CHROMIUMOS: compat-wireless: fixup rate control configuration

CHROMIUMOS: compat-wireless: fixup rate control configuration

The intended rate control configuration was lost in the recent update.
Enable ath9k and minstrel_ht rate control modules and change the default
algorithm to minstrel_ht. Remove the configuration of the default algorithm
used by the in-kernel code; it's not used by compat wireless and just
causes build complaints.

Signed-off-by: sleffler@chromium.org

Change-Id: I842060b3362b603a993d3a23911e535230d537f2

BUG=none (complaints about recent throughput drops)
TEST=TBA

Committed: http://chrome-svn/viewvc/chromeos?view=rev&revision=88a7dc5

[Sam Leffler, 2011-04-25 15:53:49]

Bring back the ath9k rc support that we're using (lost in the compat wireless
update).  This fixes regression in handling upstream traffic (legacy rates, no
ampdu).

Bring in minstrel_ht in case we want to run tests and for other drivers and make
it the default (been working ok on our openwrt configs and should be a superset
of minstrel).

[mukesh agrawal, 2011-04-25 17:41:51]

The config options set in the patch make sense, and the before/after performance
numbers are in the right ballpark. But could you help me understand where the
regression came from (see comment on line 56)?

---
performance numbers with patch, from Sam's testing:

client_iperf_00_throughput                         69.8
network_WiFiPerf/network_WiFiPerf.070Perf5HT20       PASS
network_WiFiPerf/network_WiFiPerf.071Perf5HT40       PASS
client_iperf_00_throughput                         116.0

before patch, from Sam's testing:  23-24 Mb/s

http://codereview.chromium.org/6880162/diff/1/chromeos/scripts/compat_wireles...
File chromeos/scripts/compat_wireless_config (right):

http://codereview.chromium.org/6880162/diff/1/chromeos/scripts/compat_wireles...
chromeos/scripts/compat_wireless_config:56: CONFIG_ATH9K_RATE_CONTROL=y
I see that we enable ath9k rate control here. But I'm curious why we do it
differently in this kernel, compared to kernel-next (where the config item is
set in config.common.i386 (and x86_64).
Is the reason for the difference that compat_wireless doesn't pay attention to
that file?

Or, perhaps what I'm really asking is how compat_wireless gets its config
settings. AFAICT, compat_wireless_config has never had CONFIG_ATH9K_RATE_CONTROL
set before. So did I miss a chnage in this file, or was the regression due to a
change in some other file?

[Sam Leffler, 2011-04-25 18:26:41]

http://codereview.chromium.org/6880162/diff/1/chromeos/scripts/compat_wireles...
File chromeos/scripts/compat_wireless_config (right):

http://codereview.chromium.org/6880162/diff/1/chromeos/scripts/compat_wireles...
chromeos/scripts/compat_wireless_config:56: CONFIG_ATH9K_RATE_CONTROL=y
We're using compat-wireless to bring in newer wifi support.  That has a separate
config mechanism than kernel-next and different config knobs to avoid conflict
w/ the in-kernel code.  The config mechanism is in the file 
scripts/compat_wireless_config which is run by the kernel ebuild. I haven't
looked at kernel-next to see if it's config setup is correct; we aren't using it
yet for x86 and for arm things are different.

The reason we lost the ath9k rc module was the change to
drivers/net/wireless/ath/ath9k/Makefile that moved rc.o from a standard builtin
component to an optional component.  Here's the relevant snippet of the diff:

git diff c45eaff59eae2b9b8b9921b8ce580daf100b99fa Makefile
diff --git a/chromeos/compat-wireless/drivers/net/wireless/ath/ath9k/Makefile b/
index 973ae4f..aca0162 100644
--- a/chromeos/compat-wireless/drivers/net/wireless/ath/ath9k/Makefile
+++ b/chromeos/compat-wireless/drivers/net/wireless/ath/ath9k/Makefile
@@ -5,8 +5,8 @@ ath9k-y +=      beacon.o \
                recv.o \
                xmit.o \
                virtual.o \
-               rc.o
 
+ath9k-$(CONFIG_ATH9K_RATE_CONTROL) += rc.o

[Paul Stewart, 2011-04-25 19:10:34]

LGTM.  Still confused as to why the sysfs file was wrong.  I did a check on
kernel next and its sysfs entry says:


localhost ~ # cat /sys/module/mac80211/parameters/ieee80211_default_rc_algo
minstrel_ht


so I'll need to do some fixing there, but that doesn't affect this change.

[mukesh agrawal, 2011-04-25 20:12:59]

LGTM.

http://codereview.chromium.org/6880162/diff/1/chromeos/scripts/compat_wireles...
File chromeos/scripts/compat_wireless_config (right):

http://codereview.chromium.org/6880162/diff/1/chromeos/scripts/compat_wireles...
chromeos/scripts/compat_wireless_config:56: CONFIG_ATH9K_RATE_CONTROL=y
Cool -- makes sense now.