On Fri, Jun 27, 2025 at 7:01 PM Jason Xing <kerneljasonxing@xxxxxxxxx> wrote:
>
> From: Jason Xing <kernelxing@xxxxxxxxxxx>
>
> This patch provides a setsockopt method to let applications leverage to
> adjust how many descs to be handled at most in one send syscall. It
> mitigates the situation where the default value (32) that is too small
> leads to higher frequency of triggering send syscall.
>
> Considering the prosperity/complexity the applications have, there is no
> absolutely ideal suggestion fitting all cases. So keep 32 as its default
> value like before.
>
> The patch does the following things:
> - Add XDP_MAX_TX_BUDGET socket option.
> - Convert TX_BATCH_SIZE to tx_budget_spent.
> - Set tx_budget_spent to 32 by default in the initialization phase as a
> per-socket granular control. 32 is also the min value for
> tx_budget_spent.
> - Set the range of tx_budget_spent as [32, xs->tx->nentries].
>
> The idea behind this comes out of real workloads in production. We use a
> user-level stack with xsk support to accelerate sending packets and
> minimize triggering syscalls. When the packets are aggregated, it's not
> hard to hit the upper bound (namely, 32). The moment user-space stack
> fetches the -EAGAIN error number passed from sendto(), it will loop to try
> again until all the expected descs from tx ring are sent out to the driver.
> Enlarging the XDP_MAX_TX_BUDGET value contributes to less frequency of
> sendto() and higher throughput/PPS.
>
> Here is what I did in production, along with some numbers as follows:
> For one application I saw lately, I suggested using 128 as max_tx_budget
> because I saw two limitations without changing any default configuration:
> 1) XDP_MAX_TX_BUDGET, 2) socket sndbuf which is 212992 decided by
> net.core.wmem_default. As to XDP_MAX_TX_BUDGET, the scenario behind
> this was I counted how many descs are transmitted to the driver at one
> time of sendto() based on [1] patch and then I calculated the
> possibility of hitting the upper bound. Finally I chose 128 as a
> suitable value because 1) it covers most of the cases, 2) a higher
> number would not bring evident results. After twisting the parameters,
> a stable improvement of around 4% for both PPS and throughput and less
> resources consumption were found to be observed by strace -c -p xxx:
> 1) %time was decreased by 7.8%
> 2) error counter was decreased from 18367 to 572
More interesting numbers are arriving here as I run some benchmarks
from xdp-project/bpf-examples/AF_XDP-example/ in my VM.
Running "sudo taskset -c 2 ./xdpsock -i eth0 -q 1 -l -N -t -b 256"
Using the default configure 32 as the max budget iteration:
sock0@eth0:1 txonly xdp-drv
pps pkts 1.01
rx 0 0
tx 48,574 49,152
Enlarging the value to 256:
sock0@eth0:1 txonly xdp-drv
pps pkts 1.00
rx 0 0
tx 148,277 148,736
Enlarging the value to 512:
sock0@eth0:1 txonly xdp-drv
pps pkts 1.00
rx 0 0
tx 226,306 227,072
The performance of pps goes up by 365% (with max budget set as 512)
which is an incredible number :)
If the next-respin is needed, I will attach those convincing numbers
in the commit message.
Thanks,
Jason
