a side page to be used for the descriptor (xdp_buff / skb) and
the linear part. When an XDP program is attached, we structure the
xdp_buff so that it contains no data in the linear part, and the whole
packet resides in the fragments.
In case of XDP_PASS, where an SKB still needs to be created, we copy up
to 256 bytes to its linear part, to match the current behavior, and
satisfy functions that assume finding the packet headers in the SKB
linear part (like eth_type_trans).
Performance testing:
Packet rate test, 64 bytes, 32 channels, MTU 9000 bytes.
CPU: Intel(R) Xeon(R) Platinum 8380 CPU @ 2.30GHz.
NIC: ConnectX-6 Dx, at 100 Gbps.
+----------+-------------+-------------+---------+
| Test | Legacy RQ | Striding RQ | Speedup |
+----------+-------------+-------------+---------+
| XDP_DROP | 101,615,544 | 117,191,020 | +15% |
+----------+-------------+-------------+---------+
| XDP_TX | 95,608,169 | 117,043,422 | +22% |
+----------+-------------+-------------+---------+
Reviewed-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Params info print was meant to be printed on load.
With time, new calls to mlx5e_init_rq_type_params and
mlx5e_build_rq_params were added, mistakenly printing
the params once again.
Move the print to were it belongs, in mlx5e_probe.
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Gal Pressman <gal@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
Instead of passing the unaligned flag, pass an enum that indicates the
UMR mode. The next commit will add the third mode (KLM for certain
configurations of XSK), which will be added to this enum instead of
adding another bool flag everywhere.
Signed-off-by: Maxim Mikityanskiy <maximmi@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
In the initial implementation of XSK in mlx5e, XSK RQs coexisted with
regular RQs in the same channel. The main idea was to allow RSS work the
same for regular traffic, without need to reconfigure RSS to exclude XSK
queues.
However, this scheme didn't prove to be beneficial, mainly because of
incompatibility with other vendors. Some tools don't properly support
using higher indices for XSK queues, some tools get confused with the
double amount of RQs exposed in sysfs. Some use cases are purely XSK,
and allocating the same amount of unused regular RQs is a waste of
resources.
This commit changes the queuing scheme to the standard one, where XSK
RQs replace regular RQs on the channels where XSK sockets are open. Two
RQs still exist in the channel to allow failsafe disable of XSK, but
only one is exposed at a time. The next commit will achieve the desired
memory save by flushing the buffers when the regular RQ is unused.
As the result of this transition:
1. It's possible to use RSS contexts over XSK RQs.
2. It's possible to dedicate all queues to XSK.
3. When XSK RQs coexist with regular RQs, the admin should make sure no
unwanted traffic goes into XSK RQs by either excluding them from RSS or
settings up the XDP program to return XDP_PASS for non-XSK traffic.
4. When using a mixed fleet of mlx5e devices and other netdevs, the same
configuration can be applied. If the application supports the fallback
to copy mode on unsupported drivers, it will work too.
Signed-off-by: Maxim Mikityanskiy <maximmi@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
UMR MTTs used in striding RQ have certain alignment requirements. While
it's guaranteed to work when UMR pages are aligned to the UMR page size,
in practice it works then UMR pages are aligned to 8 bytes. However,
it's still not enough flexibility for the unaligned mode of XSK. This
patch leverages KSM to map UMR pages without alignment requirements,
when unaligned XSK is active. The downside is that KSM entries are twice
as big as MTTs, which limits the maximum WQE size, so regular RQs and
aligned XSK continue using MTTs.
Signed-off-by: Maxim Mikityanskiy <maximmi@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
This commit allows striding RQ to determine MTT page size at runtime,
instead of sticking to the compile-time PAGE_SIZE. This functionality
will be used by a following commit that adjusts the MTT page size to the
XSK frame size.
Stick with PAGE_SIZE for XSK on legacy RQ, as frag_stride is not used in
data path, it only helps calculate how pages are partitioned into
fragments, and PAGE_SIZE will ensure each fragment starts at the
beginning of a new allocation unit (XSK frame).
Signed-off-by: Maxim Mikityanskiy <maximmi@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Some of the parameters of striding RQ are compile-time constants, but
they are going to become dynamically calculated at runtime in a
following commit. This commit prepares the datapath to take cached
runtime parameters, prefilled at queue creation.
New fields added to struct mlx5e_rq fit into an existing 7-byte hole.
