HW expects the data size in DUMP WQEs to be up to MTU.
Make sure they are in range.
We elevate the frag page refcount by 'n-1', in addition to the
one obtained in tx_sync_info_get(), having an overall of 'n'
references. We bulk increments by using a single page_ref_add()
command, to optimize perfermance.
The refcounts are released one by one, by the corresponding completions.
Fixes: d2ead1f360 ("net/mlx5e: Add kTLS TX HW offload support")
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Reviewed-by: Eran Ben Elisha <eranbe@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
In TX resync flow where DUMP WQEs are posted, keep a pointer to
the fragment page to unref it upon completion, instead of saving
the whole fragment.
In addition, move it the end of the arguments list in tx_fill_wi().
Fixes: d2ead1f360 ("net/mlx5e: Add kTLS TX HW offload support")
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Reviewed-by: Eran Ben Elisha <eranbe@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Move function which indicates whether tunnel inner flow table is
supported from en.h to en_fs.c. It fits better right after tunnel
protocol rules definitions.
Signed-off-by: Aya Levin <ayal@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
When the VF LAG is in use, round-robin the TX affinity of channels among
the different ports, if supported by the firmware. Create a set of TISes
per port, while doing round-robin of the channels over the different
sets. Let all SQs of a channel share the same set of TISes.
If lag_tx_port_affinity HCA cap bit is supported, num_lag_ports > 1 and
we aren't the LACP owner (PF in the regular use), assign the affinities,
otherwise use tx_affinity == 0 in TIS context to let the FW assign the
affinities itself. The TISes of the LACP owner are mapped only to the
native physical port.
For VFs, the starting port for round-robin is determined by its vhca_id,
because a VF may have only one channel if attached to a single-core VM.
Signed-off-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Mark Bloch <markb@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
For better modularity and code sharing.
Function internal change to be introduced in the next patches.
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
HV VHCA stats agent is responsible on running a preiodic rx/tx
packets/bytes stats update. Currently the supported format is version
MLX5_HV_VHCA_STATS_VERSION. Block ID 1 is dedicated for statistics data
transfer from the VF to the PF.
The reporter fetch the statistics data from all opened channels, fill it
in a buffer and send it to mlx5_hv_vhca_write_agent.
As the stats layer should include some metadata per block (sequence and
offset), the HV VHCA layer shall modify the buffer before actually send it
over block 1.
Signed-off-by: Eran Ben Elisha <eranbe@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: Haiyang Zhang <haiyangz@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add support for report and recovery from error on completion on RQ by
setting the queue back to ready state. Handle only errors with a
syndrome indicating the RQ might enter error state and could be
recovered.
Signed-off-by: Aya Levin <ayal@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Add support for report and recovery from error on completion on ICOSQ.
Deactivate RQ and flush, then deactivate ICOSQ. Set the queue back to
ready state (firmware) and reset the ICOSQ and the RQ (software
resources). Finally, activate the ICOSQ and the RQ.
Signed-off-by: Aya Levin <ayal@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Cited patch deleted ethtool flash device support, as ethtool core can
fallback into devlink flash callback. However, this is supported only if
there is a devlink port registered over the corresponding netdevice.
As mlx5e do not have devlink port support over native netdevice, it broke
the ability to flash device via ethtool.
This patch re-add the ethtool callback to avoid user functionality breakage
when trying to flash device via ethtool.
Fixes: 9c8bca2637 ("mlx5: Move firmware flash implementation to devlink")
Signed-off-by: Eran Ben Elisha <eranbe@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
The TLS progress params context WQE should not include an
Eth segment, drop it.
In addition, align the tls_progress_params layout with the
HW specification document:
- fix the tisn field name.
- remove the valid bit.
Fixes: a12ff35e0f ("net/mlx5: Introduce TLS TX offload hardware bits and structures")
Fixes: d2ead1f360 ("net/mlx5e: Add kTLS TX HW offload support")
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
If capable, use zero inline mode in TX WQE for non-VLAN packets.
