| Age | Commit message (Collapse) | Author |
|---|
|
Send and complete XSK pool frames within TX NAPI context. NAPI context
is triggered by ndo_xsk_wakeup.
Test results with A53 1.2GHz:
xdpsock txonly copy mode, 64 byte frames:
pps pkts 1.00
tx 284,409 11,398,144
Two CPUs with 100% and 10% utilization.
xdpsock txonly zero-copy mode, 64 byte frames:
pps pkts 1.00
tx 511,929 5,890,368
Two CPUs with 100% and 1% utilization.
xdpsock l2fwd copy mode, 64 byte frames:
pps pkts 1.00
rx 248,985 7,315,885
tx 248,921 7,315,885
Two CPUs with 100% and 10% utilization.
xdpsock l2fwd zero-copy mode, 64 byte frames:
pps pkts 1.00
rx 254,735 3,039,456
tx 254,735 3,039,456
Two CPUs with 100% and 4% utilization.
Packet rate increases and CPU utilization is reduced in both cases.
Signed-off-by: Gerhard Engleder <gerhard@engleder-embedded.com>
Reviewed-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Add support for XSK zero-copy to RX path. The setup of the XSK pool can
be done at runtime. If the netdev is running, then the queue must be
disabled and enabled during reconfiguration. This can be done easily
with functions introduced in previous commits.
A more important property is that, if the netdev is running, then the
setup of the XSK pool shall not stop the netdev in case of errors. A
broken netdev after a failed XSK pool setup is bad behavior. Therefore,
the allocation and setup of resources during XSK pool setup is done only
before any queue is disabled. Additionally, freeing and later allocation
of resources is eliminated in some cases. Page pool entries are kept for
later use. Two memory models are registered in parallel. As a result,
the XSK pool setup cannot fail during queue reconfiguration.
In contrast to other drivers, XSK pool setup and XDP BPF program setup
are separate actions. XSK pool setup can be done without any XDP BPF
program. The XDP BPF program can be added, removed or changed without
any reconfiguration of the XSK pool.
Test results with A53 1.2GHz:
xdpsock rxdrop copy mode, 64 byte frames:
pps pkts 1.00
rx 856,054 10,625,775
Two CPUs with both 100% utilization.
xdpsock rxdrop zero-copy mode, 64 byte frames:
pps pkts 1.00
rx 889,388 4,615,284
Two CPUs with 100% and 20% utilization.
Packet rate increases and CPU utilization is reduced.
100% CPU load seems to the base load. This load is consumed by ksoftirqd
just for dropping the generated packets without xdpsock running.
Using batch API reduced CPU utilization slightly, but measurements are
not stable enough to provide meaningful numbers.
Signed-off-by: Gerhard Engleder <gerhard@engleder-embedded.com>
Reviewed-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
TX/RX ring size is static and power of 2 to enable compiler to optimize
modulo operation to mask operation. Make this optimization already in
the code and don't rely on the compiler.
CPU utilisation during high packet rate has not changed. So no
performance improvement has been measured. But it is best practice to
prevent modulo operations.
Suggested-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Signed-off-by: Gerhard Engleder <gerhard@engleder-embedded.com>
Reviewed-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Implement setup of BPF programs for XDP RX path with command
XDP_SETUP_PROG of ndo_bpf(). This is the final step for XDP RX path
support.
There is no need to reinit the RX queues as they are always prepared for
XDP.
Additionally remove $(tsnep-y) from $(tsnep-objs) because it is added
automatically.
Test results with A53 1.2GHz:
XDP_DROP (samples/bpf/xdp1)
proto 17: 883878 pkt/s
XDP_TX (samples/bpf/xdp2)
proto 17: 255693 pkt/s
XDP_REDIRECT (samples/bpf/xdpsock)
sock0@eth2:0 rxdrop xdp-drv
pps pkts 1.00
rx 855,582 5,404,523
tx 0 0
XDP_REDIRECT (samples/bpf/xdp_redirect)
eth2->eth1 613,267 rx/s 0 err,drop/s 613,272 xmit/s
Signed-off-by: Gerhard Engleder <gerhard@engleder-embedded.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
If BPF program is set up, then run BPF program for every received frame
and execute the selected action.