Signed-off-by: Maxim Mikityanskiy <maximmi@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Normally, the MTU change requires reopening the channels, but it can be
skipped if the new MTU doesn't change any of the queue parameters and if
MTU is not used in the data path.
The shortcut is applicable to the non-linear mode of striding RQ,
because the only thing affected by MTU is the queue length. As ethtool
sets the queue length in packets, but striding RQ length is defined in
strides or bytes, we estimate the RQ length to be at least as big as the
requested number of MTU-sized packets, that's why it depends on MTU.
Improve the shortcut by actually checking whether the RQ length stayed
the same, instead of an intermediate step in the calculation.
As MTU also affects the SHAMPO parameters, skip the shortcut if SHAMPO
is in use.
Signed-off-by: Maxim Mikityanskiy <maximmi@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
One of the checks in mlx5e_rx_is_linear_skb verifies that the RX buffer
fits into the XSK frame size. Remove the duplicating check from
mlx5e_validate_xsk_param. It allows to make mlx5e_rx_get_min_frag_sz
static.
Remove mlx5e_rx_is_xdp altogether, as its only usage is located in a
branch where xsk == NULL.
Signed-off-by: Maxim Mikityanskiy <maximmi@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Currently, the driver can silently fall back to legacy RQ after enabling
XDP, even if striding RQ was active before. It happens when PAGE_SIZE is
bigger than the maximum supported stride size. This commit changes this
behavior to more straightforward: if an operation (enabling XDP) doesn't
support the current parameters (striding RQ mode), it fails.
Signed-off-by: Maxim Mikityanskiy <maximmi@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
mlx5e_verify_rx_mpwqe_strides is only used in en/params.c, so it can be
made static and removed from en/params.h.
Signed-off-by: Maxim Mikityanskiy <maximmi@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
When MPWQE is disabled, mlx5e_open_xdpsq() prefills the common fields of
WQEs in the XDP SQ to save time when sending packets.
mlx5e_xmit_xdp_frame() runs on the prefilled fields, however, sending
multi buffer XDP frames would require changing some of these fields on a
per-packet basis. Besides that, mlx5e_xmit_xdp_frame() will be used as a
fallback to send multi buffer XDP frames when MPWQE is enabled (MPWQE
can only handle linear packets).
In order to prepare for XDP multi buffer support, this commit introduces
a mode for mlx5e_xmit_xdp_frame() that fills all the fields itself.
Signed-off-by: Maxim Mikityanskiy <maximmi@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
In RQs of type multi-packet WQE (Striding RQ), each WQE is relatively
large (typically 256KB) but their number is relatively small (8 in
default).
Re-mapping the descriptors' buffers before re-posting them is done via
UMR (User-Mode Memory Registration) operations.
On the one hand, posting UMR WQEs in bulks reduces communication overhead
with the HW and better utilizes its processing units.
On the other hand, delaying the WQE repost operations for a small RQ
(say, of 4 WQEs) might drastically hit its performance, causing packet
drops due to no receive buffer, for high or bursty incoming packets rate.
Here we restrict the bulk size for too small RQs. Effectively, with the current
constants, RQ of size 4 (minimum allowed) would have no bulking, while larger
RQs will continue working with bulks of 2.
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
This commit introduces the control path infrastructure for SHAMPO feature.
SHAMPO feature enables packet stitching by splitting packets to
header and payload, the header is placed on a dedicated buffer
and the payload on the RX ring, this allows stitching the data part
of a flow together continuously in the receive buffer.
SHAMPO feature is implemented as linked list striding RQ feature.
To support packets splitting and payload stitching:
- Enlarge the ICOSQ and the correspond CQ to support the header buffer
memory regions.
- Add support to create linked list striding RQ with SHAMPO feature set
in the open_rq function.
- Add deallocation function and corresponded calls for SHAMPO header
buffer.
- Add mlx5e_create_umr_klm_mkey to support KLM mkey for the header
buffer.
- Rename mlx5e_create_umr_mkey to mlx5e_create_umr_mtt_mkey.
Signed-off-by: Ben Ben-Ishay <benishay@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
This series introduces new packet merge type, therefore rename lro
functions to packet merge to support the new merge type:
- Generalize + rename mlx5e_build_tir_ctx_lro to
mlx5e_build_tir_ctx_packet_merge.