For VLAN ones, keep the enforcement of at least L2 inline mode,
unless the WQE VLAN insertion offload cap is on.
Performance:
Tested single core packet rate of 64Bytes.
NIC: ConnectX-5
CPU: Intel(R) Xeon(R) Gold 6154 CPU @ 3.00GHz
pktgen:
Before: 12.46 Mpps
After: 14.65 Mpps (+17.5%)
XDP_TX:
The MPWQE flow is not affected, as it already has this optimization.
So we test with priv-flag xdp_tx_mpwqe: off.
Before: 9.90 Mpps
After: 10.20 Mpps (+3%)
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Tested-by: Noam Stolero <noams@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
In MPWQE mode, when transmitting packets with XDP, a packet that is smaller
than a certain size (set to 256 bytes) would be sent inline within its WQE
TX descriptor (mem-copied), in case the hardware tx queue is congested
beyond a pre-defined water-mark.
If a MPWQE cannot contain an additional inline packet, we close this
MPWQE session, and send the packet inlined within the next MPWQE.
To save some MPWQE session close+open operations, we don't open MPWQE
sessions that are contiguously smaller than certain size (set to the
HW MPWQE maximum size). If there isn't enough contiguous room in the
send queue, we fill it with NOPs and wrap the send queue index around.
This way, qualified packets are always sent inline.
Perf tests:
Tested packet rate for UDP 64Byte multi-stream
over two dual port ConnectX-5 100Gbps NICs.
CPU: Intel(R) Xeon(R) CPU E5-2680 v3 @ 2.50GHz
XDP_TX:
With 24 channels:
| ------ | bounced packets | inlined packets | inline ratio |
| before | 113.6Mpps | 96.3Mpps | 84% |
| after | 115Mpps | 99.5Mpps | 86% |
With one channel:
| ------ | bounced packets | inlined packets | inline ratio |
| before | 6.7Mpps | 0pps | 0% |
| after | 6.8Mpps | 0pps | 0% |
As we can see, there is improvement in both inline ratio and overall
packet rate for 24 channels. Also, we see no degradation for the
one-channel case.
Signed-off-by: Shay Agroskin <shayag@mellanox.com>
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
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: db05815b36 ("net/mlx5e: Add XSK zero-copy support")
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
-----BEGIN PGP SIGNATURE-----
iQEzBAABCAAdFiEEGhZs6bAKwk/OTgTpSD+KveBX+j4FAl0ng7AACgkQSD+KveBX
+j41FQgAjcUgjCiq6gmwI62EsMpZjgjwKhGfYz56jwSmNDgG7Ltj9o/K/+b+oL3u
ZVPO9YgNstWuoPyApE12MzsA+kx/ydRz8GW6IrI5v71JK3yy8Mw/HrYVyUKEnzH4
BVnSRnvHgvCnjZ3SjmUb6Xs74vmoxti9cYr0w0XtBogzuXlfpSYILYFiqrkks6sH
VEmD9P3dTBNQUviYwkRcHhJoI7NqdQBApn9/30F4dQsmnUsikoaom5GovrbBw817
L3Q+U/9+2Xp5k1i7J2MLFSL6Vyaxs/48eRKxs3YubBmrtwBjPj+aOYdRdDOscguQ
WcDRiYPOW72cGIlFtYyuzApqo+J+xw==
=mq0x
-----END PGP SIGNATURE-----
Merge tag 'mlx5-fixes-2019-07-11' of git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux
Saeed Mahameed says:
====================
Mellanox, mlx5 fixes 2019-07-11
This series introduces some fixes to mlx5 driver.
Please pull and let me know if there is any problem.
