On Wed, Jun 10, 2009 at 3:39 PM, Hin-Tak Leung<hintak.leung@xxxxxxxxx> wrote: > On Wed, Jun 10, 2009 at 8:30 AM, Johannes Berg<johannes@xxxxxxxxxxxxxxxx> wrote: >> On Wed, 2009-06-10 at 04:18 +0100, Hin-Tak Leung wrote: >> >>> But I have been using vendor driver 3.0 for a few weeks in AP mode and >>> I am as happy with it as I could be, I think; I have >>> suspended//resumed clients. It works adequately. >> >> We're not concerned about suspend/hibernate, we're concerned about >> 802.11 power saving, where the AP is required to buffer frames until the >> clients wake up again. This will work adequately, but not perfectly, for >> unicast frames, but we have not found a way of making zd1211 send >> buffered multicast frames after the DTIM beacon as required. >> >> This means that your sleeping clients will not be receiving multicast, >> and as such be invisible to the network once they fall off the ARP/NDP >> caches. This is the reason we have said that it will not be possible to >> support AP mode with this card. I'm curious how, if at all, the vendor >> driver handles that. > > I was talking about the client suspending/resuming (i.e. the client > disappearing). > > This is what appears in the AP machine's dmesg during the client's sleep: > --------------------------------------------------------- > *****Age one***** > aid:1 > now:264313809 > ttl:264163665 > idleTime:150144 > zd1205_notify_disjoin_event > Send Deasoc Req to 00:16:44:8f:71:93 RSN=4 > STA_DISASSOCIATED:00:16:44:8f:71:93 > Reject Auth Due to ar2524drv/src/zdpsmon.c,568. staSte=4 > Update BCN @ 286757537 > Re_Asoc: 00:16:44:8f:71:93, aid=1 > ------------------------------------------------------------ > > It seems that the vendor driver on the AP machine simply diassociates > a sleeping cliient for inactivity after 10 minutes during the sleep > and let the client does reassociating when it wakes again. I just grep > for 'idleTime' in the 3.0 source and 10 minutes is exactly what it > does. > But again, that is not the problem that prevents proper AP mode; it's 802.11 dynamic power saving of the card, unrelated to suspend/resume/powersave mode of the host. (Note that I'm also for allowing such "buggy" AP mode on devices without a HW multicast buffer, perhaps with hostapd spitting out a big, fat warning on startup; but Johannes has the final say in mac80211-related questions.) (To Johannes: exactly why is it required for the multicast buffer to be in hardware?) -- Vista: [V]iruses, [I]ntruders, [S]pyware, [T]rojans and [A]dware. :-) -- To unsubscribe from this list: send the line "unsubscribe linux-wireless" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
