This patch enables the support for parsing pktlog from trace
points taken on a QCA9888 & QCA9984 chipsets.
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
Updating the hw enumeration here to keep in sync with ath10k.
kvalo: ath10k accidentally broke the backwards compatibility in
hw revision in these commits:
e565c3125e03 ath10k: enable support for QCA9888
651b4cdcf97e ath10k: enable support for QCA9984
But this should not happen anymore. I added a comment to document
that.
With "ath10k: simplify pktlog htt event processing" approach, driver passes
entire payload stipping the htt_response header. By this method, we could
receive pktlog events that has payload size higher than mentioned in pktlog
header (for alignment reasons)
This issue is seen with 10.2 FW for ATH10K_PKTLOG_TYPE_TX_MSDU_ID type events.
So throw exception only when payload size is lesser than the one mentioned in
pktlog header
txctl log type manipulates pktlog header and log data as well. payload
length should be handled differently for 10.2 and 10.4 based firmware.
Not doing so, will generate invalid pktlog data file.
Currently tracing plugin works only for QCA988X chipsets. This patch
extend the support to generate pktlog for QCA99X0.
With the ath10k commit "ath10k: add hw_rev to trace events to support
pktlog" we now have a new attribute passed with tracing events
"hw_type" that helps to determine the chipset used for taking the trace
points. This is helpful in getting FW variant used thereby generating
pktlog.dat accordingly.
To support backward compatibility for older versions of ath10k, assume the
default value of the hw_type, if not provided, to be HW_QCA988X (0).
ath10k_htt_rx_desc has QCA988X,QCA99X0 and QCA6174 packet information.
ath10k_htt_rx_desc has QCA99X0 relted structures at various offsets. The current
script has not taken care of QCA99X0 changes.
This modification has fix for QCA988X chipsets.
We will fix this script for all the supported chipsets later by sending
one more trcing attribute from the driver.
Whenever trace record is started in middle of data flow, the
msdu_len_tbl is not fully populated. For the initial MSDU_ID log
types, msdu_id can not be found in msdu_len_tbl. If the entries
are not found set the msdu_len to 0. This will also fix below
errors.
File "ath10k_pktlog.py", line 297, in <lambda>
ath10k_htt_pktlog_handler(pevent, *args))
File "ath10k_pktlog.py", line 238, in ath10k_htt_pktlog_handler
pktlog_tx_msdu_id(buf)
File "ath10k_pktlog.py", line 218, in pktlog_tx_msdu_id
msdu_len = msdu_len_tbl[msdu_id]
Since tsf is valid only for last mpdu and it is already part
of rx descriptor dump, tsf argument is removed from
ath10k_htt_rx_desc tracepoint in ath10k. Hence updating the
parser accordingly.
