Clone of official tools

Revision:
40:7d3fa6b99b2b
Parent:
36:96847d42f010
Child:
41:2a77626a4c21
--- a/targets/REALTEK_RTL8195AM.py	Wed Jul 19 16:44:30 2017 -0500
+++ b/targets/REALTEK_RTL8195AM.py	Tue Oct 10 16:56:30 2017 -0500
@@ -1,141 +1,276 @@
 """
-mbed REALTEK_RTL8195AM elf2bin script
-Copyright (c) 2011-2016 Realtek Semiconductor Corp.
-
-Licensed under the Apache License, Version 2.0 (the "License");
-you may not use this file except in compliance with the License.
-You may obtain a copy of the License at
+Realtek Semiconductor Corp.
 
-    http://www.apache.org/licenses/LICENSE-2.0
-
-Unless required by applicable law or agreed to in writing, software
-distributed under the License is distributed on an "AS IS" BASIS,
-WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-See the License for the specific language governing permissions and
-limitations under the License.
-
-LIBRARIES BUILD
+RTL8195A elf2bin script
 """
 
 import sys, array, struct, os, re, subprocess
 import hashlib
+import shutil
 
 from tools.paths import TOOLS_BOOTLOADERS
 from datetime import datetime
 
 # Constant Variables
-RAM2_RSVD = 0x3131373835393138
+RAM2_RSVD = 0x00000000
+RAM2_VER = 0x8195FFFF00000000
+RAM2_TAG = 0x81950001
+RAM2_SHA = '0'
 
 def write_fixed_width_string(value, width, output):
-    # cut string to list & reverse
-    line = [value[i:i+2] for i in range(0, len(value), 2)]
-    output.write("".join([chr(long(b, 16)) for b in line]))
+        # cut string to list & reverse
+        line = [value[i:i+2] for i in range(0, len(value), 2)]
+        output.write("".join([chr(long(b, 16)) for b in line]))
 
 def write_fixed_width_value(value, width, output):
-    # convert to string
-    line = format(value, '0%dx' % (width))
-    if len(line) > width:
-        print "[ERROR] value 0x%s cannot fit width %d" % (line, width)
-        sys.exit(-1)
-    # cut string to list & reverse
-    line = [line[i:i+2] for i in range(0, len(line), 2)]
-    line.reverse()
-    # convert to write buffer
-    output.write("".join([chr(long(b, 16)) for b in line]))
+        # convert to string
+        line = format(value, '0%dx' % (width))
+        if len(line) > width:
+            print "[ERROR] value 0x%s cannot fit width %d" % (line, width)
+            sys.exit(-1)
+        # cut string to list & reverse
+        line = [line[i:i+2] for i in range(0, len(line), 2)]
+        line.reverse()
+        # convert to write buffer
+        output.write("".join([chr(long(b, 16)) for b in line]))
 
 def append_image_file(image, output):
     input = open(image, "rb")
     output.write(input.read())
     input.close()
 
-def prepend(image, image_prepend, toolchain, info):
-    output = open(image_prepend, "wb")  
-    write_fixed_width_value(info['size'], 8, output)
-    write_fixed_width_value(info['addr'], 8, output)
-    write_fixed_width_value(RAM2_RSVD, 16, output)
-    with open(image, "rb") as input:
-        if toolchain == "IAR":
-            input.seek(info['addr'])
-        output.write(input.read(info['size']))
+def write_padding_bytes(output_name, size):
+    current_size = os.stat(output_name).st_size
+    padcount = size - current_size
+    if padcount < 0:
+        print "[ERROR] image is larger than expected size"
+        sys.exit(-1)
+    output = open(output_name, "ab")
+    output.write('\377' * padcount)
     output.close()
 