For -stable v4.15
('net/mlx5e: IPoIB, Add error path in mlx5_rdma_setup_rn')
For -stable v5.1
('net/mlx5e: Fix port tunnel GRE entropy control')
('net/mlx5e: Rx, Fix checksum calculation for new hardware')
('net/mlx5e: Fix return value from timeout recover function')
('net/mlx5e: Fix error flow in tx reporter diagnose')
For -stable v5.2
('net/mlx5: E-Switch, Fix default encap mode')
Conflict note: This pull request will produce a small conflict when
merged with net-next.
In drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c
Take the hunk from net and replace:
esw_offloads_steering_init(esw, vf_nvports, total_nvports);
with:
esw_offloads_steering_init(esw);
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
CQE checksum full mode in new HW, provides a full checksum of rx frame.
Covering bytes starting from eth protocol up to last byte in the received
frame (frame_size - ETH_HLEN), as expected by the stack.
Fixing up skb->csum by the driver is not required in such case. This fix
is to avoid wrong checksum calculation in drivers which already support
the new hardware with the new checksum mode.
Fixes: 85327a9c41 ("net/mlx5: Update the list of the PCI supported devices")
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Add support for transmit side kernel-TLS acceleration.
Offload the crypto encryption to HW.
Per TLS connection:
- Use a separate TIS to maintain the HW context.
- Use a separate encryption key.
- Maintain static and progress HW contexts by posting the proper
WQEs at creation time, or upon resync.
- Use a special DUMP opcode to replay the previous frags and sync
the HW context.
To make sure the SQ is able to serve an xmit request, increase
SQ stop room to cover:
- static params WQE,
- progress params WQE, and
- resync DUMP per frag.
Currently supporting TLS 1.2, and key size 128bit.
Tested over SimX simulator.
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Eran Ben Elisha <eranbe@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Let the EN TIS creation function (mlx5e_create_tis) be responsible
for applying common mdev related fields.
Other specific fields must be set by the caller and passed within
the inbox.
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Reviewed-by: Boris Pismenny <borisp@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use an SQ field for stop_room, and use the larger value only if TLS
is supported.
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Take datapath helper functions to a new header file en/txrx.h.
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Eran Ben Elisha <eranbe@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This series adds mlx5 support for devlink fw versions query.
1) Implement the required low level firmware commands
2) Implement the devlink knobs and callbacks for fw versions query.
-----BEGIN PGP SIGNATURE-----
iQEzBAABCAAdFiEEGhZs6bAKwk/OTgTpSD+KveBX+j4FAl0eZOQACgkQSD+KveBX
+j6a0gf/SbSY2yDz6iyDZy7Zt0e7onD395UyWIoXbXJAI/aMggvdNE233dh369xa
d4Zo8zvFijiFpPKmP0p+WZfvR5YPPWgtOYDv1MbaTz5lqsGYUh2kN9FaTyCNZW3O
jeNidxjsne70VBrpmOZJX2mxduKwRdxchRg2iQPdeZQGO5Uq3nEdOfJVx/mhsIGw
iRczF4IAQpArZEyKXgqMSrtJ99FJ7JSJXnsqjVJyUC1zyCZ+Cq3VV55+u5VCBokd
qr/VSYqFd38SEHR5l4nXDXLX23pOwWNhUdPgGopjak8bN3k19vk6GfQ2bzLFnaOW
4fx1k4uTixKOrbLkgzay/YSsoF2TDw==
=Bwl7
-----END PGP SIGNATURE-----
Merge tag 'mlx5-updates-2019-07-04-v2' of git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux
Saeed Mahameed says:
====================
mlx5-update-2019-07-04
This series adds mlx5 support for devlink fw versions query.
1) Implement the required low level firmware commands
2) Implement the devlink knobs and callbacks for fw versions query.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Misc updates from mlx5-next branch:
1) Add the required HW definitions and structures for upcoming TLS
support.
2) Add support for MCQI and MCQS hardware registers for fw version query.
3) Added hardware bits and structures definitions for sub-functions
4) Small code cleanup and improvement for PF pci driver.
