Re: misc/172955: [ath] 11n does not work in adhoc mode
On Tuesday 23 October 2012 23:29:16 Chuck Burns wrote: > On 10/23/2012 2:23 PM, Adrian Chadd wrote: > > I'd really appreciate someone stepping in here. I'm trying to finish > > off the power save queue handling changes and then move to finishing > > correcting the ps-poll support. Then I have the whole TX and RX data > > and control path to jiggle around to make things (more) predictable > > and better behaved. I really don't want to stray from this. :-) > > > > Do I have to promise free hardware and t-shirts? :-) > > I could be mistaken, but I was under the impression that adhoc only > supported B or G speeds, and that if you wanted N speeds, you have to > use STA rather than ADHOC. > > I've heard this discussion outside of relation to FreeBSD, so I know > it's a well-discussed issue. I also think I remember reading something > about certain chipsets supporting 11n in ADHOC mode, but only when > connecting to other chipsets. I think it was MAYBE an atheros chipset, > but I don't really remember -exactly- what I read, or where it was... > > Here is a linux ath9k forum post regarding it: > http://comments.gmane.org/gmane.linux.drivers.ath9k.devel/2660 Seems to > indicate that 11n in ADHOC isn't supported there either.. Sooo Saw this yesterday on another mailing list. --- ad-hoc mode is supported in 802.11n since at least a year with the ath9k driver. It might be broken in a specific version of Openwrt trunk, though. --- Johann ___ freebsd-wireless@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-wireless To unsubscribe, send any mail to "freebsd-wireless-unsubscr...@freebsd.org"
Re: misc/172955: [ath] 11n does not work in adhoc mode
On 23 October 2012 14:29, Chuck Burns wrote: > I could be mistaken, but I was under the impression that adhoc only > supported B or G speeds, and that if you wanted N speeds, you have to use > STA rather than ADHOC. I don't see why that is. > I've heard this discussion outside of relation to FreeBSD, so I know it's a > well-discussed issue. I also think I remember reading something about > certain chipsets supporting 11n in ADHOC mode, but only when connecting to > other chipsets. I think it was MAYBE an atheros chipset, but I don't really > remember -exactly- what I read, or where it was... I doubt there's a problem with it in the specification. I am guessing that ADDBA/DELBA will go between adhoc nodes as per normal. But I've not read the spec to see how it should work. In any case I don't see why we can't do 11n rates (with no aggregation.) The fundamental problem here are: * that we're not at all parsing the HT fields in the IBSS beacons; * the default channel lookup doesn't seem to find ht20/ht40 channels; only 11a/11bg. I had to manually configure one end to be 6:ht/40 for it to come up; then it advertised HT fine. > Here is a linux ath9k forum post regarding it: > http://comments.gmane.org/gmane.linux.drivers.ath9k.devel/2660 Seems to > indicate that 11n in ADHOC isn't supported there either.. Sooo That thread just says "it doesn't work, no reason why it shouldn't." All the basic code is there. What we're missing is the glue that parses the WMM, HTINFO and HTCAP fields, then uses that when configuring up the rate control to the new node. I just haven't sat down and figured out how the IBSS node "learning" works. It discovers neighbors via hearing beacons BUT it doesn't use the normal path for setting things up. The STA side does it via association responses rather than beacons. It updates the WMM/HTINFO/HTCAP fields during beacon RX, but it actually uses the probe response to configure things. Adrian ___ freebsd-wireless@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-wireless To unsubscribe, send any mail to "freebsd-wireless-unsubscr...@freebsd.org"
Re: misc/172955: [ath] 11n does not work in adhoc mode
