Add MLXSW_REG_MPAT_SPAN_TYPE_REMOTE_ETH to support VLAN-encapsulated
port mirroring.
Signed-off-by: Petr Machata <petrm@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Extend the boolean interpretation of msi_x module parameter
to numerical, as follows:
0 - Don't use MSI-X.
1 - Use MSI-X, driver decides the num of MSI-X irqs.
>=2 - Use MSI-X, limit number of MSI-X irqs to msi_x.
In SRIOV, this limits the number of MSI-X irqs per VF.
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Cc: Ajaykumar Hotchandani <ajaykumar.hotchandani@oracle.com>
Reviewed-by: Ajaykumar Hotchandani <ajaykumar.hotchandani@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Implement suspend/resume callbacks in struct pci_driver.
Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If supported, write a driver version string to FW as part of the
INIT_HCA command.
Example of driver version: "Linux,mlx4_core,4.0-0"
Signed-off-by: Eran Ben Elisha <eranbe@mellanox.com>
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
mlx5 core driver misc cleanups and updates:
- fix spelling mistake: "modfiy" -> "modify"
- Cleanup unused field in Work Queue parameters
- dump_command mailbox length printed
- Refactor num of blocks in mailbox calculation
- Decrease level of prints about non-existent MKEY
- remove some extraneous spaces in indentations
-----BEGIN PGP SIGNATURE-----
iQEcBAABAgAGBQJa8Lh4AAoJEEg/ir3gV/o+0gwH/1nJwrnIfFkFgjgN6TzNFhIW
P2P1FwkuwiAiOnrJOr+r1VX17BSSsjZ9cBxd2jRZOHqVbrLfhS5vkY5Lul9wWOdu
u8h8JvL1qXCAqilxH+SLeKyZYjDVTIDreUIOFXuuthcz+HcF41NbvzM/9WsVj9kQ
cx8ineozMrx85PxSxJIKTvsNpaJc0A0GR1+tKKn3R8djgID8BpCAgXjW3XCJ/Ou9
CoPscduksdLJ6enjqI0JZPZljgi+l4D33819Mqz1yeGpCLFATRaEXe6rLQO0U+zD
n1vSZytgBvdpmADUdnCLL9BssWISIjZQzzCIjQxoSERTKFFKMtTCGRMqJXgOS7Y=
=2Lzw
-----END PGP SIGNATURE-----
Merge tag 'mlx5-updates-2018-05-07' of git://git.kernel.org/pub/scm/linux/kernel/git/mellanox/linux
Saeed Mahameed says:
====================
mlx5-updates-2018-05-07
mlx5 core driver misc cleanups and updates:
- fix spelling mistake: "modfiy" -> "modify"
- Cleanup unused field in Work Queue parameters
- dump_command mailbox length printed
- Refactor num of blocks in mailbox calculation
- Decrease level of prints about non-existent MKEY
- remove some extraneous spaces in indentations
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Trivial fix to spelling mistake in netdev_warn warning message
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Remove the 'linear' field from struct mlx5_wq_param.
It is redundant, set but never read.
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Dump command mailbox length printed was correct only if data_only flag
was set. For the case that data_only flag was clear the offset to stop
printing at was wrong and so the buffer printed was too short.
Changed the print loop to stop according to number of buffers in
mailbox.
Fixes: e126ba97db ("mlx5: Add driver for Mellanox Connect-IB adapters")
Signed-off-by: Moshe Shemesh <moshe@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Get the logic that calculates the number of blocks in a command mailbox
into a dedicated function.
Signed-off-by: Moshe Shemesh <moshe@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
User-controlled application can cause multiple prints as below to flood
dmesg. Since knowledge of failed MKey release is important for debug,
let's decrease its level to debug.
mlx5_core 0000:00:04.0: mlx5_core_destroy_mkey:127:(pid 2352): failed
radix tree delete of mkey 0x1ed700
Reported-by: Noa Osherovich <noaos@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
While trying to support CHECKSUM_COMPLETE for IPV6 fragments,
I had to experiments various hacks in get_fixed_ipv6_csum().