5) Bluefield (ECPF) updates and refactoring for better E-Switch
management on ECPF embedded CPU NIC:
5.1) Consolidate querying eswitch number of VFs
5.2) Register event handler at the correct E-Switch init stage
5.3) Setup PF's inline mode and vlan pop when the ECPF is the
E-Swtich manager ( the host PF is basically a VF ).
5.4) Handle Vport UC address changes in switchdev mode.
6) Cleanup the rep and netdev reference when unloading IB rep.
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
i# All conflicts fixed but you are still merging.
Daniel Borkmann says:
====================
pull-request: bpf-next 2019-07-03
The following pull-request contains BPF updates for your *net-next* tree.
There is a minor merge conflict in mlx5 due to 8960b38932 ("linux/dim:
Rename externally used net_dim members") which has been pulled into your
tree in the meantime, but resolution seems not that bad ... getting current
bpf-next out now before there's coming more on mlx5. ;) I'm Cc'ing Saeed
just so he's aware of the resolution below:
** First conflict in drivers/net/ethernet/mellanox/mlx5/core/en_main.c:
<<<<<<< HEAD
static int mlx5e_open_cq(struct mlx5e_channel *c,
struct dim_cq_moder moder,
struct mlx5e_cq_param *param,
struct mlx5e_cq *cq)
=======
int mlx5e_open_cq(struct mlx5e_channel *c, struct net_dim_cq_moder moder,
struct mlx5e_cq_param *param, struct mlx5e_cq *cq)
>>>>>>> e5a3e259ef
Resolution is to take the second chunk and rename net_dim_cq_moder into
dim_cq_moder. Also the signature for mlx5e_open_cq() in ...
drivers/net/ethernet/mellanox/mlx5/core/en.h +977
... and in mlx5e_open_xsk() ...
drivers/net/ethernet/mellanox/mlx5/core/en/xsk/setup.c +64
... needs the same rename from net_dim_cq_moder into dim_cq_moder.
** Second conflict in drivers/net/ethernet/mellanox/mlx5/core/en_main.c:
<<<<<<< HEAD
int cpu = cpumask_first(mlx5_comp_irq_get_affinity_mask(priv->mdev, ix));
struct dim_cq_moder icocq_moder = {0, 0};
struct net_device *netdev = priv->netdev;
struct mlx5e_channel *c;
unsigned int irq;
=======
struct net_dim_cq_moder icocq_moder = {0, 0};
>>>>>>> e5a3e259ef
Take the second chunk and rename net_dim_cq_moder into dim_cq_moder
as well.
Let me know if you run into any issues. Anyway, the main changes are:
1) Long-awaited AF_XDP support for mlx5e driver, from Maxim.
2) Addition of two new per-cgroup BPF hooks for getsockopt and
setsockopt along with a new sockopt program type which allows more
fine-grained pass/reject settings for containers. Also add a sock_ops
callback that can be selectively enabled on a per-socket basis and is
executed for every RTT to help tracking TCP statistics, both features
from Stanislav.
3) Follow-up fix from loops in precision tracking which was not propagating
precision marks and as a result verifier assumed that some branches were
not taken and therefore wrongly removed as dead code, from Alexei.
4) Fix BPF cgroup release synchronization race which could lead to a
double-free if a leaf's cgroup_bpf object is released and a new BPF
program is attached to the one of ancestor cgroups in parallel, from Roman.
5) Support for bulking XDP_TX on veth devices which improves performance
in some cases by around 9%, from Toshiaki.
6) Allow for lookups into BPF devmap and improve feedback when calling into
bpf_redirect_map() as lookup is now performed right away in the helper
itself, from Toke.
7) Add support for fq's Earliest Departure Time to the Host Bandwidth
Manager (HBM) sample BPF program, from Lawrence.
8) Various cleanups and minor fixes all over the place from many others.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Report EQE data upon CQ completion to let upper layers use this data.
Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Acked-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Refreshing TIRs is done in order to update the TIRs with the current
state of SQs in the transport domain, so that the TIRs can filter out
undesired self-loopback packets based on the source SQ of the packet.
Representor TIRs will only receive packets that originate from their
associated vport, due to dedicated steering, and therefore will never
receive self-loopback packets, whose source vport will be the vport of
the E-Switch manager, and therefore not the vport associated with the
representor. As such, it is not necessary to refresh the representors'
TIRs, since self-loopback packets can't reach them.
Since representors only exist in switchdev mode, and there is no
scenario in which a representor will exist in the transport domain
alongside a non-representor, it is not necessary to refresh the
transport domain's TIRs upon changing the state of a representor's
queues. Therefore, do not refresh TIRs upon such a change. Achieve
this by adding an update_rx callback to the mlx5e_profile, which
refreshes TIRs for non-representors and does nothing for representors,
and replace instances of mlx5e_refresh_tirs() upon changing the state
of the queues with update_rx().
Signed-off-by: Gavi Teitz <gavi@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Reviewed-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>
Put the XDP SQ that is used for XDP_TX into the channel. It used to be a
part of the RQ, but with introduction of AF_XDP there will be one more
RQ that could share the same XDP SQ. This patch is a preparation for
that change.
Separate XDP_TX statistics per RQ were implemented in one of the previous
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>
Currently, struct mlx5e_xdp_info has some issues that have to be cleaned
up before the upcoming AF_XDP support makes things too complicated and
messy. This structure is used both when sending the packet and on
completion. Moreover, the cleanup procedure on completion depends on the
origin of the packet (XDP_REDIRECT, XDP_TX). Adding AF_XDP support will
add new flows that use this structure even differently. To avoid
overcomplicating the code, this commit refactors the usage of this
structure in the following ways:
1. struct mlx5e_xdp_info is split into two different structures. One is
struct mlx5e_xdp_xmit_data, a transient structure that doesn't need to
be stored and is only used while sending the packet. The other is still
struct mlx5e_xdp_info that is stored in a FIFO and contains the fields
needed on completion.
2. The fields of struct mlx5e_xdp_info that are used in different flows
are put into a union. A special enum indicates the cleanup mode and
helps choose the right union member. This approach is clear and
explicit. Although it could be possible to "guess" the mode by looking
at the values of the fields and at the XDP SQ type, it wouldn't be that
clear and extendable and would require looking through the whole chain
to understand what's going on.
For the reference, there are the fields of struct mlx5e_xdp_info that
are used in different flows (including AF_XDP ones):
Packet origin | Fields used on completion | Cleanup steps
-----------------------+---------------------------+------------------
XDP_REDIRECT, | xdpf, dma_addr | DMA unmap and
XDP_TX from XSK RQ | | xdp_return_frame.
-----------------------+---------------------------+------------------
XDP_TX from regular RQ | di | Recycle page.
-----------------------+---------------------------+------------------
AF_XDP TX | (none) | Increment the
| | producer index in
| | Completion Ring.
On send, the same set of mlx5e_xdp_xmit_data fields is used in all
flows: DMA and virtual addresses and length.
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>
Prepare to creation of the XSK RQ, which will require posting UMRs, too.
The same ICO SQ will be used for both RQs and also to trigger interrupts
by posting NOPs. UMR WQEs can't be reused any more. Optimization
introduced in commit ab966d7e4f ("net/mlx5e: RX, Recycle buffer of
UMR WQEs") is reverted.
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>
From: Tal Gilboa and Yamin Fridman
Implement net DIM over a generic DIM library, add RDMA DIM
dim.h lib exposes an implementation of the DIM algorithm for
dynamically-tuned interrupt moderation for networking interfaces.
We want a similar functionality for other protocols, which might need to
optimize interrupts differently. Main motivation here is DIM for NVMf
storage protocol.