- Rename mlx5e_modify_tirs_lro to mlx5e_modify_tirs_packet_merge.
- Rename lro bit in mlx5_ifc_modify_tir_bitmask_bits to packet_merge.
- Rename lro_en in mlx5e_params to packet_merge_type type and combine
packet_merge params into one struct mlx5e_packet_merge_param.
Signed-off-by: Khalid Manaa <khalidm@nvidia.com>
Signed-off-by: Ben Ben-Ishay <benishay@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
The functions that build TIR context for TIR create and modify commands
used to depend on struct mlx5e_priv and fetch some values directly from
different places. It increased coupling of code and the chance of weird
misbehavior due to hidden complex dependencies.
As the first step, this commit removes the priv parameter from these
functions. Instead, the necessary values are passed directly.
Signed-off-by: Maxim Mikityanskiy <maximmi@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
Striding RQ attributes below are mutually dependent. An unaware
change to one might take the others out of the valid range derived
by the HW caps:
- The MPWQE size in bytes
- The number of strides in a MPWQE
- The stride size
Add checks to verify they are valid and comply to the HW spec
and SW assumptions/requirements.
This is not a fix, no particular issue exists today.
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
When the TLS logic finds a tcp seq match for a kTLS RX resync
request, it calls the driver callback function mlx5e_ktls_resync()
to handle it and communicate it to the device.
Errors might occur during mlx5e_ktls_resync(), however, they are not
reported to the stack. Moreover, there is no error handling in the
stack for these errors.
In this patch, the driver obtains responsibility on errors handling,
adding queue and retry mechanisms to these resyncs.
We maintain a linked list of resync matches, and try posting them
to the async ICOSQ in the NAPI context.
Only possible failure that demands driver handling is ICOSQ being full.
By relying on the NAPI mechanism, we make sure that the entries in list
will be handled when ICOSQ completions arrive and make some room
available.
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
Do not use generic struct mlx5e_priv as a parameter to param
functions, as it is too generic. All calculations of the channel's
param should be mainly based on struct mlx5_core_dev and
struct mlx5e_params. Additional info can be explicitly passed.
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Aya Levin <ayal@nvidia.com>
Reviewed-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
Take params logic out of en_main.c, into the dedicated params.c.
Some functions are now hidden and become static.
No functional change here.
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Reviewed-by: Aya Levin <ayal@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
Pass q_counter idintifier, instead of reading it from mlx5e_priv
parameter.
This is a step towards removing the mlx5e_priv parameter from all
params function and logic in the next patches of the series.
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Aya Levin <ayal@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
This commit adds support for HTB offload in the mlx5e driver.
Performance:
NIC: Mellanox ConnectX-6 Dx
CPU: Intel(R) Xeon(R) CPU E5-2680 v3 @ 2.50GHz (24 cores with HT)
100 Gbit/s line rate, 500 UDP streams @ ~200 Mbit/s each
48 traffic classes, flower used for steering
No shaping (rate limits set to 4 Gbit/s per TC) - checking for max
throughput.
Baseline: 98.7 Gbps, 8.25 Mpps
HTB: 6.7 Gbps, 0.56 Mpps
HTB offload: 95.6 Gbps, 8.00 Mpps
Limitations:
1. 256 leaf nodes, 3 levels of depth.
2. Granularity for ceil is 1 Mbit/s. Rates are converted to weights, and
the bandwidth is split among the siblings according to these weights.
Other parameters for classes are not supported.
Ethtool statistics support for QoS SQs are also added. The counters are
called qos_txN_*, where N is the QoS queue number (starting from 0, the
numeration is separate from the normal SQs), and * is the counter name
(the counters are the same as for the normal SQs).
Signed-off-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Create a function to fill the fields of struct mlx5e_create_cq_param
based on a channel. The purpose is code reuse between normal CQs, XSK
CQs and the upcoming QoS CQs.
Signed-off-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
Transmitted packet timestamping accuracy can be improved when using
timestamp from the port, instead of packet CQE creation timestamp, as
it better reflects the actual time of a packet's transmit.
TX port timestamping is supported starting from ConnectX6-DX hardware.
Although at the original completion, only CQE timestamp can be attached,
we are able to get TX port timestamping via an additional completion over
a special CQ associated with the SQ (in addition to the regular CQ).
