Thread: [Madwifi-devel] Re: how to identify echo'd frames? (Russell Senior)
Status: Beta
Brought to you by:
otaku
From: Bell K. <bel...@pe...> - 2005-11-29 06:29:54
|
Hi, I think this is a protocol layer violation in madwifi-ng: it should not combine station and wds in one virtual ethernet device, upper layer(802) have no way to identify where(which port) the packet comes from, and doing peer/loop detection within driver is a bad idea, it should be done in 802 bridge layer, like hostap wds design. there is also a hostap mode wds in madwifi-ng, use it instead. -- Bell Kin System Analyst PEK Inc. Consult & Design TEL:+886-3-5290115 FAX:+886-3-5297445 No.4 Lane 356, Xinxiang St. Hsinchu City 300 Taiwan, ROC |
From: Michael R. <ma...@no...> - 2005-11-29 08:24:52
|
Hi. On Tue, 2005-11-29 at 14:48 +0800, Bell Kin wrote: > I think this is a protocol layer violation in madwifi-ng: > it should not combine station and wds in one virtual ethernet device, > upper layer(802) have no way to identify where(which port) the packet > comes from, So you suggest that 4-address-header support should only be available for virtual devices that have been set to WDS mode? Bye, Mike |
From: Bell K. <bel...@pe...> - 2005-11-29 11:13:32
|
Michael Renzmann wrote: >So you suggest that 4-address-header support should only be available >for virtual devices that have been set to WDS mode? > > we can add notes or better examples to the user guide, or just remove the wds flag of ap/sta to simplify the packet tx path. the bridge code in net80211 is conflicting with linux bridge, it is useful for large number or dynamic wds nodes, but like an extra dumb switch in a spanning tree environment, if you link it with other bridge or have loop bad things happen. how do you think? -- Bell Kin System Analyst PEK Inc. Consult & Design TEL:+886-3-5290115 FAX:+886-3-5297445 No.4 Lane 356, Xinxiang St. Hsinchu City 300 Taiwan, ROC |