I must admit I could not find how to implement this :/
However, get_fixed_ipv6_csum() does a lot of redundant operations,
calling csum_partial() twice.
First csum_partial() computes the checksum of saddr and daddr,
put in @csum_pseudo_hdr. Undone later in the second csum_partial()
computed on whole ipv6 header.
Then nexthdr is added once, added a second time, then substracted.
payload_len is added once, then substracted.
Really all this can be reduced to two add_csum(), to add back 6 bytes
that were removed by mlx4 when providing hw_checksum in RX descriptor.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Saeed Mahameed <saeedm@mellanox.com>
Cc: Tariq Toukan <tariqt@mellanox.com>
Reviewed-by: Saeed Mahameed <saeedm@mellanox.com>
Acked-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The following patch enables sending notifications also for events on FDB
entries that weren't added by the user. Give the drivers the information
necessary to distinguish between the two origins of FDB entries.
To maintain the current behavior, have switchdev-implementing drivers
bail out on notifications about non-user-added FDB entries. In case of
mlxsw driver, allow a call to mlxsw_sp_span_respin() so that SPAN over
bridge catches up with the changed FDB.
Signed-off-by: Petr Machata <petrm@mellanox.com>
Reviewed-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Acked-by: Ivan Vecera <ivecera@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Check number of CQEs for CQE version 2 reported by QUERY_AQ_CAP command.
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use previously added resources to query FW support for multiple versions
of CQEs. Use the biggest version supported. For SDQs, it has no sense to
use version 2 as it does not introduce any new features, but it is
twice the size of CQE version 1.
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Introduce definitions of fields in CQE version 1 and 2. Also, introduce
common helpers that would call appropriate version-specific helpers
according to the version enum passed.
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add resources that FW uses to report supported CQE versions.
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Trivial fix to spelling mistake in netdev_err error message
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We currently do not perform accounting in the driver and thus can't
reject routes before resources are exceeded.
However, in order to make users aware of the fact that routes are no
longer offloaded we can return an error for routes configured after the
abort mechanism was triggered.
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Acked-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Since commit 9776d32537 ("net: Move call_fib_rule_notifiers up in
fib_nl_newrule") it is possible to forbid the installation of
unsupported FIB rules.
Have mlxsw return an error for non-default FIB rules in addition to the
existing extack message.
Example:
# ip rule add from 198.51.100.1 table 10
Error: mlxsw_spectrum: FIB rules not supported.
Note that offload is only aborted when non-default FIB rules are already
installed and merely replayed during module initialization.
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Acked-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
trivial fix to spelling mistake in mlx4_warn message.
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add statistics for rare TLS related errors.
Since the errors are rare we have a counter per netdev
rather then per SQ.
Signed-off-by: Ilya Lesokhin <ilyal@mellanox.com>
Signed-off-by: Boris Pismenny <borisp@mellanox.com>
Acked-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Implement the TLS tx offload data path according to the
requirements of the TLS generic NIC offload infrastructure.
Special metadata ethertype is used to pass information to
the hardware.
Signed-off-by: Ilya Lesokhin <ilyal@mellanox.com>
Signed-off-by: Boris Pismenny <borisp@mellanox.com>
Acked-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add NETIF_F_HW_TLS_TX capability and expose tlsdev_ops to work with the
TLS generic NIC offload infrastructure.
The NETIF_F_HW_TLS_TX capability will be added in the next patch.
Signed-off-by: Ilya Lesokhin <ilyal@mellanox.com>
Signed-off-by: Boris Pismenny <borisp@mellanox.com>
Acked-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add routines for manipulating TLS TX offload contexts.
In Innova TLS, TLS contexts are added or deleted
via a command message over the SBU connection.
The HW then sends a response message over the same connection.
Add implementation for Innova TLS (FPGA-based) hardware.
These routines will be used by the TLS offload support in a later patch
mlx5/accel is a middle acceleration layer to allow mlx5e and other ULPs
to work directly with mlx5_core rather than Innova FPGA or other mlx5
acceleration providers.
