On Wednesday 25 April 2007 14:16, John W. Linville wrote: > On Tue, Apr 24, 2007 at 10:03:03PM -0700, James Ketrenos wrote: > > (I would like to see hw_scan remain since iwlwifi uses it -- which we > > hope to submit as soon as the code restructuring is complete) > > I was going to mention this in response to Michael's patch, but then > I checked your current iwlwifi git tree. There your hw_scan routine > is commented-out, so I figured you didn't want/need it anymore...? > It was never needed in the first place except as an iwlwifi specific hack to minimize any disruption to the current connection. (which often seems to be commented out due to firmware issues) mac80211 now does its own PS mode scanning trick along with some frame queuing as a generic means to minimize disruptions to the current connection during scans. It still needs some improvements in order to quickly empty the current hardware TX queue before commencing a scan and possibly time/segment the scanning such that beacons are not lost, but it can certainly be done. Besides the fact that what hw_scan does can be done by mac80211 for all drivers, the use of the hw_scan callback breaks the AP autoconfiguration code in ieee80211_sta.c due to its inadequate design. Calling hw_scan starts a hardware scan, but there is no way to know when the scan is completed. This means the code has no way of knowing when it should look through its scan results and configure itself to connect to an AP if a scan was started by mac80211 for the purpose of finding an AP to connect to. Even if that problem were addressed, I still wouldn't like it as the design of the hw_scan callback is deficient in a number of other ways and is completely inapplicable to all other hardware/firmware softmac designs that I know. -Michael Wu
Attachment:
pgpzAjPARnfyC.pgp
Description: PGP signature
