diff options
| -rw-r--r-- | gdb/NEWS | 12 | ||||
| -rw-r--r-- | gdb/cli/cli-cmds.c | 6 | ||||
| -rw-r--r-- | gdb/disasm-flags.h | 1 | ||||
| -rw-r--r-- | gdb/disasm.c | 43 | ||||
| -rw-r--r-- | gdb/doc/gdb.texinfo | 48 | ||||
| -rw-r--r-- | gdb/mi/mi-cmd-disas.c | 6 | ||||
| -rw-r--r-- | gdb/record.c | 3 | ||||
| -rw-r--r-- | gdb/testsuite/gdb.mi/mi-disassemble.exp | 6 |
8 files changed, 109 insertions, 16 deletions
@@ -59,6 +59,18 @@ * gdb now supports zstd compressed debug sections (ELFCOMPRESS_ZSTD) for ELF. +* The format of 'disassemble /r' and 'record instruction-history /r' + has changed. The instruction bytes could now be grouped together, + and displayed in the endianness of the instruction. This is the + same layout as used by GNU objdump when disassembling. + + There is now 'disassemble /b' and 'record instruction-history /b' + which will always display the instructions bytes one at a time in + memory order, that is, the byte at the lowest address first. + + For both /r and /b GDB is now better at using whitespace in order to + align the disassembled instruction text. + * New commands maintenance set ignore-prologue-end-flag on|off diff --git a/gdb/cli/cli-cmds.c b/gdb/cli/cli-cmds.c index d570719..c78b93f 100644 --- a/gdb/cli/cli-cmds.c +++ b/gdb/cli/cli-cmds.c @@ -1508,6 +1508,9 @@ disassemble_current_function (gdb_disassembly_flags flags) A /r modifier will include raw instructions in hex with the assembly. + A /b modifier is similar to /r except the instruction bytes are printed + as separate bytes with no grouping, or endian switching. + A /s modifier will include source code with the assembly, like /m, with two important differences: 1) The output is still in pc address order. @@ -1546,6 +1549,9 @@ disassemble_command (const char *arg, int from_tty) case 'r': flags |= DISASSEMBLY_RAW_INSN; break; + case 'b': + flags |= DISASSEMBLY_RAW_BYTES; + break; case 's': flags |= DISASSEMBLY_SOURCE; break; diff --git a/gdb/disasm-flags.h b/gdb/disasm-flags.h index 025b689..5a7371b 100644 --- a/gdb/disasm-flags.h +++ b/gdb/disasm-flags.h @@ -33,6 +33,7 @@ enum gdb_disassembly_flag DISASSEMBLY_OMIT_PC = (0x1 << 4), DISASSEMBLY_SOURCE = (0x1 << 5), DISASSEMBLY_SPECULATIVE = (0x1 << 6), + DISASSEMBLY_RAW_BYTES = (0x1 << 7), }; DEF_ENUM_FLAGS_TYPE (enum gdb_disassembly_flag, gdb_disassembly_flags); diff --git a/gdb/disasm.c b/gdb/disasm.c index ba6ac2d..b5e503f 100644 --- a/gdb/disasm.c +++ b/gdb/disasm.c @@ -457,7 +457,7 @@ gdb_pretty_print_disassembler::pretty_print_insn (const struct disasm_insn *insn throw ex; } - if (flags & DISASSEMBLY_RAW_INSN) + if ((flags & (DISASSEMBLY_RAW_INSN | DISASSEMBLY_RAW_BYTES)) != 0) { /* Build the opcodes using a temporary stream so we can write them out in a single go for the MI. */ @@ -467,14 +467,51 @@ gdb_pretty_print_disassembler::pretty_print_insn (const struct disasm_insn *insn m_opcode_data.resize (size); read_code (pc, m_opcode_data.data (), size); - for (int i = 0; i < size; ++i) + /* The disassembler provides information about the best way to + display the instruction bytes to the user. We provide some sane + defaults in case the disassembler gets it wrong. */ + const struct disassemble_info *di = m_di.disasm_info (); + int bytes_per_line = std::max (di->bytes_per_line, size); + int bytes_per_chunk = std::max (di->bytes_per_chunk, 