2011-03-09 01:49:56

by Daniel Halperin

Subject: [PATCH] mac80211: update minstrel_ht sample rate when probe is set

Waiting until the status is received can cause the same rate to be
probed multiple times consecutively.

Cc: Felix Fietkau <[email protected]>
Signed-off-by: Daniel Halperin <[email protected]>
---
net/mac80211/rc80211_minstrel_ht.c | 15 +++++----------
1 files changed, 5 insertions(+), 10 deletions(-)

diff --git a/net/mac80211/rc80211_minstrel_ht.c
b/net/mac80211/rc80211_minstrel_ht.c
index 775cf15..bce14fb 100644
--- a/net/mac80211/rc80211_minstrel_ht.c
+++ b/net/mac80211/rc80211_minstrel_ht.c
@@ -415,10 +415,8 @@ minstrel_ht_tx_status(void *priv, struct
ieee80211_supported_band *sband,
mi->sample_count--;
}

- if (info->flags & IEEE80211_TX_CTL_RATE_CTRL_PROBE) {
+ if (info->flags & IEEE80211_TX_CTL_RATE_CTRL_PROBE)
mi->sample_packets += info->status.ampdu_len;
- minstrel_next_sample_idx(mi);
- }

for (i = 0; !last; i++) {
last = (i == IEEE80211_TX_MAX_RATES - 1) ||
@@ -551,13 +549,14 @@ minstrel_get_sample_rate(struct minstrel_priv
*mp, struct minstrel_ht_sta *mi)
sample_idx = sample_table[mg->column][mg->index];
mr = &mg->rates[sample_idx];
sample_idx += mi->sample_group * MCS_GROUP_RATES;
+ minstrel_next_sample_idx(mi);

/*
* When not using MRR, do not sample if the probability is already
* higher than 95% to avoid wasting airtime
*/
if (!mp->has_mrr && (mr->probability > MINSTREL_FRAC(95, 100)))
- goto next;
+ return -1;

/*
* Make sure that lower rates get sampled only occasionally,
@@ -566,17 +565,13 @@ minstrel_get_sample_rate(struct minstrel_priv
*mp, struct minstrel_ht_sta *mi)
if (minstrel_get_duration(sample_idx) >
minstrel_get_duration(mi->max_tp_rate)) {
if (mr->sample_skipped < 20)
- goto next;
+ return -1;

if (mi->sample_slow++ > 2)
- goto next;
+ return -1;
}

return sample_idx;
-
-next:
- minstrel_next_sample_idx(mi);
- return -1;
}

static void
--
1.7.0.4

2011-03-09 03:34:17

by Felix Fietkau

Subject: Re: [PATCH] mac80211: update minstrel_ht sample rate when probe is set

On 2011-03-09 2:41 AM, Daniel Halperin wrote:
> Waiting until the status is received can cause the same rate to be
> probed multiple times consecutively.
>
> Cc: Felix Fietkau <[email protected]>
> Signed-off-by: Daniel Halperin <[email protected]>
Mangled by line wrapping, please resend. Looks good otherwise.

- Felix