static u16 iwl_limit_dwell(struct iwl_priv *priv, u16 dwell_time)
{
struct iwl_rxon_context *ctx;
+ int limits[NUM_IWL_RXON_CTX] = {};
+ int n_active = 0;
+ u16 limit;
+
+ BUILD_BUG_ON(NUM_IWL_RXON_CTX != 2);
/*
* If we're associated, we clamp the dwell time 98%
- * of the smallest beacon interval (minus 2 * channel
- * tune time)
+ * of the beacon interval (minus 2 * channel tune time)
+ * If both contexts are active, we have to restrict to
+ * 1/2 of the minimum of them, because they might be in
+ * lock-step with the time inbetween only half of what
+ * time we'd have in each of them.
*/
for_each_context(priv, ctx) {
- u16 value;
-
switch (ctx->staging.dev_type) {
case RXON_DEV_TYPE_P2P:
/* no timing constraints */
break;
}
- value = ctx->beacon_int;
- if (!value)
- value = IWL_PASSIVE_DWELL_BASE;
- value = (value * 98) / 100 - IWL_CHANNEL_TUNE_TIME * 2;
- dwell_time = min(value, dwell_time);
+ limits[n_active++] = ctx->beacon_int ?: IWL_PASSIVE_DWELL_BASE;
}
- return dwell_time;
+ switch (n_active) {
+ case 0:
+ return dwell_time;
+ case 2:
+ limit = (limits[1] * 98) / 100 - IWL_CHANNEL_TUNE_TIME * 2;
+ limit /= 2;
+ dwell_time = min(limit, dwell_time);
+ /* fall through to limit further */
+ case 1:
+ limit = (limits[0] * 98) / 100 - IWL_CHANNEL_TUNE_TIME * 2;
+ limit /= n_active;
+ return min(limit, dwell_time);
+ default:
+ WARN_ON_ONCE(1);
+ return dwell_time;
+ }
}
static u16 iwl_get_passive_dwell_time(struct iwl_priv *priv,