There are several issues existed in start_xmit():
- Transmitted packets need to be freed before sending a packet, this
introduces delay and increases the average packets transmit
time. This also increase the time that spent in holding the TX lock.
- Notification is enabled after free_old_xmit_skbs() which will
introduce unnecessary interrupts if TX notification happens on the
same CPU that is doing the transmission now (actually, virtio-net
driver are optimized for this case).
So this patch tries to avoid those issues by not cleaning transmitted
packets in start_xmit() when TX NAPI is enabled and disable
notifications even more aggressively. Notification will be since the
beginning of the start_xmit(). But we can't enable delayed
notification after TX is stopped as we will lose the
notifications. Instead, the delayed notification needs is enabled
after the virtqueue is kicked for best performance.
Performance numbers:
1) single queue 2 vcpus guest with pktgen_sample03_burst_single_flow.sh
(burst 256) + testpmd (rxonly) on the host:
- When pinning TX IRQ to pktgen VCPU: split virtqueue PPS were
increased 55% from 6.89 Mpps to 10.7 Mpps and 32% TX interrupts were
eliminated. Packed virtqueue PPS were increased 50% from 7.09 Mpps to
10.7 Mpps, 99% TX interrupts were eliminated.
- When pinning TX IRQ to VCPU other than pktgen: split virtqueue PPS
were increased 96% from 5.29 Mpps to 10.4 Mpps and 45% TX interrupts
were eliminated; Packed virtqueue PPS were increased 78% from 6.12
Mpps to 10.9 Mpps and 99% TX interrupts were eliminated.
2) single queue 1 vcpu guest + vhost-net/TAP on the host: single
session netperf from guest to host shows 82% improvement from
31Gb/s to 58Gb/s, %stddev were reduced from 34.5% to 1.9% and 88%
of TX interrupts were eliminated.
Signed-off-by: Jason Wang <jasowang@redhat.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add an implementation for DMTF DSP0283, which defines a MCTP-over-USB
transport. As per that spec, we're restricted to full speed mode,
requiring 512-byte transfers.
Each MCTP-over-USB interface is a peer-to-peer link to a single MCTP
endpoint, so no physical addressing is required (of course, that MCTP
endpoint may then bridge to further MCTP endpoints). Consequently,
interfaces will report with no lladdr data:
# mctp link
dev lo index 1 address 00:00:00:00:00:00 net 1 mtu 65536 up
dev mctpusb0 index 6 address none net 1 mtu 68 up
This is a simple initial implementation, with single rx & tx urbs, and
no multi-packet tx transfers - although we do accept multi-packet rx
from the device.
Includes suggested fixes from Santosh Puranik <spuranik@nvidia.com>.
Signed-off-by: Jeremy Kerr <jk@codeconstruct.com.au>
Cc: Santosh Puranik <spuranik@nvidia.com>
Link: https://patch.msgid.link/20250221-dev-mctp-usb-v3-2-3353030fe9cc@codeconstruct.com.au
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
stmmac_init_dma_engine() uses dev_err() which leads to errors being
reported as e.g:
dwc-eth-dwmac 2490000.ethernet: Failed to reset the dma
dwc-eth-dwmac 2490000.ethernet eth0: stmmac_hw_setup: DMA engine initialization failed
stmmac_init_dma_engine() is only called from stmmac_hw_setup() which
itself uses netdev_err(), and we will have a net_device setup. So,
change the dev_err() to netdev_err() to give consistent error
messages.
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Link: https://patch.msgid.link/E1tl5y1-004UgG-8X@rmk-PC.armlinux.org.uk
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The current test marks all unexpected return values as failed and sets ret
to 1. If a test is skipped, the entire test also returns 1, incorrectly
indicating failure.
To fix this, add a skipped variable and set ret to 4 if it was previously
0. Otherwise, keep ret set to 1.
Signed-off-by: Hangbin Liu <liuhangbin@gmail.com>
Link: https://patch.msgid.link/20250220085326.1512814-1-liuhangbin@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Ido Schimmel says:
====================
net: fib_rules: Add DSCP mask support
In some deployments users would like to encode path information into
certain bits of the IPv6 flow label, the UDP source port and the DSCP
field and use this information to route packets accordingly.
Redirecting traffic to a routing table based on specific bits in the
DSCP field is not currently possible. Only exact match is currently
supported by FIB rules.
This patchset extends FIB rules to match on the DSCP field with an
optional mask.
Patches #1-#5 gradually extend FIB rules to match on the DSCP field with
an optional mask.
Patch #6 adds test cases for the new functionality.
iproute2 support can be found here [1].
[1] https://github.com/idosch/iproute2/tree/submit/fib_rule_mask_v1
====================
Link: https://patch.msgid.link/20250220080525.831924-1-idosch@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add tests for FIB rules that match on DSCP with a mask. Test both good
and bad flows and both the input and output paths.
