coredump: implement host parser

coredump_parser/Makefile

@@ -0,0 +1,21 @@
+#
+# Makefile for Phoenix-RTOS coredump parser
+#
+# Copyright 2025 Phoenix Systems
+#
+
+NAME := coredump_parser
+LOCAL_DIR := $(call my-dir)
+SRCS := $(wildcard $(LOCAL_DIR)*.cpp)
+LOCAL_LDLIBS := -lstdc++
+SCRIPTS := $(wildcard $(LOCAL_DIR)*.py)
+
+
+include $(binary.mk)
+
+
+$(NAME)-scripts:
+	mkdir -p $(PREFIX_BUILD)/scripts
+	cp $(SCRIPTS) $(PREFIX_BUILD)/scripts
+
+ALL_COMPONENTS += $(NAME)-scripts

coredump_parser/gdb-phoenix-core.py

@@ -0,0 +1,223 @@
+# Phoenix-RTOS
+#
+# Coredump parser - gdb script
+#
+# Copyright 2025 Phoenix Systems
+# Authors: Jakub Klimek
+#
+# This file is part of Phoenix-RTOS.
+#
+# %LICENSE%
+#
+
+from elftools.elf.elffile import ELFFile
+from elftools.elf.constants import SH_FLAGS
+from itertools import pairwise
+from pathlib import Path
+import argparse
+import struct
+import sys
+import os
+import subprocess
+import tempfile
+
+NT_LMA = 0x00414D4C
+
+COREDUMP_PARSER_PATH = (Path(__file__).parent.parent / "prog" / "coredump_parser").absolute()
+COREDUMP_PARSER_OUTPUT = Path(tempfile.gettempdir()) / "phoenix-coredumps"
+
+os.makedirs(COREDUMP_PARSER_OUTPUT, exist_ok=True)
+
+
+def has_lma_note(elffile):
+    for segment in elffile.iter_segments():
+        if segment['p_type'] != 'PT_NOTE':
+            continue
+        for note in segment.iter_notes():
+            if note['n_type'] == NT_LMA:
+                return True
+    return False
+
+def parse_core_file(elffile):
+    for segment in elffile.iter_segments():
+        if segment['p_type'] != 'PT_NOTE':
+            continue
+        for note in segment.iter_notes():
+            if note['n_type'] != NT_LMA:
+                continue
+            values = struct.unpack(f'<{len(note["n_desc"]) // 4}I', note['n_desc'])
+            res = {}
+            for lma, vma in pairwise(values):
+                res[vma] = lma
+            return res
+    return {}
+
+def parse_symbol_file(elffile):
+    section_to_segment_mapping = {}
+
+    for section in elffile.iter_sections():
+        if not (section['sh_flags'] & SH_FLAGS.SHF_ALLOC) or section['sh_size'] == 0:
+            continue
+
+        sec_start_addr = section['sh_addr']
+        sec_end_addr = sec_start_addr + section['sh_size']
+
+        for segment in elffile.iter_segments():
+            if segment['p_type'] != 'PT_LOAD':
+                continue
+            seg_start_addr = segment['p_vaddr']
+            seg_end_addr = seg_start_addr + segment['p_memsz']
+
+            if sec_start_addr >= seg_start_addr and sec_end_addr <= seg_end_addr:
+                section_to_segment_mapping[section.name] = {
+                    'seg_addr': seg_start_addr,
+                    'offset': section['sh_addr'] - seg_start_addr,
+                }
+                break
+
+    return section_to_segment_mapping
+
+def create_mapping_args(core_file, symbol_file):
+    core_segment_mapping = parse_core_file(core_file)
+    elf_section_mapping = parse_symbol_file(symbol_file)
+    args = []
+    for section_name, section_info in elf_section_mapping.items():
+        if section_info['seg_addr'] not in core_segment_mapping:
+            print(f"Warning: Segment containing '{section_name}' not found in core NT_FILE mapping.", file=sys.stderr)
+            if section_name == '.text':
+                print("Warning: No valid segment mapping found for '.text' section!", file=sys.stderr)
+                return []
+            continue
+        new_addr = core_segment_mapping[section_info['seg_addr']] + section_info['offset']
+        args.append(f"-s {section_name} {new_addr:#x}")
+    if len(args) == 0:
+        print("Error: No valid segment mappings found.", file=sys.stderr)
+    return args
+
+def run_parser(coredump=None, symbolfile=None):
+    arglist = [COREDUMP_PARSER_OUTPUT]
+    if symbolfile: arglist.append(Path(symbolfile).name)
+    try:
+        if coredump:
+            with open(coredump, "rb") as f:
+                result = subprocess.run([COREDUMP_PARSER_PATH, *arglist], check=True, stdin=f, stdout=subprocess.PIPE)
+                return result.stdout.decode('utf-8').strip()
+        else:
+            result = subprocess.run([COREDUMP_PARSER_PATH, *arglist], check=True, stdout=subprocess.PIPE)
+            return result.stdout.decode('utf-8').strip()