-def parse_section(toolchain, elf, section):
-    info = {'addr':None, 'size':0};
-    if toolchain not in ["GCC_ARM", "ARM_STD", "ARM", "ARM_MICRO", "IAR"]:
-        print "[ERROR] unsupported toolchain " + toolchain
-        sys.exit(-1)
-    
-    mapfile = elf.rsplit(".", 1)[0] + ".map"
-    
-    with open(mapfile, 'r') as infile:
-        # Search area to parse
-        for line in infile:
-            if toolchain == "GCC_ARM":
-                # .image2.table   0x[00000000]30000000       0x18
-                #                 0x[00000000]30000000       __image2_start__ = .
-                #                 0x[00000000]30000000       __image2_entry_func__ = .
-                match = re.match(r'^' + section + \
-                        r'\s+0x0{,8}(?P<addr>[0-9A-Fa-f]{8})\s+0x(?P<size>[0-9A-Fa-f]+).*$', line)
-            elif toolchain in ["ARM_STD", "ARM", "ARM_MICRO"]:
-                # Memory Map of the image
-                #   Load Region LR_DRAM (Base: 0x30000000, Size: 0x00006a74, Max: 0x00200000, ABSOLUTE)
-                #     Execution Region IMAGE2_TABLE (Base: 0x30000000, Size: 0x00000018, Max: 0xffffffff, ABSOLUTE, FIXED)
-                #     Base Addr    Size         Type   Attr      Idx    E Section Name        Object
-                #     0x30000000   0x00000004   Data   RO         5257    .image2.ram.data    rtl8195a_init.o
-                match = re.match(r'^.*Region\s+' + section + \
-                        r'\s+\(Base: 0x(?P<addr>[0-9A-Fa-f]{8}),\s+Size: 0x(?P<size>[0-9A-Fa-f]+), .*\)$', line)
-            elif toolchain == "IAR":
-                #   Section                 Kind        Address     Size  Object
-                #   -------                 ----        -------     ----  ------
-                # "A3":                                           0x8470
-                #   IMAGE2                           0x10006000   0x5d18  <Block>
-                #     .ram_image2.text               0x10006000   0x5bbc  <Block>
-                #       .rodata             const    0x10006000     0x14  retarget.o [17]
-                match = re.match(r'^\s+' + section + \
-                        r'\s+0x(?P<addr>[0-9A-Fa-f]{8})\s+0x(?P<size>[0-9A-Fa-f]+)\s+.*<Block>$', line)
-            if match:
-                info['addr'] = int(match.group("addr"), 16)
-                try:
-                    info['size'] = int(match.group("size"), 16)
-                except IndexError:
-                    print "[WARNING] cannot find the size of section " + section
-                return info
+def sha256_checksum(filename, block_size=65536):
+    sha256 = hashlib.sha256()
+    with open(filename, 'rb') as f:
+        for block in iter(lambda: f.read(block_size), b''):
+            sha256.update(block)
+    return sha256.hexdigest()
 
-    print "[ERROR] cannot find the address of section " + section    
-    return info
+def get_version_by_time():
+    secs = int((datetime.now()-datetime(2016,11,1)).total_seconds())
+    return RAM2_VER + secs
 
