arm64 iperf performance suboptimal
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
linux (Ubuntu) |
Fix Released
|
High
|
Unassigned | ||
Trusty |
Fix Released
|
Undecided
|
dann frazier | ||
Utopic |
Fix Released
|
High
|
Unassigned |
Bug Description
[Impact]
The copy_{to,from}_user implementations in the Ubuntu kernel are suboptimal. Optimized implementations have been submitted upstream and have shown a significant improvement in network performance.
Iperf performance increase:
-l (size) 1 core result
Optimized 64B 44-51Mb/s
Original 64B 34-50.7Mb/s
[Test Case]
Generate traffic from one node to another using iperf (see above for config).
[Regression Risk]
These functions are obviously used heavily throughout the kernel, so a defect here could have significant impact. This risk is mitigated by using an implementation heavily based on the linaro string libraries (which are used in other places already, e.g. glibc), and through active testing of this patch on real hardware using a trusty-kernel base.
Changed in linux (Ubuntu Trusty): | |
assignee: | nobody → dann frazier (dannf) |
status: | New → Fix Committed |
Changed in linux (Ubuntu Utopic): | |
status: | Triaged → Fix Committed |
tags: | added: arm-hs-vivid |
Can you also send this patch to the Ubuntu kernel-team mailing list for review?