+    except subprocess.CalledProcessError as e:
+        print(f"Running coredump_parser failed with exit status: {e.returncode}", file=sys.stderr)
+    return []
+
+def generate_gdb_commands(args):
+    """
+    If symbols is None, only corefile will be loaded
+    If core is None, corefile will be loaded from stdin
+    """
+    coreelf = None
+    symbolelf = None
+    corefile = None
+    symbolfile = None
+
+    try:
+        corefile = open(args.core, 'rb')
+        coreelf = ELFFile(corefile)
+    except Exception as e:
+        args.core = run_parser(args.core, args.symbol)
+        if args.core:
+            print(f"Using core file '{args.core}' with symbol file '{args.symbol}'.", file=sys.stderr)
+            corefile = open(args.core, 'rb')
+            coreelf = ELFFile(corefile)
+
+    if args.symbol:
+        symbolfile = open(args.symbol, 'rb')
+        symbolelf = ELFFile(symbolfile)
+
+    if not coreelf:
+        print("No core file found.", file=sys.stderr)
+        if corefile: corefile.close()
+        if symbolfile: symbolfile.close()
+        return []
+
+    commands = []
+    if has_lma_note(coreelf):
+        if not symbolelf:
+            print("Warning: NOMMU coreelf detected, but no symbol file found. Symbols won't be resolved properly when loading symbols separately!", file=sys.stderr)
+        else:
+            mapping_args = create_mapping_args(coreelf, symbolelf)
+            commands.append(f"symbol-file") # clear previous
+            commands.append(f"exec-file") # clear previous
+            commands.append(f"add-symbol-file {args.symbol} {' '.join(mapping_args)}")
+            commands.append(f"exec-file {args.symbol} {' '.join(mapping_args)}")
+    elif symbolelf:
+        commands.append(f"symbol-file") # clear previous
+        commands.append(f"exec-file") # clear previous
+        commands.append(f"add-symbol-file {args.symbol}")
+        commands.append(f"exec-file {args.symbol}")
+
+    commands.append(f"core-file {args.core}")
+    if corefile: corefile.close()
+    if symbolfile: symbolfile.close()
+    return commands
+
+def parse_args(args):
+    parser = argparse.ArgumentParser(add_help=True)
+    parser.add_argument("-c", "--core", help="Path to the core file or to text containing dump (by default stdin will be used)")
+    parser.add_argument("-s", "--symbol", "--sym", help="Path to the symbol file (by default in gdb currently loaded symbol file will be used)")
+    try:
+        args = parser.parse_args(args)
+    except SystemExit:
+        return {"help": True}
+    return args
+
+is_in_gdb = True
+try:
+    import gdb
+except ImportError:
+    is_in_gdb = False
+
+if is_in_gdb:
+    class PhoenixCoreCommand(gdb.Command):
+        """
+        Auto-load Phoenix core file and offset symbols on NOMMU targets.
+        Use `phoenix-load --help` for more information.
+        """
+
+        def __init__(self):
+            super(PhoenixCoreCommand, self).__init__("phoenix-core", gdb.COMMAND_USER)
+            self.target_arch = None
+
+        def invoke(self, arg, from_tty):
+            args = parse_args(gdb.string_to_argv(arg))
+            if "help" in args:
+                return
+            if not args.symbol:
+                try:
+                    if gdb.objfiles():
+                        args.symbol = gdb.objfiles()[0].filename
+                except gdb.error as e:
+                    print(f"Phoenix-core: Note: Could not determine current GDB symbol file:", e, file=sys.stderr)
+            commands = generate_gdb_commands(args)
+            for command in commands:
+                gdb.execute(command)
+        def complete(self, text, word):
+            options = ["-c", "--core", "-s", "--symbol", "--sym", "-h", "--help"]
+            if word and word.startswith("-"):
+                return [opt for opt in options if opt.startswith(word)]
+            return gdb.COMPLETE_FILENAME
+
+
+    PhoenixCoreCommand()
+
+elif __name__ == "__main__":
+    args = parse_args(sys.argv[1:])
+    if "help" in args:
+       exit(0)
+    commands = generate_gdb_commands(args)
+    if commands:
+        print("\n\n# Generated GDB commands:\n")
+        for cmd in commands:
+            print(cmd)
+        print()
+    else:
+        print("No commands generated.", file=sys.stderr)