linux.git/net/mac80211/main.c, branch v4.8

mac80211: implement fair queueing per txq

2016-06-09T09:34:08+00:00

mac80211's software queues were designed to work
very closely with device tx queues. They are
required to make use of 802.11 packet aggregation
easily and efficiently.

Due to the way 802.11 aggregation is designed it
only makes sense to keep fair queuing as close to
hardware as possible to reduce induced latency and
inertia and provide the best flow responsiveness.

This change doesn't translate directly to
immediate and significant gains. End result
depends on driver's induced latency. Best results
can be achieved if driver keeps its own tx
queue/fifo fill level to a minimum.

Signed-off-by: Michal Kazior 
Signed-off-by: Johannes Berg

mac80211: skip netdev queue control with software queuing

2016-06-09T09:31:27+00:00

Qdiscs are designed with no regard to 802.11
aggregation requirements and hand out
packet-by-packet with no guarantee they are
destined to the same tid. This does more bad than
good no matter how fairly a given qdisc may behave
on an ethernet interface.

Software queuing used per-AC netdev subqueue
congestion control whenever a global AC limit was
hit. This meant in practice a single station or
tid queue could starve others rather easily. This
could resonate with qdiscs in a bad way or could
just end up with poor aggregation performance.
Increasing the AC limit would increase induced
latency which is also bad.

Disabling qdiscs by default and performing
taildrop instead of netdev subqueue congestion
control on the other hand makes it possible for
tid queues to fill up "in the meantime" while
preventing stations starving each other.

This increases aggregation opportunities and
should allow software queuing based drivers
achieve better performance by utilizing airtime
more efficiently with big aggregates.

Signed-off-by: Michal Kazior 
Signed-off-by: Johannes Berg

cfg80211: remove enum ieee80211_band

2016-04-12T13:56:15+00:00

This enum is already perfectly aliased to enum nl80211_band, and
the only reason for it is that we get IEEE80211_NUM_BANDS out of
it. There's no really good reason to not declare the number of
bands in nl80211 though, so do that and remove the cfg80211 one.

Signed-off-by: Johannes Berg

mac80211: add NETIF_F_RXCSUM to features white list

2016-04-05T09:45:51+00:00

NETIF_F_RXCSUM is not in the white list, though some
drivers may want to set it in order to enable seeing the
actual RX checksum status in ethtool.

Signed-off-by: Sara Sharon 
Signed-off-by: Emmanuel Grumbach 
Signed-off-by: Johannes Berg

mac80211: Set global RRM capability

2016-04-05T09:44:46+00:00

Allow publishing RRM capabilities for features that are not
HW dependent.

Signed-off-by: Emmanuel Grumbach 
Signed-off-by: Johannes Berg

mac80211: handle sched_scan_stopped vs. hw restart race

2016-01-14T10:10:20+00:00

On hw restart, mac80211 might try to reconfigure already
stopped sched scan, if ieee80211_sched_scan_stopped_work()
wasn't scheduled yet.

This in turn will keep the device driver with scheduled scan
configured, while both mac80211 and cfg80211 will clear
their sched scan state once the work is scheduled.

Fix it by ignoring ieee80211_sched_scan_stopped() calls
while in hw restart, and flush the work before starting
the reconfiguration.

Signed-off-by: Eliad Peller 
Signed-off-by: Emmanuel Grumbach 
Signed-off-by: Johannes Berg

mac80211: avoid ROC during hw restart

2016-01-14T10:10:14+00:00

Defer ROC requests during hw restart, as the driver
might not be fully configured in this stage (e.g.
channel contexts were not added yet)

Signed-off-by: Eliad Peller 
Signed-off-by: Emmanuel Grumbach 
Signed-off-by: Johannes Berg

mac80211: rewrite remain-on-channel logic

2015-12-04T13:43:32+00:00

Jouni found a bug in the remain-on-channel logic: when a short item
is queued, a long item is combined with it extending the original
one, and then the long item is deleted, the timeout doesn't go back
to the short one, and the short item ends up taking a long time. In
this case, this showed as blocking scan when running two test cases
back to back - the scan from the second was delayed even though all
the remain-on-channel items should long have been gone.

Fixing this with the current data structures turns out to be a bit
complicated, we just remove the long item from the dependents list
right now and don't recalculate the timeouts.

There's a somewhat similar bug where we delete the short item and
all the dependents go with it; to fix this we'd have to move them
from the dependents to the real list.

Instead of trying to do that, rewrite the code to not have all this
complexity in the data structures: use a single list and allow more
than one entry in it being marked as started. This makes the code a
bit more complex, the worker needs to understand that it might need
to just remove one of the started items, while keeping the device
off-channel, but that's not more complicated than the nested data
structures.

This then fixes both issues described, and makes it easier to also
limit the overall off-channel time when combining.

TODO: as before, with hardware remain-on-channel, deleting an item
after combining results in cancelling them all - we can keep track
of the time elapsed and only cancel after that to fix this.

Signed-off-by: Johannes Berg

Revert "mac80211: don't advertise NL80211_FEATURE_FULL_AP_CLIENT_STATE"

2015-12-04T13:43:32+00:00

This reverts commit 45bb780a2147b9995f3d288c44ecb87ca8a330e2,
the previous two patches fixed the functionality.

Signed-off-by: Johannes Berg

mac80211: don't advertise NL80211_FEATURE_FULL_AP_CLIENT_STATE

2015-11-17T14:48:50+00:00

For now, this feature doesn't actually work. To avoid shipping a
kernel that has it enabled but where it can't be used disable it
for now - we can re-enable it when it's fixed.

This partially reverts 44674d9c2267 ("mac80211: advertise support
for full station state in AP mode").

Cc: Ayala Beker 
Signed-off-by: Johannes Berg