# ./fib_rule_tests.sh
IPv6 FIB rule tests
[...]
TEST: rule6 check: dscp redirect to table [ OK ]
TEST: rule6 check: dscp no redirect to table [ OK ]
TEST: rule6 del by pref: dscp redirect to table [ OK ]
TEST: rule6 check: iif dscp redirect to table [ OK ]
TEST: rule6 check: iif dscp no redirect to table [ OK ]
TEST: rule6 del by pref: iif dscp redirect to table [ OK ]
TEST: rule6 check: dscp masked redirect to table [ OK ]
TEST: rule6 check: dscp masked no redirect to table [ OK ]
TEST: rule6 del by pref: dscp masked redirect to table [ OK ]
TEST: rule6 check: iif dscp masked redirect to table [ OK ]
TEST: rule6 check: iif dscp masked no redirect to table [ OK ]
TEST: rule6 del by pref: iif dscp masked redirect to table [ OK ]
[...]
Tests passed: 316
Tests failed: 0
Reviewed-by: Petr Machata <petrm@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Guillaume Nault <gnault@redhat.com>
Link: https://patch.msgid.link/20250220080525.831924-7-idosch@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Extend IPv6 FIB rules to match on DSCP using a mask. Unlike IPv4, also
initialize the DSCP mask when a non-zero 'tos' is specified as there is
no difference in matching between 'tos' and 'dscp'. As a side effect,
this makes it possible to match on 'dscp 0', like in IPv4.
Reviewed-by: Petr Machata <petrm@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Guillaume Nault <gnault@redhat.com>
Link: https://patch.msgid.link/20250220080525.831924-4-idosch@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add an attribute that allows matching on DSCP with a mask. Matching on
DSCP with a mask is needed in deployments where users encode path
information into certain bits of the DSCP field.
Temporarily set the type of the attribute to 'NLA_REJECT' while support
is being added.
Reviewed-by: Petr Machata <petrm@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Guillaume Nault <gnault@redhat.com>
Link: https://patch.msgid.link/20250220080525.831924-2-idosch@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Martin KaFai Lau says:
====================
pull-request: bpf-next 2025-02-20
We've added 19 non-merge commits during the last 8 day(s) which contain
a total of 35 files changed, 1126 insertions(+), 53 deletions(-).
The main changes are:
1) Add TCP_RTO_MAX_MS support to bpf_set/getsockopt, from Jason Xing
2) Add network TX timestamping support to BPF sock_ops, from Jason Xing
3) Add TX metadata Launch Time support, from Song Yoong Siang
* tag 'for-netdev' of https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next:
igc: Add launch time support to XDP ZC
igc: Refactor empty frame insertion for launch time support
net: stmmac: Add launch time support to XDP ZC
selftests/bpf: Add launch time request to xdp_hw_metadata
xsk: Add launch time hardware offload support to XDP Tx metadata
selftests/bpf: Add simple bpf tests in the tx path for timestamping feature
bpf: Support selective sampling for bpf timestamping
bpf: Add BPF_SOCK_OPS_TSTAMP_SENDMSG_CB callback
bpf: Add BPF_SOCK_OPS_TSTAMP_ACK_CB callback
bpf: Add BPF_SOCK_OPS_TSTAMP_SND_HW_CB callback
bpf: Add BPF_SOCK_OPS_TSTAMP_SND_SW_CB callback
bpf: Add BPF_SOCK_OPS_TSTAMP_SCHED_CB callback
net-timestamp: Prepare for isolating two modes of SO_TIMESTAMPING
bpf: Disable unsafe helpers in TX timestamping callbacks
bpf: Prevent unsafe access to the sock fields in the BPF timestamping callback
bpf: Prepare the sock_ops ctx and call bpf prog for TX timestamping
bpf: Add networking timestamping support to bpf_get/setsockopt()
selftests/bpf: Add rto max for bpf_setsockopt test
bpf: Support TCP_RTO_MAX_MS for bpf_setsockopt
====================
Link: https://patch.msgid.link/20250221022104.386462-1-martin.lau@linux.dev
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Not all the devices have the capability for the driver to query for the
registered RSS configuration. The driver can discover this by checking
the relevant device option during setup. If it cannot, the driver needs
to store the RSS config cache and directly return such cache when
queried by the ethtool. RSS config is inited when driver probes. Also the
default RSS config will be adjusted when there is RX queue count change.
At this point, only keys of GVE_RSS_KEY_SIZE and indirection tables of
GVE_RSS_INDIR_SIZE are supported.
