1LLVM-OBJCOPY(1) LLVM LLVM-OBJCOPY(1)
2
6 llvm-objcopy - object copying and editing tool
7
9 llvm-objcopy [options] input [output]
10
12 llvm-objcopy is a tool to copy and manipulate objects. In basic usage,
13 it makes a semantic copy of the input to the output. If any options are
14 specified, the output may be modified along the way, e.g. by removing
15 sections.
16 If no output file is specified, the input file is modified in-place. If
18 "-" is specified for the input file, the input is read from the pro‐
19 gram's standard input stream. If "-" is specified for the output file,
20 the output is written to the standard output stream of the program.
21 If the input is an archive, any requested operations will be applied to
23 each archive member individually.
24 The tool is still in active development, but in most scenarios it works
26 as a drop-in replacement for GNU's objcopy.
27
29 The following options are either agnostic of the file format, or apply
30 to multiple file formats.
31 --add-gnu-debuglink <debug-file>
33 Add a .gnu_debuglink section for <debug-file> to the output.
34 --add-section <section=file>
36 Add a section named <section> with the contents of <file> to the
37 output. For ELF objects the section will be of type SHT_NOTE, if
38 the name starts with ".note". Otherwise, it will have type
39 SHT_PROGBITS. Can be specified multiple times to add multiple
40 sections.
41 For MachO objects, <section> must be formatted as <segment
43 name>,<section name>.
44 --binary-architecture <arch>, -B
46 Ignored for compatibility.
47 --disable-deterministic-archives, -U
49 Use real values for UIDs, GIDs and timestamps when updating ar‐
50 chive member headers.
51 --discard-all, -x
53 Remove most local symbols from the output. Different file for‐
54 mats may limit this to a subset of the local symbols. For exam‐
55 ple, file and section symbols in ELF objects will not be dis‐
56 carded. Additionally, remove all debug sections.
57 --dump-section <section>=<file>
59 Dump the contents of section <section> into the file <file>. Can
60 be specified multiple times to dump multiple sections to differ‐
61 ent files. <file> is unrelated to the input and output files
62 provided to llvm-objcopy and as such the normal copying and
63 editing operations will still be performed. No operations are
64 performed on the sections prior to dumping them.
65 For MachO objects, <section> must be formatted as <segment
67 name>,<section name>.
68 --enable-deterministic-archives, -D
70 Enable deterministic mode when copying archives, i.e. use 0 for
71 archive member header UIDs, GIDs and timestamp fields. On by de‐
72 fault.
73 --help, -h
75 Print a summary of command line options.
76 --only-keep-debug
78 Produce a debug file as the output that only preserves contents
79 of sections useful for debugging purposes.
80 For ELF objects, this removes the contents of SHF_ALLOC sections
82 that are not SHT_NOTE by making them SHT_NOBITS and shrinking
83 the program headers where possible.
84 --only-section <section>, -j
86 Remove all sections from the output, except for sections named
87 <section>. Can be specified multiple times to keep multiple
88 sections.
89 For MachO objects, <section> must be formatted as <segment
91 name>,<section name>.
92 --redefine-sym <old>=<new>
94 Rename symbols called <old> to <new> in the output. Can be spec‐
95 ified multiple times to rename multiple symbols.
96 --redefine-syms <filename>
98 Rename symbols in the output as described in the file <file‐
99 name>. In the file, each line represents a single symbol to re‐
100 name, with the old name and new name separated by whitespace.
101 Leading and trailing whitespace is ignored, as is anything fol‐
102 lowing a '#'. Can be specified multiple times to read names from
103 multiple files.
104 --regex
106 If specified, symbol and section names specified by other
107 switches are treated as extended POSIX regular expression pat‐
108 terns.
109 --remove-section <section>, -R
111 Remove the specified section from the output. Can be specified
112 multiple times to remove multiple sections simultaneously.
113 For MachO objects, <section> must be formatted as <segment
115 name>,<section name>.
116 --set-section-alignment <section>=<align>
118 Set the alignment of section <section> to <align>. Can be speci‐
119 fied multiple times to update multiple sections.
120 --set-section-flags <section>=<flag>[,<flag>,...]
122 Set section properties in the output of section <section> based
123 on the specified <flag> values. Can be specified multiple times
124 to update multiple sections.
125 Supported flag names are alloc, load, noload, readonly, exclude,
127 debug, code, data, rom, share, contents, merge and strings. Not
128 all flags are meaningful for all object file formats.