In the future, when IPSec/TLS or any other acceleration gets integrated
into ConnectX chip, mlx5/accel layer will provide the integrated
acceleration, rather than the Innova one.
Signed-off-by: Ilya Lesokhin <ilyal@mellanox.com>
Signed-off-by: Boris Pismenny <borisp@mellanox.com>
Acked-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The defines are not IPSEC specific.
Signed-off-by: Ilya Lesokhin <ilyal@mellanox.com>
Signed-off-by: Boris Pismenny <borisp@mellanox.com>
Acked-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When handling mirroring to a gretap or ip6gretap netdevice in mlxsw, the
underlay address (i.e. the remote address of the tunnel) may be routed
to a bridge.
In that case, look up the resolved neighbor Ethernet address in that
bridge's FDB. Then configure the offload to direct the mirrored traffic
to that port, possibly with tagging.
Signed-off-by: Petr Machata <petrm@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Reviewed-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Changes to switchdev artifact can make a SPAN entry offloadable or
unoffloadable. To that end:
- Listen to SWITCHDEV_FDB_*_TO_BRIDGE notifications in addition to
the *_TO_DEVICE ones, to catch whatever activity is sent to the
bridge (likely by mlxsw itself).
On each FDB notification, respin SPAN to reconcile it with the FDB
changes.
- Also respin on switchdev port attribute changes (which currently
covers changes to STP state of ports) and port object additions and
removals.
Signed-off-by: Petr Machata <petrm@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Since switchdev events can trigger SPAN respin, it is necessary that the
data structures are available. Register SPAN first, with a commentary on
what the dependencies are.
Signed-off-by: Petr Machata <petrm@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Publish the existing function mlxsw_sp_bridge_port_find(), and add
another service accessor mlxsw_sp_bridge_port_stp_state(). Publish both
in a new file spectrum_switchdev.h.
Signed-off-by: Petr Machata <petrm@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Instead of duplicating the decision regarding port forwarding state made
by mlxsw_sp_port_vid_stp_set(), extract the decision-making into a new
function and reuse.
Signed-off-by: Petr Machata <petrm@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The code in the fallback path has supported XDP in conjunction with the Tx
traffic classification for TCs for over a year now. So instead of just
calling skb_tx_hash for every packet we are better off using the fallback
since that will record the Tx queue to the socket and then that can be used
instead of having to recompute the hash every time.
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
-----BEGIN PGP SIGNATURE-----
iQEcBAABAgAGBQJa4ixLAAoJEEg/ir3gV/o+PB0IAKlexiYEyID3N3cvG/I0m4Zh
yxIN4H8sUu3kz7dG6dkvpL5Oo5c43/huE+tspOYfFmydMWvOV+DakwqAhkE+KUfe
VIpF0cM7+T8Els8e7OzuT0Zu5ggeN1wU0uPRhAn1F592BH0ppXGBn8WIKTffYb8c
5XeB2JSZyw4yMgk1zurm4tVtFvHbYO7SAkLZZG5E0m7EeIujVWi1lTnhXNi9zdm/
48LoLZ/1Rmx0e/Qpey2fm9HEPRPTgCNSBLEsx2hIDiJG56YyWPH6+N7U9Acf2PaI
lKu2JMqLYe3du8hhPtCbPYH0i74af/LbNCQgXXgPstAI49v+MydlBcv2NX9J6NY=
=nw/Q
-----END PGP SIGNATURE-----
Merge tag 'mlx5-fixes-2018-04-25' of git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux
Saeed Mahameed says:
====================
Mellanox, mlx5 fixes 2018-04-26
This pull request includes fixes for mlx5 core and netdev driver.
Please pull and let me know if there's any problems.
For -stable v4.12
net/mlx5e: TX, Use correct counter in dma_map error flow
For -stable v4.13
net/mlx5: Avoid cleaning flow steering table twice during error flow
For -stable v4.14
net/mlx5e: Allow offloading ipv4 header re-write for icmp
For -stable v4.15
net/mlx5e: DCBNL fix min inline header size for dscp
For -stable v4.16
net/mlx5: Fix mlx5_get_vector_affinity function
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
When IGMP snooping is enabled on a bridge, traffic forwarded by an MDB
entry should be sent to both ports member in the MDB's ports list and
mrouter ports.
