Slated for introduction in the next kernel cycle (Linux 6.20~7.0) is introducing large receive buffer support for IO_uring’s zero-copy receive code path. This large receive buffer support can be very beneficial for those with higher-end networking hardware capable of handling the larger buffers for some significant performance and efficiency wins.
For enhancing the IO_uring network zero-copy receive support that has been in the mainline kernel since Linux 6.15, larger (greater than 4K) receive buffer support is on the way for the next version of the Linux kernel.
Linux block subsystem maintainer and IO_uring lead developer Jens Axboe has …
Slated for introduction in the next kernel cycle (Linux 6.20~7.0) is introducing large receive buffer support for IO_uring’s zero-copy receive code path. This large receive buffer support can be very beneficial for those with higher-end networking hardware capable of handling the larger buffers for some significant performance and efficiency wins.
For enhancing the IO_uring network zero-copy receive support that has been in the mainline kernel since Linux 6.15, larger (greater than 4K) receive buffer support is on the way for the next version of the Linux kernel.
Linux block subsystem maintainer and IO_uring lead developer Jens Axboe has queued up a patch by Pavel Begunkov of Meta for this large receive buffer support:
"There are network cards that support receive buffers larger than 4K, and that can be vastly beneficial for performance, and benchmarks for this patch showed up to 30% CPU util improvement for 32K vs 4K buffers."
Not bad for less than 40 lines of new code.
The patch has worked its way into Axboe’s "for-next" Git branch as part of the "for-7.0/io_uring-zcrx-large-buffers" changes.