129 For ELF objects, the flags have the following effects:
131 • alloc = add the SHF_ALLOC flag.
133 • load = if the section has SHT_NOBITS type, mark it as a
135 SHT_PROGBITS section.
136 • readonly = if this flag is not specified, add the SHF_WRITE
138 flag.
139 • exclude = add the SHF_EXCLUDE flag.
141 • code = add the SHF_EXECINSTR flag.
143 • merge = add the SHF_MERGE flag.
145 • strings = add the SHF_STRINGS flag.
147 • contents = if the section has SHT_NOBITS type, mark it as a
149 SHT_PROGBITS section.
150 For COFF objects, the flags have the following effects:
152 • alloc = add the IMAGE_SCN_CNT_UNINITIALIZED_DATA and IM‐
154 AGE_SCN_MEM_READ flags, unless the load flag is specified.
155 • noload = add the IMAGE_SCN_LNK_REMOVE and IMAGE_SCN_MEM_READ
157 flags.
158 • readonly = if this flag is not specified, add the IM‐
160 AGE_SCN_MEM_WRITE flag.
161 • exclude = add the IMAGE_SCN_LNK_REMOVE and IMAGE_SCN_MEM_READ
163 flags.
164 • debug = add the IMAGE_SCN_CNT_INITIALIZED_DATA, IM‐
166 AGE_SCN_MEM_DISCARDABLE and IMAGE_SCN_MEM_READ flags.
167 • code = add the IMAGE_SCN_CNT_CODE, IMAGE_SCN_MEM_EXECUTE and
169 IMAGE_SCN_MEM_READ flags.
170 • data = add the IMAGE_SCN_CNT_INITIALIZED_DATA and IM‐
172 AGE_SCN_MEM_READ flags.
173 • share = add the IMAGE_SCN_MEM_SHARED and IMAGE_SCN_MEM_READ
175 flags.
176 --strip-all-gnu
178 Remove all symbols, debug sections and relocations from the out‐
179 put. This option is equivalent to GNU objcopy's --strip-all
180 switch.
181 --strip-all, -S
183 For ELF objects, remove from the output all symbols and non-al‐
184 loc sections not within segments, except for .gnu.warning,
185 .ARM.attribute sections and the section name table.
186 For COFF and Mach-O objects, remove all symbols, debug sections,
188 and relocations from the output.
189 --strip-debug, -g
191 Remove all debug sections from the output.
192 --strip-symbol <symbol>, -N
194 Remove all symbols named <symbol> from the output. Can be speci‐
195 fied multiple times to remove multiple symbols.
196 --strip-symbols <filename>
198 Remove all symbols whose names appear in the file <filename>,
199 from the output. In the file, each line represents a single sym‐
200 bol name, with leading and trailing whitespace ignored, as is
201 anything following a '#'. Can be specified multiple times to
202 read names from multiple files.
203 --strip-unneeded-symbol <symbol>
205 Remove from the output all symbols named <symbol> that are local
206 or undefined and are not required by any relocation.
207 --strip-unneeded-symbols <filename>
209 Remove all symbols whose names appear in the file <filename>,
210 from the output, if they are local or undefined and are not re‐
211 quired by any relocation. In the file, each line represents a
212 single symbol name, with leading and trailing whitespace ig‐
213 nored, as is anything following a '#'. Can be specified multiple
214 times to read names from multiple files.
215 --strip-unneeded
217 Remove from the output all local or undefined symbols that are
218 not required by relocations. Also remove all debug sections.
219 --update-section <name>=<file>
221 Replace the contents of the section <name> with contents from
222 the file <file>. If the section <name> is part of a segment, the
223 new contents cannot be larger than the existing section.
224 --version, -V
226 Display the version of the llvm-objcopy executable.
227 --wildcard, -w
229 Allow wildcard syntax for symbol-related flags. On by default
230 for section-related flags. Incompatible with --regex.
231 Wildcard syntax allows the following special symbols:
233 ┌───────────────┬─────────────────────┬────────────┐
235 │Character │ Meaning │ Equivalent │
236 ├───────────────┼─────────────────────┼────────────┤
237 │* │ Any number of char‐ │ .* │
238 │ │ acters │ │
239 ├───────────────┼─────────────────────┼────────────┤
240 │? │ Any single charac‐ │ . │
241 │ │ ter │ │
242 ├───────────────┼─────────────────────┼────────────┤
243 │\ │ Escape the next │ \ │
244 │ │ character │ │
245 ├───────────────┼─────────────────────┼────────────┤
246 │[a-z] │ Character class │ [a-z] │
247 ├───────────────┼─────────────────────┼────────────┤
248 │[!a-z], [^a-z] │ Negated character │ [^a-z] │
249 │ │ class │ │
250 └───────────────┴─────────────────────┴────────────┘
251 Additionally, starting a wildcard with '!' will prevent a match,
253 even if another flag matches. For example -w -N '*' -N '!x' will
254 strip all symbols except for x.