In case a port needs to be removed from an MDB's ports list, but this
port is also configured as an mrouter port, then do not update the
device so that it will continue to forward traffic through that port.
Fix a copy-paste error that checked that IGMP snooping is enabled twice
instead of checking the port's mrouter state.
Fixes: ded711c87a ("mlxsw: spectrum_switchdev: Consider mrouter status for mdb changes")
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Reported-by: Colin King <colin.king@canonical.com>
Reviewed-by: Nogah Frankel <nogahf@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When deleting a flow counter, the modify mask should be the action and
the flow counter. Otherwise the flow counter is not deleted and we'll
get a firmware warning when deleting the remaining destinations on the
same FTE.
It only happens in the presence of flow counter and multiple vport
destinations. If there is only one vport destination, there is no
need to update the FTE when deleting the only vport destination,
we just delete the FTE.
Fixes: ae05831424 ("net/mlx5: Add option to add fwd rule with counter")
Signed-off-by: Chris Mi <chrism@mellanox.com>
Signed-off-by: Jianbo Liu <jianbol@mellanox.com>
Reviewed-by: Mark Bloch <markb@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
After the cited commit, WQE RQ size is calculated based on sw_mtu but it
was not set for representors. This commit fixes that.
Fixes: 472a1e44b3 ("net/mlx5e: Save MTU in channels params")
Signed-off-by: Shahar Klein <shahark@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
When we fail to initialize the RX root namespace, we need
to clean only that and not the entire flow steering.
Currently the code may try to clean the flow steering twice
on error witch leads to null pointer deference.
Make sure we clean correctly.
Fixes: fba53f7b57 ("net/mlx5: Introduce mlx5_flow_steering structure")
Signed-off-by: Talat Batheesh <talatb@mellanox.com>
Reviewed-by: Mark Bloch <markb@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
In case of a dma_mapping_error, do not use wi->num_dma
as a parameter for dma unmap function because it's yet
to be set, and holds an out-of-date value.
Use actual value (local variable num_dma) instead.
Fixes: 34802a42b3 ("net/mlx5e: Do not modify the TX SKB")
Fixes: e586b3b0ba ("net/mlx5: Ethernet Datapath files")
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
When the trust state is set to dscp and the netdev is down, the inline
header size is not updated. When netdev is up, the inline header size
stays at L2 instead of IP.
Fix this issue by updating the private parameter when the netdev is in
down so that when netdev is up, it picks up the right header size.
Fixes: fbcb127e89 ("net/mlx5e: Support DSCP trust state ...")
Signed-off-by: Huy Nguyen <huyn@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
For ICMPv4, the checksum is calculated from the ICMP headers and data.
Since the ICMPv4 checksum doesn't cover the IP header, we can allow to
do L3 header re-write for this protocol.
Fixes: bdd66ac0ae ('net/mlx5e: Disallow TC offloading of unsupported match/action combinations')
Signed-off-by: Jianbo Liu <jianbol@mellanox.com>
Reviewed-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Add support for adaptive TX moderation. This greatly reduces TX interrupt
rate and increases bandwidth, mostly for TCP bandwidth over ARM
architecture (below). There is a slight single stream TCP with very large
message sizes degradation (x86). In this case if there's any moderation on
transmitted packets the bandwidth would reduce due to hitting TCP output limit.
Since this is a synthetic case, this is still worth doing.
Performance improvement (ConnectX-4Lx 40GbE, ARM)
TCP 64B bandwidth with 1-50 streams increased 6-35%.
TCP 64B bandwidth with 100-500 streams increased 20-70%.
Performance improvement (ConnectX-5 100GbE, x86)
Bandwidth: increased up to 40% (1024B with 10s of streams).
Interrupt rate: reduced up to 50% (1024B with 1000s of streams).
Performance degradation (ConnectX-5 100GbE, x86)
Bandwidth: up to 10% decrease single stream TCP (1MB message size from
51Gb/s to 47Gb/s).