 # ----------------------------
 #       main function
 # ----------------------------
-def rtl8195a_elf2bin(toolchain, image_elf, image_bin):
+def prepend(image, entry, segment, image_ram2, image_ota):
+
+    # parse input arguments
+    output = open(image_ram2, "wb")
+
+    write_fixed_width_value(os.stat(image).st_size, 8, output)
+    write_fixed_width_value(int(entry), 8, output)
+    write_fixed_width_value(int(segment), 8, output)
+
+    RAM2_SHA = sha256_checksum(image)
+    write_fixed_width_value(RAM2_TAG, 8, output)
+    write_fixed_width_value(get_version_by_time(), 16, output)
+    write_fixed_width_string(RAM2_SHA, 64, output)
+    write_fixed_width_value(RAM2_RSVD, 8, output)
+
+    append_image_file(image, output)
+    output.close()
+
+    ota = open(image_ota, "wb")
+    write_fixed_width_value(os.stat(image).st_size, 8, ota)
+    write_fixed_width_value(int(entry), 8, ota)
+    write_fixed_width_value(int(segment), 8, ota)
+    write_fixed_width_value(0xFFFFFFFF, 8, ota)
+    write_fixed_width_value(get_version_by_time(), 16, ota)
+    write_fixed_width_string(RAM2_SHA, 64, ota)
+    write_fixed_width_value(RAM2_RSVD, 8, ota)
+
+    append_image_file(image, ota)
+    ota.close()
+
+def find_symbol(toolchain, mapfile, symbol):
+    ret = None
+
+    HEX = '0x0{,8}(?P<addr>[0-9A-Fa-f]{8})'
     if toolchain == "GCC_ARM":
-        img2_sections = [".image2.table", ".text", ".data"]
+        SYM = re.compile(r'^\s+' + HEX + r'\s+' + symbol + '\r?$')
     elif toolchain in ["ARM_STD", "ARM", "ARM_MICRO"]:
-        img2_sections = [".image2.table", ".text", ".data"]
+        SYM = re.compile(r'^\s+' + HEX + r'\s+0x[0-9A-Fa-f]{8}\s+Code.*\s+i\.' + symbol + r'\s+.*$')
     elif toolchain == "IAR":
-        # actually it's block
-        img2_sections = ["IMAGE2"]
+        SYM = re.compile(r'^' + symbol + r'\s+' + HEX + '\s+.*$')
+
+    with open(mapfile, 'r') as infile:
+        for line in infile:
+            match = re.match(SYM, line)
+            if match:
+                ret = match.group("addr")
+
+    if not ret:
+        print "[ERROR] cannot find the address of symbol " + symbol
+        return 0
+
+    return int(ret,16) | 1
+
+def parse_load_segment_gcc(image_elf):
+    # Program Headers:
+    #   Type           Offset   VirtAddr   PhysAddr   FileSiz MemSiz  Flg Align
+    #   LOAD           0x000034 0x10006000 0x10006000 0x026bc 0x026bc RW  0x8
+    #   LOAD           0x0026f0 0x30000000 0x30000000 0x06338 0x06338 RWE 0x4
+    segment_list = []
+    cmd = 'arm-none-eabi-readelf -l ' + image_elf
+    for line in subprocess.check_output(cmd, shell=True, universal_newlines=True).split("\n"):
+        if not line.startswith("  LOAD"):
+            continue
+        segment = line.split()
+        if len(segment) != 8:
+            continue
+        offset = int(segment[1][2:], 16)
+        addr   = int(segment[2][2:], 16)
+        size   = int(segment[4][2:], 16)
+        if addr != 0 and size != 0:
+            segment_list.append((offset, addr, size))
+    return segment_list
+
+def parse_load_segment_armcc(image_elf):
+    # ====================================
+    #
+    # ** Program header #2
+    #
+    #     Type          : PT_LOAD (1)
+    #     File Offset   : 52 (0x34)
+    #     Virtual Addr  : 0x30000000
+    #     Physical Addr : 0x30000000
+    #     Size in file  : 27260 bytes (0x6a7c)
+    #     Size in memory: 42168 bytes (0xa4b8)
+    #     Flags         : PF_X + PF_W + PF_R + PF_ARM_ENTRY (0x80000007)
+    #     Alignment     : 8
+    #
+    (offset, addr, size) = (0, 0, 0)
+    segment_list = []
+    in_segment = False
+    cmd = 'fromelf --text -v --only=none ' + image_elf
+    for line in subprocess.check_output(cmd, shell=True, universal_newlines=True).split("\n"):
+        if line == "":
+            pass
+        elif line.startswith("** Program header"):
+            in_segment = True
+        elif in_segment == False:
+            pass
+        elif line.startswith("============"):
+            if addr != 0 and size != 0:
+                segment_list.append((offset, addr, size))
+            in_segment = False
+            (offset, addr, size) = (0, 0, 0)
+        elif line.startswith("    Type"):
+            if not re.match(r'\s+Type\s+:\s+PT_LOAD\s.*$', line):
+                in_segment = False
+        elif line.startswith("    File Offset"):
+            match = re.match(r'^\s+File Offset\s+:\s+(?P<offset>\d+).*$', line)
+            if match:
+                offset = int(match.group("offset"))
+        elif line.startswith("    Virtual Addr"):
+            match = re.match(r'^\s+Virtual Addr\s+:\s+0x(?P<addr>[0-9a-f]+).*$', line)
+            if match:
+                addr = int(match.group("addr"), 16)
+        elif line.startswith("    Size in file"):
+            match = re.match(r'^\s+Size in file\s+:.*\(0x(?P<size>[0-9a-f]+)\).*$', line)
+            if match:
+                size = int(match.group("size"), 16)
+    return segment_list
+
+
+def parse_load_segment_iar(image_elf):
+    #   SEGMENTS:
+    #
+    #      Type Offset    Virtual   Physical File Sz Mem Sz Flags   Align
+    #      ---- ------    -------   -------- ------- ------ -----   -----
+    #   0: load   0x34 0x10006000 0x10006000  0x26bc 0x26bc 0x6  WR   0x8
+    #   1: load 0x26f0 0x30000000 0x30000000  0x6338 0x6338 0x7 XWR   0x4
+    #
+    #   SECTIONS:
+    #
+    #       Name            Type     Addr         Offset    Size Aln Lnk Inf  ESz Flags
+    #       ----            ----     ----         ------    ---- --- --- ---  --- -----
+    #    1: .shstrtab       strtab               0xfc4d8    0x60 0x4
+    #    2: .strtab         strtab               0xfc538  0xbb3f 0x4
+
+    segment_list = []
+    in_segment = False
+    cmd = 'ielfdumparm ' + image_elf
+    for line in subprocess.check_output(cmd, shell=True, universal_newlines=True).split("\n"):
+        if line.startswith("  SEGMENTS:"):
+            in_segment = True
+        elif in_segment == False:
+            pass
+        elif line.startswith("  SECTIONS:"):
+            break
+        elif re.match(r'^\s+\w+:\s+load\s+.*$', line):
+            segment = line.split()
+            offset = int(segment[2][2:], 16)
+            addr   = int(segment[3][2:], 16)
+            size   = int(segment[5][2:], 16)
+            if addr < 0x10007000:
+                continue
+            if addr != 0 and size != 0:
+                segment_list.append((offset, addr, size))
+    return segment_list
+
+def parse_load_segment(toolchain, image_elf):
+    if toolchain == "GCC_ARM":
+        return parse_load_segment_gcc(image_elf)
+    elif toolchain in ["ARM_STD", "ARM", "ARM_MICRO"]:
+        return parse_load_segment_armcc(image_elf)
+    elif toolchain == "IAR":
+        return parse_load_segment_iar(image_elf)
     else:
-        print("[error] unsupported toolchain") + toolchain
-        return
-    ram2_info = {'addr':None, 'size':0}
+        return []
+
+def write_load_segment(image_elf, image_bin, segment):
+    file_elf = open(image_elf, "rb")
+    file_bin = open(image_bin, "wb")
+    for (offset, addr, size) in segment:
+        file_elf.seek(offset)
+        # write image header - size & addr
+        write_fixed_width_value(addr, 8, file_bin)
+        write_fixed_width_value(size, 8, file_bin)
+        # write load segment
+        file_bin.write(file_elf.read(size))
+        delta = size % 4
+        if delta != 0:
+            padding = 4 - delta
+            write_fixed_width_value(0x0, padding * 2, file_bin)
+    file_bin.close()
+    file_elf.close()
+
+# ----------------------------
+#       main function
+# ----------------------------
+def rtl8195a_elf2bin(t_self, image_elf, image_bin):
+    # remove target binary file/path
+    if os.path.isfile(image_bin):
+        os.remove(image_bin)
+    else:
+        shutil.rmtree(image_bin)
+
+    segment = parse_load_segment(t_self.name, image_elf)
+    write_load_segment(image_elf, image_bin, segment)
+
     image_name = os.path.splitext(image_elf)[0]
+    image_map = image_name + '.map'
 