255 The order of wildcards does not matter. For example, -w -N '*'
257 -N '!x' is the same as -w -N '!x' -N '*'.
258 @<FILE>
260 Read command-line options and commands from response file
261 <FILE>.
262
264 The following options are implemented only for ELF objects. If used
265 with other objects, llvm-objcopy will either emit an error or silently
266 ignore them.
267 --add-symbol <name>=[<section>:]<value>[,<flags>]
269 Add a new symbol called <name> to the output symbol table, in
270 the section named <section>, with value <value>. If <section> is
271 not specified, the symbol is added as an absolute symbol. The
272 <flags> affect the symbol properties. Accepted values are:
273 • global = the symbol will have global binding.
275 • local = the symbol will have local binding.
277 • weak = the symbol will have weak binding.
279 • default = the symbol will have default visibility.
281 • hidden = the symbol will have hidden visibility.
283 • protected = the symbol will have protected visibility.
285 • file = the symbol will be an STT_FILE symbol.
287 • section = the symbol will be an STT_SECTION symbol.
289 • object = the symbol will be an STT_OBJECT symbol.
291 • function = the symbol will be an STT_FUNC symbol.
293 • indirect-function = the symbol will be an STT_GNU_IFUNC sym‐
295 bol.
296 Additionally, the following flags are accepted but ignored: de‐
298 bug, constructor, warning, indirect, synthetic, unique-object,
299 before.
300 Can be specified multiple times to add multiple symbols.
302 --allow-broken-links
304 Allow llvm-objcopy to remove sections even if it would leave in‐
305 valid section references. Any invalid sh_link fields will be set
306 to zero.
307 --change-start <incr>, --adjust-start
309 Add <incr> to the program's start address. Can be specified mul‐
310 tiple times, in which case the values will be applied cumula‐
311 tively.
312 --compress-debug-sections [<format>]
314 Compress DWARF debug sections in the output, using the specified
315 format. Supported formats are zlib. Use zlib if <format> is
316 omitted.
317 --decompress-debug-sections
319 Decompress any compressed DWARF debug sections in the output.
320 --discard-locals, -X
322 Remove local symbols starting with ".L" from the output.
323 --extract-dwo
325 Remove all sections that are not DWARF .dwo sections from the
326 output.
327 --extract-main-partition
329 Extract the main partition from the output.
330 --extract-partition <name>
332 Extract the named partition from the output.
333 --globalize-symbol <symbol>
335 Mark any defined symbols named <symbol> as global symbols in the
336 output. Can be specified multiple times to mark multiple sym‐
337 bols.
338 --globalize-symbols <filename>
340 Read a list of names from the file <filename> and mark defined
341 symbols with those names as global in the output. In the file,
342 each line represents a single symbol, with leading and trailing
343 whitespace ignored, as is anything following a '#'. Can be spec‐
344 ified multiple times to read names from multiple files.
345 --input-target <format>, -I
347 Read the input as the specified format. See SUPPORTED FORMATS
348 for a list of valid <format> values. If unspecified, llvm-obj‐
349 copy will attempt to determine the format automatically.
350 --keep-file-symbols
352 Keep symbols of type STT_FILE, even if they would otherwise be
353 stripped.
354 --keep-global-symbol <symbol>, -G
356 Make all symbols local in the output, except for symbols with
357 the name <symbol>. Can be specified multiple times to ignore
358 multiple symbols.
359 --keep-global-symbols <filename>
361 Make all symbols local in the output, except for symbols named
362 in the file <filename>. In the file, each line represents a sin‐
363 gle symbol, with leading and trailing whitespace ignored, as is
364 anything following a '#'. Can be specified multiple times to
365 read names from multiple files.
366 --keep-section <section>
368 When removing sections from the output, do not remove sections
369 named <section>. Can be specified multiple times to keep multi‐
370 ple sections.
371 --keep-symbol <symbol>, -K
373 When removing symbols from the output, do not remove symbols
374 named <symbol>. Can be specified multiple times to keep multiple
375 symbols.