Signed-off-by: Tal Gilboa <talgi@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Acked-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Preparation for introducing adaptive TX to net DIM.
Signed-off-by: Tal Gilboa <talgi@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Daniel Borkmann says:
====================
pull-request: bpf-next 2018-04-21
The following pull-request contains BPF updates for your *net-next* tree.
The main changes are:
1) Initial work on BPF Type Format (BTF) is added, which is a meta
data format which describes the data types of BPF programs / maps.
BTF has its roots from CTF (Compact C-Type format) with a number
of changes to it. First use case is to provide a generic pretty
print capability for BPF maps inspection, later work will also
add BTF to bpftool. pahole support to convert dwarf to BTF will
be upstreamed as well (https://github.com/iamkafai/pahole/tree/btf),
from Martin.
2) Add a new xdp_bpf_adjust_tail() BPF helper for XDP that allows
for changing the data_end pointer. Only shrinking is currently
supported which helps for crafting ICMP control messages. Minor
changes in drivers have been added where needed so they recalc
the packet's length also when data_end was adjusted, from Nikita.
3) Improve bpftool to make it easier to feed hex bytes via cmdline
for map operations, from Quentin.
4) Add support for various missing BPF prog types and attach types
that have been added to kernel recently but neither to bpftool
nor libbpf yet. Doc and bash completion updates have been added
as well for bpftool, from Andrey.
5) Proper fix for avoiding to leak info stored in frame data on page
reuse for the two bpf_xdp_adjust_{head,meta} helpers by disallowing
to move the pointers into struct xdp_frame area, from Jesper.
6) Follow-up compile fix from BTF in order to include stdbool.h in
libbpf, from Björn.
7) Few fixes in BPF sample code, that is, a typo on the netdevice
in a comment and fixup proper dump of XDP action code in the
tracepoint exception, from Wang and Jesper.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Change the prefix for fib6_info struct elements from rt6i_ to fib6_.
rt6i_pcpu and rt6i_exception_bucket are left as is given that they
point to rt6_info entries.
Rename only; not functional change intended.
Signed-off-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
w/ bpf_xdp_adjust_tail helper xdp's data_end pointer could be changed as
well (only "decrease" of pointer's location is going to be supported).
changing of this pointer will change packet's size.
for mlx4 driver we will just calculate packet's length unconditionally
(the same way as it's already being done in mlx5)
Acked-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Nikita V. Shirokov <tehnerd@tehnerd.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Convert all code paths referencing a FIB entry from
rt6_info to fib6_info.
Signed-off-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Introduce fib6_nh structure and move nexthop related data from
rt6_info and rt6_info.dst to fib6_nh. References to dev, gateway or
lwtstate from a FIB lookup perspective are converted to use fib6_nh;
datapath references to dst version are left as is.
Signed-off-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Changing API xdp_return_frame() to take struct xdp_frame as argument,
seems like a natural choice. But there are some subtle performance
details here that needs extra care, which is a deliberate choice.
When de-referencing xdp_frame on a remote CPU during DMA-TX
completion, result in the cache-line is change to "Shared"
state. Later when the page is reused for RX, then this xdp_frame
cache-line is written, which change the state to "Modified".
This situation already happens (naturally) for, virtio_net, tun and
cpumap as the xdp_frame pointer is the queued object. In tun and
cpumap, the ptr_ring is used for efficiently transferring cache-lines
(with pointers) between CPUs. Thus, the only option is to
de-referencing xdp_frame.
It is only the ixgbe driver that had an optimization, in which it can
avoid doing the de-reference of xdp_frame. The driver already have
TX-ring queue, which (in case of remote DMA-TX completion) have to be
transferred between CPUs anyhow. In this data area, we stored a
struct xdp_mem_info and a data pointer, which allowed us to avoid
de-referencing xdp_frame.
To compensate for this, a prefetchw is used for telling the cache
coherency protocol about our access pattern. My benchmarks show that
this prefetchw is enough to compensate the ixgbe driver.