-    ram1_prepend_bin = os.path.join(TOOLS_BOOTLOADERS, "REALTEK_RTL8195AM", "ram_1_prepend.bin")
-    ram2_prepend_bin = image_name + '-ram_2_prepend.bin'
-    
-    old_bin = image_name + '.bin'
-    for section in img2_sections:
-        section_info = parse_section(toolchain, image_elf, section)
-        if ram2_info['addr'] is None or ram2_info['addr'] > section_info['addr']:
-            ram2_info['addr'] = section_info['addr']
-        ram2_info['size'] = ram2_info['size'] + section_info['size']
+    ram2_ent = find_symbol(t_self.name, image_map, "PLAT_Start")
+    ram1_bin = os.path.join(TOOLS_BOOTLOADERS, "REALTEK_RTL8195AM", "ram_1.bin")
+    ram2_bin = image_name + '-ram_2.bin'
+    ota_bin = image_name + '-ota.bin'
+    prepend(image_bin, ram2_ent, len(segment), ram2_bin, ota_bin)
 
-    prepend(old_bin, ram2_prepend_bin, toolchain, ram2_info)
     # write output file
     output = open(image_bin, "wb")
-    append_image_file(ram1_prepend_bin, output)
-    append_image_file(ram2_prepend_bin, output)
+    append_image_file(ram1_bin, output)
+    append_image_file(ram2_bin, output)
     output.close()
-    # post built done
-