Driver to ignore the original packet completion timestamp, and report
back the timestamp of the special CQ completion. If the absolute timestamp
diff between the two completions is greater than 1 / 128 second, ignore
the TX port timestamp as it has a jitter which is too big.
No skb will be generate out of the extra completion.
Allocate additional CQ per ptpsq, to receive the TX port timestamp.
Driver to hold an skb FIFO in order to map between transmitted skb to
the two expected completions. When using ptpsq, hold double refcount on
the skb, to gaurantee it will not get released before both completions
arrive.
Expose dedicated counters of the ptp additional CQ and connect it to the
TX health reporter.
This patch improves TX Hardware timestamping offset to be less than 40ns
at a 100Gbps line rate, compared to 600ns before.
With that, making our HW compliant with G.8273.2 class C, and allow Linux
systems to be deployed in the 5G telco edge, where this standard is a
must.
Signed-off-by: Eran Ben Elisha <eranbe@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
Add TX PTP port object support for better TX timestamping accuracy.
Currently, driver supports CQE based TX port timestamp. Device
also offers TX port timestamp, which has less jitter and better
reflects the actual time of a packet's transmit.
Define new driver layout called ptpsq, on which driver will create
SQs that will support TX port timestamp for their transmitted packets.
Driver to identify PTP TX skbs and steer them to these dedicated SQs
as part of the select queue ndo.
Driver to hold ptpsq per TC and report them at
netif_set_real_num_tx_queues().
Add support for all needed functionality in order to xmit and poll
completions received via ptpsq.
Add ptpsq to the TX reporter recover, diagnose and dump methods.
Creation of ptpsqs is disabled by default, and can be enabled via
tx_port_ts private flag.
This patch steer all timestamp related packets to a ptpsq, but it
does not open the port timestamp support for it. The support will
be added in the following patch.
Signed-off-by: Eran Ben Elisha <eranbe@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
Stop room is a space that may be taken by WQEs in the SQ during a packet
transmit. It is used to check if next packet has enough room in the SQ.
Stop room guarantees this packet can be served and if not, the queue is
stopped, so no more packets are passed to the driver until it's ready.
Currently, stop_room size is calculated and validated upon tx queues
allocation. This makes it impossible to know if user provided valid
input for certain parameters when interface is down.
Instead, store stop_room in mlx5e_sq_param and create
mlx5e_validate_params(), to validate its fields upon user input even
when the interface is down.
Signed-off-by: Vladyslav Tarasiuk <vladyslavt@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
Take the CQ params into their respective RQ/SQ params.
Split the params build of the different ICOSQs (sync and async),
as they require different init values.
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Reviewed-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
Relax the requirements to the XSK frame size to allow it to be smaller
than a page and even not a power of two. The current implementation can
work in this mode, both with Striding RQ and without it.
The code that checks `mtu + headroom <= XSK frame size` is modified
accordingly. Any frame size between 2048 and PAGE_SIZE is accepted.
Functions that worked with pages only now work with XSK frames, even if
their size is different from PAGE_SIZE.
With XSK queues, regardless of the frame size, Striding RQ uses the
stride size of PAGE_SIZE, and UMR MTTs are posted using starting
addresses of frames, but PAGE_SIZE as page size. MTU guarantees that no
packet data will overlap with other frames. UMR MTT size is made equal
to the stride size of the RQ, because UMEM frames may come in random
order, and we need to handle them one by one. PAGE_SIZE is just a power
of two that is bigger than any allowed XSK frame size, and also it
doesn't require making additional changes to the code.
Signed-off-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Reviewed-by: Saeed Mahameed <saeedm@mellanox.com>
Acked-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
|
|
No XSK support in the enhanced IPoIB driver and representors.
Add a profile property to specify this, and enhance the logic
that calculates the max number of channels to take it into
account.
Fixes: db05815b36cb ("net/mlx5e: Add XSK zero-copy support")
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
This commit adds support for AF_XDP zero-copy RX and TX.
We create a dedicated XSK RQ inside the channel, it means that two
RQs are running simultaneously: one for non-XSK traffic and the other
for XSK traffic. The regular and XSK RQs use a single ID namespace split
into two halves: the lower half is regular RQs, and the upper half is
XSK RQs. When any zero-copy AF_XDP socket is active, changing the number
of channels is not allowed, because it would break to mapping between
XSK RQ IDs and channels.