V7: Adjust for commit d9314c474d ("i40e: add support for XDP_REDIRECT")
V8: Adjust for commit bd658dda42 ("net/mlx5e: Separate dma base address
and offset in dma_sync call")
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch shows how it is possible to have both the driver local page
cache, which uses elevated refcnt for "catching"/avoiding SKB
put_page returns the page through the page allocator. And at the
same time, have pages getting returned to the page_pool from
ndp_xdp_xmit DMA completion.
The performance improvement for XDP_REDIRECT in this patch is really
good. Especially considering that (currently) the xdp_return_frame
API and page_pool_put_page() does per frame operations of both
rhashtable ID-lookup and locked return into (page_pool) ptr_ring.
(It is the plan to remove these per frame operation in a followup
patchset).
The benchmark performed was RX on mlx5 and XDP_REDIRECT out ixgbe,
with xdp_redirect_map (using devmap) . And the target/maximum
capability of ixgbe is 13Mpps (on this HW setup).
Before this patch for mlx5, XDP redirected frames were returned via
the page allocator. The single flow performance was 6Mpps, and if I
started two flows the collective performance drop to 4Mpps, because we
hit the page allocator lock (further negative scaling occurs).
Two test scenarios need to be covered, for xdp_return_frame API, which
is DMA-TX completion running on same-CPU or cross-CPU free/return.
Results were same-CPU=10Mpps, and cross-CPU=12Mpps. This is very
close to our 13Mpps max target.
The reason max target isn't reached in cross-CPU test, is likely due
to RX-ring DMA unmap/map overhead (which doesn't occur in ixgbe to
ixgbe testing). It is also planned to remove this unnecessary DMA
unmap in a later patchset
V2: Adjustments requested by Tariq
- Changed page_pool_create return codes not return NULL, only
ERR_PTR, as this simplifies err handling in drivers.
- Save a branch in mlx5e_page_release
- Correct page_pool size calc for MLX5_WQ_TYPE_LINKED_LIST_STRIDING_RQ
V5: Updated patch desc
V8: Adjust for b0cedc844c ("net/mlx5e: Remove rq_headroom field from params")
V9:
- Adjust for 121e892754 ("net/mlx5e: Refactor RQ XDP_TX indication")
- Adjust for 73281b78a3 ("net/mlx5e: Derive Striding RQ size from MTU")
- Correct handling if page_pool_create fail for MLX5_WQ_TYPE_LINKED_LIST_STRIDING_RQ
V10: Req from Tariq
- Change pool_size calc for MLX5_WQ_TYPE_LINKED_LIST_STRIDING_RQ
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Acked-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Need a fast page recycle mechanism for ndo_xdp_xmit API for returning
pages on DMA-TX completion time, which have good cross CPU
performance, given DMA-TX completion time can happen on a remote CPU.
Refurbish my page_pool code, that was presented[1] at MM-summit 2016.
Adapted page_pool code to not depend the page allocator and
integration into struct page. The DMA mapping feature is kept,
even-though it will not be activated/used in this patchset.
[1] http://people.netfilter.org/hawk/presentations/MM-summit2016/generic_page_pool_mm_summit2016.pdf
V2: Adjustments requested by Tariq
- Changed page_pool_create return codes, don't return NULL, only
ERR_PTR, as this simplifies err handling in drivers.
V4: many small improvements and cleanups
- Add DOC comment section, that can be used by kernel-doc
- Improve fallback mode, to work better with refcnt based recycling
e.g. remove a WARN as pointed out by Tariq
e.g. quicker fallback if ptr_ring is empty.
V5: Fixed SPDX license as pointed out by Alexei
V6: Adjustments requested by Eric Dumazet
- Adjust ____cacheline_aligned_in_smp usage/placement
- Move rcu_head in struct page_pool
- Free pages quicker on destroy, minimize resources delayed an RCU period
- Remove code for forward/backward compat ABI interface
V8: Issues found by kbuild test robot
- Address sparse should be static warnings
- Only compile+link when a driver use/select page_pool,
mlx5 selects CONFIG_PAGE_POOL, although its first used in two patches
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