On 10/23/2012 2:23 PM, Adrian Chadd wrote: I'd really appreciate someone stepping in here. I'm trying to finish off the power save queue handling changes and then move to finishing correcting the ps-poll support. Then I have the whole TX and RX data and control path to jiggle around to make things (more) predictable and better behaved. I really don't want to stray from this. :-) Do I have to promise free hardware and t-shirts? :-) I could be mistaken, but I was under the impression that adhoc only supported B or G speeds, and that if you wanted N speeds, you have to use STA rather than ADHOC. I've heard this discussion outside of relation to FreeBSD, so I know it's a well-discussed issue. I also think I remember reading something about certain chipsets supporting 11n in ADHOC mode, but only when connecting to other chipsets. I think it was MAYBE an atheros chipset, but I don't really remember -exactly- what I read, or where it was... Here is a linux ath9k forum post regarding it: http://comments.gmane.org/gmane.linux.drivers.ath9k.devel/2660 Seems to indicate that 11n in ADHOC isn't supported there either.. Sooo -- Chuck Burns ___ freebsd-wireless@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-wireless To unsubscribe, send any mail to "freebsd-wireless-unsubscr...@freebsd.org"
Re: misc/172955: [ath] 11n does not work in adhoc mode
On 23 October 2012 03:31, Johann Hugo wrote: >> Also - the WME parameters aren't parsed either, leading to your > >> observation about performance drops between 7x and 8.x. > I'm ready to test any upgrades :-) Well, I was more hoping that someone else would be interested in tackling this particular issue. It's a general stack issue affecting all ibss 802.11 behaviour, not just ath(4) and 802.11n. I _think_ I'm starting to understand how the adhoc node stuff works (ie, it's faking nodes through beacon discovery, rather than having actual association exchanged between them) but it's not clear exactly where to slice in the missing stuff. I also had no end of trouble trying to statically configure an 11a or 11na channel on an IBSS node. 11bgn channels are fine. I'd really appreciate someone stepping in here. I'm trying to finish off the power save queue handling changes and then move to finishing correcting the ps-poll support. Then I have the whole TX and RX data and control path to jiggle around to make things (more) predictable and better behaved. I really don't want to stray from this. :-) Do I have to promise free hardware and t-shirts? :-) Adrian ___ freebsd-wireless@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-wireless To unsubscribe, send any mail to "freebsd-wireless-unsubscr...@freebsd.org"
Re: misc/172955: [ath] 11n does not work in adhoc mode
On Tuesday 23 October 2012 09:33:42 Adrian Chadd wrote: > Ok- the problem here is that unfortunately the adhoc code doesn't do > any of the HT field parsing or "upgrading". > > Adding the parsing isn't enough, it still doesn't properly negotiate > HT on both ends, so it doesn't upgrade the rate control. > > Also - the WME parameters aren't parsed either, leading to your > observation about performance drops between 7x and 8.x. > > I'm ready to test any upgrades :-) Johann ___ freebsd-wireless@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-wireless To unsubscribe, send any mail to "freebsd-wireless-unsubscr...@freebsd.org"
Re: misc/172955: [ath] 11n does not work in adhoc mode
Ok- the problem here is that unfortunately the adhoc code doesn't do any of the HT field parsing or "upgrading". Adding the parsing isn't enough, it still doesn't properly negotiate HT on both ends, so it doesn't upgrade the rate control. Also - the WME parameters aren't parsed either, leading to your observation about performance drops between 7x and 8.x. Adrian ___ freebsd-wireless@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-wireless To unsubscribe, send any mail to "freebsd-wireless-unsubscr...@freebsd.org"
Re: misc/172955: [ath] 11n does not work in adhoc mode
The following reply was made to PR kern/172955; it has been noted by GNATS. From: Adrian Chadd To: Johann Hugo Cc: bug-follo...@freebsd.org Subject: Re: misc/172955: [ath] 11n does not work in adhoc mode Date: Mon, 22 Oct 2012 12:57:42 -0700 Right, it looks like the adhoc nodes are advertising HTCAP in the beacons; however it's not actually treating that node as HT capable. Something needs to "upgrade" those nodes to HT. Adrian ___ freebsd-wireless@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-wireless To unsubscribe, send any mail to "freebsd-wireless-unsubscr...@freebsd.org"