Signed-off-by: Gerhard Engleder <gerhard@engleder-embedded.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Register xdp_rxq_info with page_pool memory model. This is needed for
XDP buffer handling.
Additionally fix error path by removing call of tsnep_phy_close() after
failed tsnep_phy_open().
Signed-off-by: Gerhard Engleder <gerhard@engleder-embedded.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Implement ndo_xdp_xmit() for XDP TX support. Support for fragmented XDP
frames is included.
Also some braces and logic cleanups are done in normal TX path to keep
both TX paths in sync.
Signed-off-by: Gerhard Engleder <gerhard@engleder-embedded.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
TX spin_lock can be eliminated, because the normal TX path is already
protected with __netif_tx_lock and this lock can be used for access to
queue outside of normal TX path too.
Signed-off-by: Gerhard Engleder <gerhard@engleder-embedded.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Refill RX queue in batches of descriptors to improve performance. Refill
is allowed to fail as long as a minimum number of descriptors is active.
Thus, a limited number of failed RX buffer allocations is now allowed
for normal operation. Previously every failed allocation resulted in a
dropped frame.
If the minimum number of active descriptors is reached, then RX buffers
are still reused and frames are dropped. This ensures that the RX queue
never runs empty and always continues to operate.
Prework for future XDP support.
Signed-off-by: Gerhard Engleder <gerhard@engleder-embedded.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Without interrupt throttling, iperf server mode generates a CPU load of
100% (A53 1.2GHz). Also the throughput suffers with less than 900Mbit/s
on a 1Gbit/s link. The reason is a high interrupt load with interrupts
every ~20us.
Reduce interrupt load by throttling of interrupts. Interrupt delay
default is 64us. For iperf server mode the CPU load is significantly
reduced to ~20% and the throughput reaches the maximum of 941MBit/s.
Interrupts are generated every ~140us.
RX and TX coalesce can be configured with ethtool. RX coalesce has
priority over TX coalesce if the same interrupt is used.
Signed-off-by: Gerhard Engleder <gerhard@engleder-embedded.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Use page pool for RX buffer handling. Makes RX path more efficient and
is required prework for future XDP support.
Signed-off-by: Gerhard Engleder <gerhard@engleder-embedded.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Received Ethernet frames are assigned to first RX queue per default.
Based on EtherType Ethernet frames can be assigned to other RX queues.
This enables processing of real-time Ethernet protocols on dedicated
RX queues.
Add RX flow classification interface for EtherType based RX queue
assignment.
Signed-off-by: Gerhard Engleder <gerhard@engleder-embedded.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Support additional TX/RX queue pairs if dedicated interrupt is
available. Interrupts are detected by name in device tree.
Signed-off-by: Gerhard Engleder <gerhard@engleder-embedded.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
For multiple queues multiple interrupts shall be used. Therefore, rework
global interrupt to per queue interrupt.
Every interrupt name shall contain interface name and queue information.
To get a valid interface name, the interrupt request needs to by done
during open like in other drivers. Additionally, this allows the removal
of some initialisation checks in the interrupt handler.
Signed-off-by: Gerhard Engleder <gerhard@engleder-embedded.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Other drivers record RX queue so it should make sense to do that also.
Signed-off-by: Gerhard Engleder <gerhard@engleder-embedded.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The following warning is fixed, by removing the unused resource size:
drivers/net/ethernet/engleder/tsnep_main.c:1155:21-24:
WARNING: Suspicious code. resource_size is maybe missing with io
Reported-by: kernel test robot <lkp@intel.com>
Reported-by: Abaci Robot <abaci@linux.alibaba.com>
Signed-off-by: Gerhard Engleder <gerhard@engleder-embedded.com>
Link: https://lore.kernel.org/r/20211124205225.13985-1-gerhard@engleder-embedded.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
The TSN endpoint Ethernet MAC is a FPGA based network device for
real-time communication.
It is integrated as Ethernet controller with ethtool and PTP support.
For real-time communcation TC_SETUP_QDISC_TAPRIO is supported.
Signed-off-by: Gerhard Engleder <gerhard@engleder-embedded.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