Signed-off-by: Ziwei Xiao <ziweixiao@google.com>
Reviewed-by: Harshitha Ramamurthy <hramamurthy@google.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: Praveen Kaligineedi <pkaligineedi@google.com>
Signed-off-by: Jeroen de Borst <jeroendb@google.com>
Link: https://patch.msgid.link/20250219200451.3348166-1-jeroendb@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
- Add test for creating link in another netns when a link of the same
name and ifindex exists in current netns.
- Add test to verify that link is created in target netns directly -
no link new/del events should be generated in link netns or current
netns.
- Add test cases to verify that link-netns is set as expected for
various drivers and combination of namespace-related parameters.
Signed-off-by: Xiao Liang <shaw.leon@gmail.com>
Link: https://patch.msgid.link/20250219125039.18024-14-shaw.leon@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Make rtnl_newlink_create() create device in target namespace directly.
Avoid extra netns change when link netns is provided.
Device drivers has been converted to be aware of link netns, that is not
assuming device netns is and link netns is the same when ops->newlink()
is called.
Signed-off-by: Xiao Liang <shaw.leon@gmail.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Link: https://patch.msgid.link/20250219125039.18024-12-shaw.leon@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Currently some IPv6 tunnel drivers set tnl->net to dev_net(dev) in
ndo_init(), which is called in register_netdevice(). However, it lacks
the context of link-netns when we enable cross-net tunnels at device
registration time.
Let's move the init of tunnel link-netns before register_netdevice().
ip6_gre has already initialized netns, so just remove the redundant
assignment.
Signed-off-by: Xiao Liang <shaw.leon@gmail.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Link: https://patch.msgid.link/20250219125039.18024-8-shaw.leon@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
When link_net is set, use it as link netns instead of dev_net(). This
prepares for rtnetlink core to create device in target netns directly,
in which case the two namespaces may be different.
Convert common ip_tunnel_newlink() to accept an extra link netns
argument.
Signed-off-by: Xiao Liang <shaw.leon@gmail.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Link: https://patch.msgid.link/20250219125039.18024-7-shaw.leon@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Device denoted by IFLA_LINK is in link_net (IFLA_LINK_NETNSID) or
source netns by design, but 6lowpan uses dev_net.
Note dev->netns_local is set to true and currently link_net is
implemented via a netns change. These together effectively reject
IFLA_LINK_NETNSID.
This patch adds a validation to ensure link_net is either NULL or
identical to dev_net. Thus it would be fine to continue using dev_net
when rtnetlink core begins to create devices directly in target netns.
Signed-off-by: Xiao Liang <shaw.leon@gmail.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Link: https://patch.msgid.link/20250219125039.18024-5-shaw.leon@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add two helper functions - rtnl_newlink_link_net() and
rtnl_newlink_peer_net() for netns fallback logic. Peer netns falls back
to link netns, and link netns falls back to source netns.
Convert the use of params->net in netdevice drivers to one of the helper
functions for clarity.
Signed-off-by: Xiao Liang <shaw.leon@gmail.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Link: https://patch.msgid.link/20250219125039.18024-4-shaw.leon@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
There are 4 net namespaces involved when creating links:
- source netns - where the netlink socket resides,
- target netns - where to put the device being created,
- link netns - netns associated with the device (backend),
- peer netns - netns of peer device.
Currently, two nets are passed to newlink() callback - "src_net"
parameter and "dev_net" (implicitly in net_device). They are set as
follows, depending on netlink attributes in the request.
+------------+-------------------+---------+---------+
| peer netns | IFLA_LINK_NETNSID | src_net | dev_net |
+------------+-------------------+---------+---------+
| | absent | source | target |
| absent +-------------------+---------+---------+
| | present | link | link |
+------------+-------------------+---------+---------+
| | absent | peer | target |
| present +-------------------+---------+---------+
| | present | peer | link |
+------------+-------------------+---------+---------+
When IFLA_LINK_NETNSID is present, the device is created in link netns
first and then moved to target netns. This has some side effects,
including extra ifindex allocation, ifname validation and link events.
These could be avoided if we create it in target netns from
the beginning.
On the other hand, the meaning of src_net parameter is ambiguous. It
varies depending on how parameters are passed. It is the effective
link (or peer netns) by design, but some drivers ignore it and use
dev_net instead.
To provide more netns context for drivers, this patch packs existing
newlink() parameters, along with the source netns, link netns and peer
netns, into a struct. The old "src_net" is renamed to "net" to avoid
confusion with real source netns, and will be deprecated later. The use
of src_net are converted to params->net trivially.
Signed-off-by: Xiao Liang <shaw.leon@gmail.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Link: https://patch.msgid.link/20250219125039.18024-3-shaw.leon@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
When creating link, lookup for existing device in target net namespace
instead of current one.
For example, two links created by:
# ip link add dummy1 type dummy
# ip link add netns ns1 dummy1 type dummy
should have no conflict since they are in different namespaces.
