af1dedf89c
This commit makes absolutely no change to the output unless debug is
manually turned on. Debug can be turned on with this other code change:
--- a/boards/xtensa/intel_adsp_cavs15/tools/logtool.py
+++ b/boards/xtensa/intel_adsp_cavs15/tools/logtool.py
@@ -45,7 +45,7 @@ def main():
else:
etrace = QEMU_ETRACE
- l = Loglist(etrace)
+ l = Loglist(etrace, True)
l.print()
Below are two examples of the new debug output. For instance the empty
and duplicate (due to a bug) slot [33] is now much more obvious:
}; {[31] : 1434444: ssp_set_config(),
}; {[32] : 1485591: ssp_set_config(),
}; {[33] : }; {[33] : 1539091: ssp_set_config(),
In this other example it is now much clearer that [7] misses a
`\n` newline:
}; {[6] : 858361: starting dma_trace_init_complete()
}; {[7] : dma_tr_...() ret=0[00000005] <inf> main}; {[8] : ipc_dma..()
}; {[9] : entering dma_trace_enable
}; {[10] : entering dma_trace_buffer_init()
Signed-off-by: Marc Herbert <marc.herbert@intel.com>
|
||
|---|---|---|
| .. | ||
| lib | ||
| README.md | ||
| adsplog.py | ||
| cavs-fw-v15.py | ||
| cavs-fw-v25.py | ||
| cavsload.sh | ||
| dump_trace.py | ||
| flash.sh | ||
| fw_loader.py | ||
| logtool.py | ||
| mbterm.py | ||
README.md
For various legacy reasons this directory has two similar log tools: logtool.py and adsplog.py
Both may be used in automation so merging them would require some coordination.
They both read from the same data from the exact same shared memory yet they have significant differences:
-
logtool.py reads /sys/kernel/debug/sof/etrace which requires the kernel driver to be loaded.
-
adsplog.py finds the memory address by scanning /sys/bus/pci/devices/; this does not require a driver.
-
logtool.py supports reading from a special QEMU location.
-
logtool.py performs a raw dump of the memory and exits immediately.
-
adsplog.py parses the data, understands the ring buffer and reads continuously. Its output is much more human-readable.
-
adsplog.py has technical details explained in a comment at the top.