Current DIM implementation prioritizes reducing interrupt overhead over
latency. Also, in order to reduce DIM's own overhead, the algorithm might
take some time to identify it needs to change profiles. While this is
acceptable for networking, it might not work well on other scenarios.
Here we propose a new structure to DIM. The idea is to allow a slightly
modified functionality without the risk of breaking Net DIM behavior for
netdev. We verified there are no degradations in current DIM behavior with
the modified solution.
Suggested solution:
- Common logic is implemented in lib/dim/dim.c
- Net DIM (existing) logic is implemented in lib/dim/net_dim.c, which uses
the common logic in dim.c
- Any new DIM logic will be implemented in "lib/dim/new_dim.c".
This new implementation will expose modified versions of profiles,
dim_step() and dim_decision().
- DIM API is declared in include/linux/dim.h for all implementations.
Pros for this solution are:
- Zero impact on existing net_dim implementation and usage
- Relatively more code reuse (compared to two separate solutions)
- Increased extensibility
-----BEGIN PGP SIGNATURE-----
iQEzBAABCAAdFiEEGhZs6bAKwk/OTgTpSD+KveBX+j4FAl0SiDAACgkQSD+KveBX
+j4mlQgAoabbtxwIapMg62tKhnlr/3hEEyuEjiQRhzMUb5y5zOIJNRL82Nerv4K3
oD7v76emw/Xb1ErPe4yJJRdkpbpabEAFKbqkM0R7xFPNK33ZbphlU1d9YGgUKp1W
Uieydn5kfubzi6p7R1TC7c2qW1Wr6rE1wnNJyC/sy8zFEP/OEH0NmgPDyOs7ckWI
xaoJ1lhNWMoGN+Qk+fxazd1VtsKdcbc9JvLD2e8ON5qWdnuNiLhawcJVoWzfsCAa
V/Tdff+Rj9bVFsi468u20og4EdV3ceAOTzMnhCYpAtYZHVbVk6MFgh87uQYkTrfY
sW8BDmWKkRdHtzHd3a3yQiVDnz/iuw==
=bAZn
-----END PGP SIGNATURE-----
Merge tag 'blk-dim-v2' of git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux
Saeed Mamameed says:
====================
Generic DIM
From: Tal Gilboa and Yamin Fridman
Implement net DIM over a generic DIM library, add RDMA DIM
dim.h lib exposes an implementation of the DIM algorithm for
dynamically-tuned interrupt moderation for networking interfaces.
We want a similar functionality for other protocols, which might need to
optimize interrupts differently. Main motivation here is DIM for NVMf
storage protocol.
Current DIM implementation prioritizes reducing interrupt overhead over
latency. Also, in order to reduce DIM's own overhead, the algorithm might
take some time to identify it needs to change profiles. While this is
acceptable for networking, it might not work well on other scenarios.
Here we propose a new structure to DIM. The idea is to allow a slightly
modified functionality without the risk of breaking Net DIM behavior for
netdev. We verified there are no degradations in current DIM behavior with
the modified solution.
Suggested solution:
- Common logic is implemented in lib/dim/dim.c
- Net DIM (existing) logic is implemented in lib/dim/net_dim.c, which uses
the common logic in dim.c
- Any new DIM logic will be implemented in "lib/dim/new_dim.c".
This new implementation will expose modified versions of profiles,
dim_step() and dim_decision().
- DIM API is declared in include/linux/dim.h for all implementations.
Pros for this solution are:
- Zero impact on existing net_dim implementation and usage
- Relatively more code reuse (compared to two separate solutions)
- Increased extensibility
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Moved all logic from dim.h and net_dim.h to dim.c and net_dim.c.
This is both more structurally appealing and would allow to only
expose externally used functions.
Signed-off-by: Tal Gilboa <talgi@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Removed 'net' prefix from functions and structs used by external drivers.
Signed-off-by: Tal Gilboa <talgi@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
The TX queue index returned by the fallback function ranges
between [0,NUM CHANNELS - 1] if QoS isn't set and
[0, (NUM CHANNELS)*(NUM TCs) -1] otherwise.