XSK requires different page allocation and release routines. Such
functions as mlx5e_{alloc,free}_rx_mpwqe and mlx5e_{get,put}_rx_frag are
generic enough to be used for both regular and XSK RQs, and they use the
mlx5e_page_{alloc,release} wrappers around the real allocation
functions. Function pointers are not used to avoid losing the
performance with retpolines. Wherever it's certain that the regular
(non-XSK) page release function should be used, it's called directly.
Only the stats that could be meaningful for XSK are exposed to the
userspace. Those that don't take part in the XSK flow are not
considered.
Note that we don't wait for WQEs on the XSK RQ (unlike the regular RQ),
because the newer xdpsock sample doesn't provide any Fill Ring entries
at the setup stage.
We create a dedicated XSK SQ in the channel. This separation has its
advantages:
1. When the UMEM is closed, the XSK SQ can also be closed and stop
receiving completions. If an existing SQ was used for XSK, it would
continue receiving completions for the packets of the closed socket. If
a new UMEM was opened at that point, it would start getting completions
that don't belong to it.
2. Calculating statistics separately.
When the userspace kicks the TX, the driver triggers a hardware
interrupt by posting a NOP to a dedicated XSK ICO (internal control
operations) SQ, in order to trigger NAPI on the right CPU core. This XSK
ICO SQ is protected by a spinlock, as the userspace application may kick
the TX from any core.
Store the pointers to the UMEMs in the net device private context,
independently from the kernel. This way the driver can distinguish
between the zero-copy and non-zero-copy UMEMs. The kernel function
xdp_get_umem_from_qid does not care about this difference, but the
driver is only interested in zero-copy UMEMs, particularly, on the
cleanup it determines whether to close the XSK RQ and SQ or not by
looking at the presence of the UMEM. Use state_lock to protect the
access to this area of UMEM pointers.
LRO isn't compatible with XDP, but there may be active UMEMs while
XDP is off. If this is the case, don't allow LRO to ensure XDP can
be reenabled at any time.
The validation of XSK parameters typically happens when XSK queues
open. However, when the interface is down or the XDP program isn't
set, it's still possible to have active AF_XDP sockets and even to
open new, but the XSK queues will be closed. To cover these cases,
perform the validation also in these flows:
1. A new UMEM is registered, but the XSK queues aren't going to be
created due to missing XDP program or interface being down.
2. MTU changes while there are UMEMs registered.
Having this early check prevents mlx5e_open_channels from failing
at a later stage, where recovery is impossible and the application
has no chance to handle the error, because it got the successful
return value for an MTU change or XSK open operation.
The performance testing was performed on a machine with the following
configuration:
- 24 cores of Intel Xeon E5-2620 v3 @ 2.40 GHz
- Mellanox ConnectX-5 Ex with 100 Gbit/s link
The results with retpoline disabled, single stream:
txonly: 33.3 Mpps (21.5 Mpps with queue and app pinned to the same CPU)
rxdrop: 12.2 Mpps
l2fwd: 9.4 Mpps
The results with retpoline enabled, single stream:
txonly: 21.3 Mpps (14.1 Mpps with queue and app pinned to the same CPU)
rxdrop: 9.9 Mpps
l2fwd: 6.8 Mpps
Signed-off-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Acked-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
|
|
structs mlx5e_{rq,sq,cq,channel}_param are going to be used in the
upcoming XSK RX and TX patches. Move them to a header file to make
them accessible from other C files.
Signed-off-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Acked-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
|
|
Use the existing mlx5e_get_linear_rq_headroom function to calculate the
headroom for mlx5e_xdp_max_mtu. This function takes the XSK headroom
into consideration, which will be used in the following patches.
Signed-off-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Acked-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
|
|
Additional conditions introduced:
- XSK implies XDP.
- Headroom includes the XSK headroom if it exists.
- No space is reserved for struct shared_skb_info in XSK mode.
- Fragment size smaller than the XSK chunk size is not allowed.
A new auxiliary function mlx5e_get_linear_rq_headroom with the support
for XSK is introduced. Use this function in the implementation of
mlx5e_get_rq_headroom. Change headroom to u32 to match the headroom
field in struct xdp_umem.
Signed-off-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Acked-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
|
|
mdev is unused in mlx5e_rx_is_linear_skb.
Signed-off-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
This commit moves the parameter calculation functions to a separate file
for better modularity and code sharing with future features.
Signed-off-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|