hi, I am trying to measure the time when the frame is inserted into queue (ath_tx_process_buffer, ath_drain_txq_list[not req actually]) and when one receives ACK from hardware in ath_tx_process_buffer() in the descriptor, ath_tx_status and eventually the timestamp is calculated in ath_tx_complete_buf() by me.
But when I compute the airtime of the frame by using frame-size/bitrate,I get the airtime greater than the difference in the tsf values. My traces have this situation about 22% by volume count. For example, for a frame of size 1542 transmitted at 18Mbps, the airtime calculated is 685 [1542/18.0], while the ACK_time- enqueue_time ==318 ! Both the values are in microseconds. Is there a factor between the clock time and the airtime I calculate ? I am unable to understand why ? https://github.com/abnarain/mac80211-ath9k/blob/master/drivers/net/wireless/ath/ath9k/xmit.c#L2075 I have neat documented code above, which does the following : get the hw tsf in ath_tx_process_buffer() [also in ath_drain_txq_list(), which is hardly called], this is stored in bf->timestamp_temp. This is then used to combine the ath_tx_status->ts_tstamp 32 bit field to get the final tsf value, using the way tsf is calculated in the receive path (ath9k/recv.c) Will be great if someone can help - Abhinav
_______________________________________________ ath9k-devel mailing list ath9k-devel@lists.ath9k.org https://lists.ath9k.org/mailman/listinfo/ath9k-devel
