kern/165969: Slower performance in adhoc mode vs Client/AP mode, same HW (atheros adapters)

Johann jhugo at meraka.csir.co.za
Mon Mar 12 11:00:27 UTC 2012


>Number:         165969
>Category:       kern
>Synopsis:       Slower performance in adhoc mode vs Client/AP mode, same HW (atheros adapters)
>Confidential:   no
>Severity:       non-critical
>Priority:       low
>Responsible:    freebsd-bugs
>State:          open
>Quarter:        
>Keywords:       
>Date-Required:
>Class:          sw-bug
>Submitter-Id:   current-users
>Arrival-Date:   Mon Mar 12 11:00:26 UTC 2012
>Closed-Date:
>Last-Modified:
>Originator:     Johann
>Release:        FreeBSD 9.0-STABLE
>Organization:
CSIR
>Environment:
FreeBSD jhugo-rtr 9.0-STABLE FreeBSD 9.0-STABLE #7: Tue Feb 14 13:50:42 SAST 2012     jhay at magrathea.meraka.csir.co.za:/var/scratch/9-stable/obj/arm.armeb/var/scratch/9-stable/src/sys/SMALL-AVILA  arm
>Description:
I get lower throughputs in adhoc mode, compared to AP/client mode for 
the same two devices.

AP -> client            iperf   27.2 MBytes  28.8 Mbits/sec
adhoc -> adhoc          iperf   12.1 MBytes  21.3 Mbits/sec

dev.ath.0.stats.ast_tx_longretry is more than double in adhoc mode.

The two devices are directly next to each other in the LAB
HW platform = Gateworks Avila GW2348-4
Wifi adapters = Compex WLM54AGP23
Atheros driver from +- 10 Jan 2012
OS = FreeBSD 9.0-STABLE

Adhoc mode for FreeBSD 9.0-STABLE (as well as FreeBSD 8.0-STABLE) are also slower than adhoc mode for FreeBSD 7.2-STABLE

9.0-STABLE  adhoc -> adhoc  iperf   21.3 Mbits/sec
8.0-STABLE  adhoc -> adhoc  iperf   21.3 Mbits/sec
7.2-STABLE  adhoc -> adhoc  iperf   25.1 Mbits/sec


>How-To-Repeat:
- Place two wifi nodes next to each other. 
- Configure one device into AP mode and the other onto client mode.
- Do an iperf between the two nodes.

iperf  28.8 Mbits/sec

- Now use the same two devices and configure them into adhoc mode.
- Do an iperf between the same two nodes.

iperf  21.3 Mbits/sec
>Fix:


>Release-Note:
>Audit-Trail:
>Unformatted:


More information about the freebsd-bugs mailing list