1); + + /* If the user has requested the instruction bytes be displayed + byte at a time, then handle that here. Also, if the instruction + is not a multiple of the chunk size (which probably indicates a + disassembler problem) then avoid that causing display problems + by switching to byte at a time mode. */ + if ((flags & DISASSEMBLY_RAW_BYTES) != 0 + || (size % bytes_per_chunk) != 0) + bytes_per_chunk = 1; + + /* Print the instruction opcodes bytes, grouped into chunks. */ + for (int i = 0; i < size; i += bytes_per_chunk) { if (i > 0) m_opcode_stb.puts (" "); - m_opcode_stb.printf ("%02x", (unsigned) m_opcode_data[i]); + + if (di->display_endian == BFD_ENDIAN_LITTLE) + { + for (int k = bytes_per_chunk; k-- != 0; ) + m_opcode_stb.printf ("%02x", (unsigned) m_opcode_data[i + k]); + } + else + { + for (int k = 0; k < bytes_per_chunk; k++) + m_opcode_stb.printf ("%02x", (unsigned) m_opcode_data[i + k]); + } + } + + /* Calculate required padding. */ + int nspaces = 0; + for (int i = size; i < bytes_per_line; i += bytes_per_chunk) + { + if (i > size) + nspaces++; + nspaces += bytes_per_chunk * 2; } m_uiout->field_stream ("opcodes", m_opcode_stb); + m_uiout->spaces (nspaces); m_uiout->text ("\t"); } diff --git a/gdb/doc/gdb.texinfo b/gdb/doc/gdb.texinfo index 238a49b..596e587 100644 --- a/gdb/doc/gdb.texinfo +++ b/gdb/doc/gdb.texinfo @@ -7945,7 +7945,10 @@ are printed in execution order. It can also print mixed source+disassembly if you specify the the @code{/m} or @code{/s} modifier, and print the raw instructions in hex -as well as in symbolic form by specifying the @code{/r} modifier. +as well as in symbolic form by specifying the @code{/r} or @code{/b} +modifier. The behaviour of the @code{/m}, @code{/s}, @code{/r}, and +@code{/b} modifiers are the same as for the @kbd{disassemble} command +(@pxref{disassemble,,@kbd{disassemble}}). The current position marker is printed for the instruction at the current program counter value. This instruction can appear multiple @@ -9859,6 +9862,7 @@ After @code{info line}, using @code{info line} again without specifying a location will display information about the next source line. +@anchor{disassemble} @table @code @kindex disassemble @cindex assembly instructions @@ -9869,16 +9873,17 @@ line. @itemx disassemble /m @itemx disassemble /s @itemx disassemble /r +@itemx disassemble /b This specialized command dumps a range of memory as machine instructions. It can also print mixed source+disassembly by specifying -the @code{/m} or @code{/s} modifier and print the raw instructions in hex -as well as in symbolic form by specifying the @code{/r} modifier. -The default memory range is the function surrounding the +the @code{/m} or @code{/s} modifier and print the raw instructions in +hex as well as in symbolic form by specifying the @code{/r} or @code{/b} +modifier. The default memory range is the function surrounding the program counter of the selected frame. A single argument to this command is a program counter value; @value{GDBN} dumps the function -surrounding this value. When two arguments are given, they should -be separated by a comma, possibly surrounded by whitespace. The -arguments specify a range of addresses to dump, in one of two forms: +surrounding this value. When two arguments are given, they should be +separated by a comma, possibly