Our HW uses different TC mapping than the fallback function
(which is denoted as 'up', user priority) so we only need to extract
a channel number out of the returned value.
Since (NUM CHANNELS)*(NUM TCs) is a relatively small number, using
reciprocal scale almost always returns zero.
We instead access the 'txq2sq' table to extract the sq (and with it the
channel number) associated with the tx queue, thus getting
a more evenly distributed channel number.
Perf:
Rx/Tx side with Intel(R) Xeon(R) Silver 4108 CPU @ 1.80GHz and ConnectX-5.
Used 'iperf' UDP traffic, 10 threads, and priority 5.
Before: 0.566Mpps
After: 2.37Mpps
As expected, releasing the existing bottleneck of steering all traffic
to TX queue zero significantly improves transmission rates.
Fixes: 7ccdd0841b ("net/mlx5e: Fix select queue callback")
Signed-off-by: Shay Agroskin <shayag@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
After we have a dedicated uplink representor, the new netdev ops
doesn't support ndo_set_feature. Because of that, we can't change
some features, eg. rxvlan. Now add it back.
In this patch, I also do a cleanup for the features flag handling,
eg. remove duplicate NETIF_F_HW_TC flag setting.
Fixes: aec002f6f8 ("net/mlx5e: Uninstantiate esw manager vport netdev on switchdev mode")
Signed-off-by: Chris Mi <chrism@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Reviewed-by: Vlad Buslov <vladbu@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Benefit from the devlink flash update implementation and ethtool
fallback to it and move firmware flash implementation there.
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Acked-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Hardware requires that all TIRs that steer traffic to the same RQ
should share identical tunneled_offload_en value.
For that, the tunneled_offload_en bit should be set/unset (according to
the HW capability) for all TIRs', not only the ones dedicated for
tunneled (inner) traffic.
Fixes: 1b223dd391 ("net/mlx5e: Fix checksum handling for non-stripped vlan packets")
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
mlx5e_trigger_irq posts a NOP to the ICO SQ just to trigger an IRQ and
enter the NAPI poll on the right CPU according to the affinity. Use it
in mlx5e_activate_rq.
Signed-off-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Upon high packet rate with multiple CPUs TX workloads, much of the HCA's
resources are spent on prefetching TX descriptors, thus affecting
transmission rates.
This patch comes to mitigate this problem by moving some workload to the
CPU and reducing the HW data prefetch overhead for small packets (<= 256B).
When forwarding packets with XDP, a packet that is smaller
than a certain size (set to ~256 bytes) would be sent inline within
its WQE TX descrptor (mem-copied), when the hardware tx queue is congested
beyond a pre-defined water-mark.
This is added to better utilize the HW resources (which now makes
one less packet data prefetch) and allow better scalability, on the
account of CPU usage (which now 'memcpy's the packet into the WQE).
To load balance between HW and CPU and get max packet rate, we use
watermarks to detect how much the HW is congested and move the work
loads back and forth between HW and CPU.
Performance:
Tested packet rate for UDP 64Byte multi-stream
over two dual port ConnectX-5 100Gbps NICs.
CPU: Intel(R) Xeon(R) CPU E5-2680 v3 @ 2.50GHz
* Tested with hyper-threading disabled
XDP_TX:
| | before | after | |
| 24 rings | 51Mpps | 116Mpps | +126% |
| 1 ring | 12Mpps | 12Mpps | same |
XDP_REDIRECT:
** Below is the transmit rate, not the redirection rate
which might be larger, and is not affected by this patch.
| | before | after | |
| 32 rings | 64Mpps | 92Mpps | +43% |
| 1 ring | 6.4Mpps | 6.4Mpps | same |
As we can see, feature significantly improves scaling, without
hurting single ring performance.
