On 11/04/2011 09:24 AM, Johannes Berg wrote:
On Fri, 2011-11-04 at 09:17 -0700, Ben Greear wrote:
The AMPDU density should only allow increasing.
I think a lot of this validation should live in cfg80211 so if another
driver wants to implement it, this kind of thing is already covered.
The ath9k driver supports 0, and then every value that corresponds to 1us or higher.
If you set it to 1/2us, for instance, it just quietly rounds up to 1us. It defaults
to 8, so it appears valid to decrease or increase this value.
What quietly rounds up?
If it defaults to 8 then I'm sure there's a reason for it, such as the
crypto engine not being fast enough and needing 8us buffer between
frames. As such, I really don't think decreasing it is valid.
See this code in ath9k, top of main.c. It appears to support more
than the default of 8. I tested it out, and it appears to work
when set to lower values. I am disabling hw-crypt since I need
multiple VIFS, but not sure that matters or not.
static u8 parse_mpdudensity(u8 mpdudensity)
Well, this is used for TX. You're advertising the value for RX.
Advertising a smaller value may work for you -- but only if the AP uses
something larger anyway. It's free to do this. The code you quoted
implements this -- it looks at what the min spacing is the peer wants,
and then uses something bigger.
So I still stand by what I said earlier -- you should not allow
advertising a smaller value than the hardware wanted to use initially.
If the AP doesn't use that smaller value, all is well, but if it
actually uses a smaller value then the hardware might fall over.
Ok, I'll add that restriction. When this all settles, maybe I
can get some info from the ath9k folks about what their hardware
can actually support.
Thanks,
Ben
johannes
--
Ben Greear <greearb@xxxxxxxxxxxxxxx>
Candela Technologies Inc http://www.candelatech.com
--
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