Signed-off-by: Xiao Liang <shaw.leon@gmail.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Link: https://patch.msgid.link/20250219125039.18024-2-shaw.leon@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add dtschema for the MDIO controller found in the RTL9300 Ethernet
switch. The controller is slightly unusual in that direct MDIO
communication is not possible. We model the MDIO controller with the
MDIO buses as child nodes and the PHYs as children of the buses. The
mapping of switch port number to MDIO bus/addr requires the
ethernet-ports sibling to provide the mapping via the phy-handle
property.
Signed-off-by: Chris Packham <chris.packham@alliedtelesis.co.nz>
Reviewed-by: Rob Herring (Arm) <robh@kernel.org>
Link: https://patch.msgid.link/20250218195216.1034220-4-chris.packham@alliedtelesis.co.nz
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The OEM SFP-2.5G-BX10-D/U SFP module pair is meant to operate with
2500Base-X. However, in their EEPROM they incorrectly specify:
Transceiver codes : 0x00 0x12 0x00 0x00 0x12 0x00 0x01 0x05 0x00
BR, Nominal : 2500MBd
Use sfp_quirk_2500basex for this module to allow 2500Base-X mode anyway.
Tested on BananaPi R3.
Signed-off-by: Birger Koblitz <mail@birger-koblitz.de>
Reviewed-by: Daniel Golle <daniel@makrotopia.org>
Link: https://patch.msgid.link/20250218-b4-lkmsub-v1-1-1e51dcabed90@birger-koblitz.de
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Jakub Kicinski says:
====================
selftests: drv-net: improve the queue test for XSK
We see some flakes in the the XSK test:
Exception| Traceback (most recent call last):
Exception| File "/home/virtme/testing-18/tools/testing/selftests/net/lib/py/ksft.py", line 218, in ksft_run
Exception| case(*args)
Exception| File "/home/virtme/testing-18/tools/testing/selftests/drivers/net/./queues.py", line 53, in check_xdp
Exception| ksft_eq(q['xsk'], {})
Exception| KeyError: 'xsk'
I think it's because the method of running the helper in the background
is racy. Add more solid infra for waiting for a background helper to be
initialized.
v1: https://lore.kernel.org/20250218195048.74692-1-kuba@kernel.org
====================
Link: https://patch.msgid.link/20250219234956.520599-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
We use wait_port_listen() extensively to wait for a process
we spawned to be ready. Not all processes will open listening
sockets. Add a method of explicitly waiting for a child to
be ready. Pass a FD to the spawned process and wait for it
to write a message to us. FD number is passed via KSFT_READY_FD
env variable.
Similarly use KSFT_WAIT_FD to let the child process for a sign
that we are done and child should exit. Sending a signal to
a child with shell=True can get tricky.
Make use of this method in the queues test to make it less flaky.
Acked-by: Stanislav Fomichev <sdf@fomichev.me>
Acked-by: Joe Damato <jdamato@fastly.com>
Tested-by: Joe Damato <jdamato@fastly.com>
Link: https://patch.msgid.link/20250219234956.520599-6-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
A previous patch introduces a build-time warning when CONFIG_DCB
is disabled:
drivers/net/ethernet/marvell/octeontx2/nic/otx2_pf.c: In function 'otx2_probe':
drivers/net/ethernet/marvell/octeontx2/nic/otx2_pf.c:3217:1: error: label 'err_free_zc_bmap' defined but not used [-Werror=unused-label]
drivers/net/ethernet/marvell/octeontx2/nic/otx2_vf.c: In function 'otx2vf_probe':
drivers/net/ethernet/marvell/octeontx2/nic/otx2_vf.c:740:1: error: label 'err_free_zc_bmap' defined but not used [-Werror=unused-label]
Add the same #ifdef check around it.
Fixes: efabce2901 ("octeontx2-pf: AF_XDP zero copy receive support")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Suman Ghosh <sumang@marvell.com>
Link: https://patch.msgid.link/20250219162239.1376865-1-arnd@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Tariq Toukan says:
====================
mlx5 misc enhancements 2025-02-19
This small series enhances the mlx5 ethtool link speed code (no
functional change), in addition to a Kconfig description enhancement.
====================
Link: https://patch.msgid.link/20250219114112.403808-1-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The function ext_requested() serves two distinct purposes: it checks
if extended link modes were requested, and it selects whether to use
extended or legacy link modes.
This change separates these two purposes. Now, ext_link_mode_requested()
is used directly for checking if extended link modes are requested,
while the selection of extended modes is handled independently based on
the autonegotiation status.
By making this distinction, the logic for determining whether to select
extended or legacy link modes is clearer.
Signed-off-by: Shahar Shitrit <shshitrit@nvidia.com>
Reviewed-by: Carolina Jubran <cjubran@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20250219114112.403808-6-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
