adding CPU Trace support

[shmuelfomberg]

is it possible to add support for CPU Trace? (new in M4 machines)

I see that inside the trace file it is under tags '0a00010000000000' (in hex)

and under that tags there are multiple files, each with the name in the beginning.

[matan1008]

Sure, could you please supply raw kdebug data for tests?

[shmuelfomberg]

test_trace-010.atrc.zip

created by the "trace" command. here is some python code that I added to pykdebugparser/kd_buf_parser.py, in parse_v3, trying to parse the data:

TRACEV3_MACH_SYSCALLS = b'\x01\x20\x01\x00\x00\x00\x00\x00' TRACEV3_TRACE_STATISTICS = b'\x16\x80\x00\x00\x01\x00\x00\x00'

        elif block.tag == TRACEV3_MACH_SYSCALLS:
            mach_syscalls_data = json.loads(block.data)
            mach_syscalls = mach_syscalls_data["mach_syscalls"]
            # mach_syscalls is an array, each is a dict such as:
            # {
            #   "number": 12,
            #   "name": "mach_vm_deallocate",
            #   "arguments": ["mach_port_name_t target", "mach_vm_address_t address", "mach_vm_size_t size"]
            # }
        elif block.tag == TRACEV3_TRACE_STATISTICS:
            # a json contain satistics of the trace. maybe interesting keys:
            # "kdebugEventsSize" "chunkCount" "recordingInstructions" 
            pass
        elif block.tag.hex() in ['2080000000000000', 'ff51000000000000', '0880000000000000', 'fa51000000000000', '0250000001000100']:
            #print("============ TAG (plist): " + block.tag.hex())
            #print(plistlib.loads(block.data))
            pass
            # 2080000000000000 is empty
            # ff51000000000000 is about IOKit, huge amount of data
            # 0880000000000000 is display data
            # fa51000000000000 is empty
            # 0250000001000100 is about AGXDriverInfo
        elif block.tag.hex() == '0a00010000000000':
            # each is block of 16324 bytes, filled with leading zeros
            print("============ TAG: " + block.tag.hex())
            data = block.data
            ix = data.index(b'}')
            key = json.loads(data[0:ix+1])["Key"]
            print(key)
            data = data[ix+1:]
            ix = find_first_non_zero_byte(data)
            if ix is None:
                print("[Total size {} Leading Zeros {} Data left {}]".format(len(data), len(data), 0))
            else:
                print("[Total size {} Leading Zeros {} Data left {}]".format(len(data), ix, len(data[ix:])))
                data = data[ix:] 
                if key == "manifest.json":
                    aData = data.decode()
                    aData.replace("\\n", "\n")
                    print(aData)
                if 'UnitMarks' in key:
                    for ix in range(int(len(data)/24)):
                        start = ix * 24
                        print("{} {} {}".format(data[start:start+8].hex(), data[start+8:start+16].hex(), data[start+16:start+24].hex())) 
                if 'Ranges' in key:
                    if len(data) % 8 != 0:
                        data = b'\x00'*(8 - (len(data) % 8)) + data
                    for ix in range(int(len(data)/24)):
                        start = ix * 24
                        address1 = hex(int.from_bytes(data[start:start+4][::-1], byteorder='big') << 2)
                        flags1 = data[start+4:start+12].hex()
                        address2 = hex(int.from_bytes(data[start+12:start+16][::-1], byteorder='big') << 2)
                        third = data[start+16:start+24].hex()
                        print("{} {}".format(address1, flags1)) 
                        print("{} {}".format(address2, third)) 
                else:
                    print(data)
        else:
            print("============ TAG: " + block.tag.hex())
            print(block.data)

[matan1008]

Sorry for the delay, do you have any resources about how to parse UnitMarks and Ranges?

[shmuelfomberg]

sorry, I don't have anything.