On Sat, Jul 07, 2012 at 08:14:34PM -0400, Brad Smith wrote: > I had posted a diff for this and a few other bug fixes here.. > > http://marc.info/?l=openbsd-tech&m=124139607313719&w=2
I've looked at the additional changes you ported from dragonfly in that diff. I don't know what impact these changes will really have so I'd rather not apply them. I cannot verify the slot-time change sephe made to fix problems with his AP, so I'd rather not apply that. The EEPROM values read for tx->txpow are all within the 24/25 range for me. I cannot tell how using these values for different channels will make a noticable difference. Maybe this works around broken EEPROMs in some cards but I'd rather have solid evidence that this is a real problem. The values we use for ARTCSR1 and ARTCSR2 are the same as the linux driver is using. I don't know the source of sephe's information so I cannot verify it. I have no idea what changing these bits really does. The problems with ral you were hoping to fix by porting these changes from dragonfly might already be fixed now in -current. I'm typing this mail across a ral AP that used to lock up very often and now apparently works without issues. I'd prefer waiting to see if we still get reports of failing hostap ral cards that require ifconfig down/up to restore before making more changes to the driver.