surrounded by whitespace. The arguments +specify a range of addresses to dump, in one of two forms: @table @code @item @var{start},@var{end} @@ -9916,6 +9921,35 @@ Dump of assembler code from 0x32c4 to 0x32e4: End of assembler dump. @end smallexample +The following two examples are for RISC-V, and demonstrates the +difference between the @code{/r} and @code{/b} modifiers. First with +@code{/b}, the bytes of the instruction are printed, in hex, in memory +order: + +@smallexample +(@value{GDBP}) disassemble /b 0x00010150,0x0001015c +Dump of assembler code from 0x10150 to 0x1015c: + 0x00010150 <call_me+4>: 22 dc sw s0,56(sp) + 0x00010152 <call_me+6>: 80 00 addi s0,sp,64 + 0x00010154 <call_me+8>: 23 26 a4 fe sw a0,-20(s0) + 0x00010158 <call_me+12>: 23 24 b4 fe sw a1,-24(s0) +End of assembler dump. +@end smallexample + +In contrast, with @code{/r} the bytes of the instruction are displayed +in the instruction order, for RISC-V this means that the bytes have been +swapped to little-endian order: + +@smallexample +(@value{GDBP}) disassemble /r 0x00010150,0x0001015c +Dump of assembler code from 0x10150 to 0x1015c: + 0x00010150 <call_me+4>: dc22 sw s0,56(sp) + 0x00010152 <call_me+6>: 0080 addi s0,sp,64 + 0x00010154 <call_me+8>: fea42623 sw a0,-20(s0) + 0x00010158 <call_me+12>: feb42423 sw a1,-24(s0) +End of assembler dump. +@end smallexample + Here is an example showing mixed source+assembly for Intel x86 with @code{/m} or @code{/s}, when the program is stopped just after function prologue in a non-optimized function with no inline code. diff --git a/gdb/mi/mi-cmd-disas.c b/gdb/mi/mi-cmd-disas.c index 387c490..c8e06cd 100644 --- a/gdb/mi/mi-cmd-disas.c +++ b/gdb/mi/mi-cmd-disas.c @@ -165,16 +165,16 @@ mi_cmd_disassemble (const char *command, char **argv, int argc) disasm_flags |= DISASSEMBLY_SOURCE_DEPRECATED; break; case 2: - disasm_flags |= DISASSEMBLY_RAW_INSN; + disasm_flags |= DISASSEMBLY_RAW_BYTES; break; case 3: - disasm_flags |= DISASSEMBLY_SOURCE_DEPRECATED | DISASSEMBLY_RAW_INSN; + disasm_flags |= DISASSEMBLY_SOURCE_DEPRECATED | DISASSEMBLY_RAW_BYTES; break; case 4: disasm_flags |= DISASSEMBLY_SOURCE; break; case 5: - disasm_flags |= DISASSEMBLY_SOURCE | DISASSEMBLY_RAW_INSN; + disasm_flags |= DISASSEMBLY_SOURCE | DISASSEMBLY_RAW_BYTES; break; default: gdb_assert_not_reached ("bad disassembly mode"); diff --git a/gdb/record.c b/gdb/record.c index 17a5df2..2390a58 100644 --- a/gdb/record.c +++ b/gdb/record.c @@ -494,6 +494,9 @@ get_insn_history_modifiers (const char **arg) case 'r': modifiers |= DISASSEMBLY_RAW_INSN; break; + case 'b': + modifiers |= DISASSEMBLY_RAW_BYTES; + break; case 'f': modifiers |= DISASSEMBLY_OMIT_FNAME; break; diff --git a/gdb/testsuite/gdb.mi/mi-disassemble.exp b/gdb/testsuite/gdb.mi/mi-disassemble.exp index b7c5247..ef3337d 100644 --- a/gdb/testsuite/gdb.mi/mi-disassemble.exp +++ b/gdb/testsuite/gdb.mi/mi-disassemble.exp @@ -245,12 +245,12 @@ proc test_disassembly_opcode_format {} { # then disassemble using the MI command. set longest_insn_bytes "" set longest_insn_addr "" - gdb_test_multiple "disassemble /r main" "" { - -re "^disassemble /r main\r\n" { + gdb_test_multiple "disassemble /b main" "" { + -re "^disassemble /b main\r\n" { exp_continue } - -re "^&\"disassemble /r main.n\"\r\n" { + -re "^&\"disassemble /b main.n\"\r\n" { exp_continue } |