Signed-off-by: Shay Agroskin <shayag@mellanox.com>
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
The XDP redirect flush indication belongs to the receive queue,
not to its XDP send queue.
For this, use a new bit on rq->flags.
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Reviewed-by: Shay Agroskin <shayag@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Values in enum mlx5e_rq_flag are used as bit indixes.
Intention was to use them with no BIT(i) wrapping.
No functional bug fix here, as the same (shifted)flag bit
is used for all set, test, and clear operations.
Fixes: 121e892754 ("net/mlx5e: Refactor RQ XDP_TX indication")
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Reviewed-by: Shay Agroskin <shayag@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
The buffers mapping of the Multi-Packet WQEs (of Striding RQ)
is done via UMR posts, one UMR WQE per an RX MPWQE.
A single MPWQE is capable of serving many incoming packets,
usually larger than the budget of a single napi cycle.
Hence, posting a single UMR WQE per napi cycle (and handling its
completion in the next cycle) works fine in many common cases,
but not always.
When an XDP program is loaded, every MPWQE is capable of serving less
packets, to satisfy the packet-per-page requirement.
Thus, for the same number of packets more MPWQEs (and UMR posts)
are needed (twice as much for the default MTU), giving less latency
room for the UMR completions.
In this patch, we add support for multiple outstanding UMR posts,
to allow faster gap closure between consuming MPWQEs and reposting
them back into the WQ.
For better SW and HW locality, we combine the UMR posts in bulks of
(at least) two.
This is expected to improve packet rate in high CPU scale.
Performance test:
As expected, huge improvement in large-scale (48 cores).
xdp_redirect_map, 64B UDP multi-stream.
Redirect from ConnectX-5 100Gbps to ConnectX-6 100Gbps.
CPU: Intel(R) Xeon(R) CPU E5-2680 v3 @ 2.50GHz.
Before: Unstable, 7 to 30 Mpps
After: Stable, at 70.5 Mpps
No degradation in other tested scenarios.
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
When requested to recover from error, the tx reporter might open new
channels and close the existing ones. Use safe channels switch flow in
order to guarantee opened channels at the end of the recover flow.
For this purpose, define mlx5e_safe_reopen_channels function and use it
within those flows.
Fixes: de8650a820 ("net/mlx5e: Add tx reporter support")
Signed-off-by: Eran Ben Elisha <eranbe@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Expose a new helper that wraps the logic for setting the
netdevice's MTU boundaries.
Use it for the different components (Eth, rep, IPoIB).
Set the netdevice min MTU to ETH_MIN_MTU, and the max according
to both the FW capability and the kernel definition.
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
The LRO WQE size is a constant, obsolete the parameter field
that holds it.
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Reviewed-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
This merge commit includes some misc shared code updates from mlx5-next branch needed
for net-next.
1) From Maxim, Remove un-used macros and spinlock from mlx5 code.
2) From Aya, Expose Management PCIE info register layout and add rate limit
print macros.
3) From Tariq, Compilation warning fix in fs_core.c
4) From Vu, Huy and Saeed, Improve mlx5 initialization flow:
The goal is to provide a better logical separation of mlx5 core
device initialization flow and will help to seamlessly support
creating different mlx5 device types such as PF, VF and SF
mlx5 sub-function virtual devices.
Mlx5_core driver needs to separate HCA resources from pci resources.
Its initialize/load/unload will be broken into stages:
1. Initialize common data structures
2. Setup function which initializes pci resources (for PF/VF)
or some other specific resources for virtual device
3. Initialize software objects according to hardware capabilities
4. Load all mlx5_core components
It is also necessary to detach mlx5_core mdev name/message from pci
device mdev->pdev name/message for a clearer report/debug of
different mlx5 device types.
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
As there is no user of mlx5_write64 that passes a spinlock to
mlx5_write64, remove this functionality and simplify the function.
Signed-off-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Reviewed-by: Eran Ben Elisha <eranbe@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Tariq Toukan