Red Hat Bugzilla – Bug 700973
Regression of Wireless performance on Atheros AR928X adapter
Last modified: 2012-06-04 15:07:50 EDT
Description of problem: Compared to Fedora 14, the wireless takes several seconds more to connect on boot and wake from sleep(going from less than 5 in F14 to 20-30 seconds in F15).
Once the wireless connection is actually made, the network performance is greatly degraded(a 20MB download took 5 minutes compared to the previous 5 seconds on F14). The slowdown can be counteracted by adding "hwcrypt=1" to the ath9k options.
Version-Release number of selected component (if applicable): 126.96.36.199-20.fc15.x86_64
How reproducible: Every time wifi networking is needed on a computer with a AR928X wireless NIC.
Steps to Reproduce:
1. Install Fedora 15 Beta on computer with AR928X wireless adapter
2. connect to wireless network
3. try to do anything with the network.
Actual results: Wireless connection takes in excess of 20 seconds to connect. Once you actually try to do anything, the network connectivity is excessively slow to the point of uselessness
Expected results: Wireless connection should be made in roughly 5 seconds and speed should be comparable to Fedora 14 and Windows.
Created attachment 495903 [details]
Created attachment 495904 [details]
Created attachment 495905 [details]
Might be a dupe/same as the issue in 702722
You can check current upstream drivers to see if problem is fixed there
I'm having a similar problem with this adapter. I get frequent packet loss, and frequent absurdly-high ping times (probably also packet loss with automatic retransmit, I guess). The same machine sees no packet loss when running Windows and seemed to be working fine before I upgraded from F13 to F15. Other devices on the same network work fine. The AP is about three metres from the F15 box, in the same room.
I have updated to 2.6.40-4, I've tried Stanislaw's latest compat-wireless and compat-wireless-next packages with their respective kernels, and I even downloaded the compat-wireless source (August 8) from here:
and built it. I also tried nohwcrypt=1. Nothing makes any difference.
Any suggestions on what to try next?
If bug is on current upstream driver send bug report to email@example.com and cc firstname.lastname@example.org. Note in email this is regression. You probably will be asked about providing more info, i.e dmesg with debug= module option.
Brandon, do these performance problems continue with 2.6.40 kernels?