376 --keep-symbols <filename>
378 When removing symbols from the output do not remove symbols
379 named in the file <filename>. In the file, each line represents
380 a single symbol, with leading and trailing whitespace ignored,
381 as is anything following a '#'. Can be specified multiple times
382 to read names from multiple files.
383 --localize-hidden
385 Make all symbols with hidden or internal visibility local in the
386 output.
387 --localize-symbol <symbol>, -L
389 Mark any defined non-common symbol named <symbol> as a local
390 symbol in the output. Can be specified multiple times to mark
391 multiple symbols as local.
392 --localize-symbols <filename>
394 Read a list of names from the file <filename> and mark defined
395 non-common symbols with those names as local in the output. In
396 the file, each line represents a single symbol, with leading and
397 trailing whitespace ignored, as is anything following a '#'. Can
398 be specified multiple times to read names from multiple files.
399 --new-symbol-visibility <visibility>
401 Specify the visibility of the symbols automatically created when
402 using binary input or --add-symbol. Valid options are:
403 • default
405 • hidden
407 • internal
409 • protected
411 The default is default.
413 --output-target <format>, -O
415 Write the output as the specified format. See SUPPORTED FORMATS
416 for a list of valid <format> values. If unspecified, the output
417 format is assumed to be the same as the value specified for
418 --input-target or the input file's format if that option is also
419 unspecified.
420 --prefix-alloc-sections <prefix>
422 Add <prefix> to the front of the names of all allocatable sec‐
423 tions in the output.
424 --prefix-symbols <prefix>
426 Add <prefix> to the front of every symbol name in the output.
427 --preserve-dates, -p
429 Preserve access and modification timestamps in the output.
430 --rename-section <old>=<new>[,<flag>,...]
432 Rename sections called <old> to <new> in the output, and apply
433 any specified <flag> values. See --set-section-flags for a list
434 of supported flags. Can be specified multiple times to rename
435 multiple sections.
436 --set-section-type <section>=<type>
438 Set the type of section <section> to the integer <type>. Can be
439 specified multiple times to update multiple sections.
440 --set-start-addr <addr>
442 Set the start address of the output to <addr>. Overrides any
443 previously specified --change-start or --adjust-start options.
444 --split-dwo <dwo-file>
446 Equivalent to running llvm-objcopy with --extract-dwo and
447 <dwo-file> as the output file and no other options, and then
448 with --strip-dwo on the input file.
449 --strip-dwo
451 Remove all DWARF .dwo sections from the output.
452 --strip-non-alloc
454 Remove from the output all non-allocatable sections that are not
455 within segments.
456 --strip-sections
458 Remove from the output all section headers and all section data
459 not within segments. Note that many tools will not be able to
460 use an object without section headers.
461 --target <format>, -F
463 Equivalent to --input-target and --output-target for the speci‐
464 fied format. See SUPPORTED FORMATS for a list of valid <format>
465 values.
466 --weaken-symbol <symbol>, -W
468 Mark any global symbol named <symbol> as a weak symbol in the
469 output. Can be specified multiple times to mark multiple symbols
470 as weak.
471 --weaken-symbols <filename>
473 Read a list of names from the file <filename> and mark global
474 symbols with those names as weak in the output. In the file,
475 each line represents a single symbol, with leading and trailing
476 whitespace ignored, as is anything following a '#'. Can be spec‐
477 ified multiple times to read names from multiple files.
478 --weaken
480 Mark all defined global symbols as weak in the output.
481
483 --keep-undefined
484 Keep undefined symbols, even if they would otherwise be
485 stripped.
486
488 --subsystem <name>[:<version>]
489 Set the PE subsystem, and optionally subsystem version.
490
492 The following values are currently supported by llvm-objcopy for the
493 --input-target, --output-target, and --target options. For GNU objcopy
494 compatibility, the values are all bfdnames.
495 • binary
497 • ihex
499 • elf32-i386
501 • elf32-x86-64
503 • elf64-x86-64
505 • elf32-iamcu
507 • elf32-littlearm
509 • elf64-aarch64
511 • elf64-littleaarch64
513 • elf32-littleriscv
515 • elf64-littleriscv
517 • elf32-powerpc
519 • elf32-powerpcle
521 • elf64-powerpc
523 • elf64-powerpcle
525 • elf32-bigmips
527 • elf32-ntradbigmips
529 • elf32-ntradlittlemips
531 • elf32-tradbigmips
533 • elf32-tradlittlemips
535 • elf64-tradbigmips
537 • elf64-tradlittlemips
539 • elf32-sparc
541 • elf32-sparcel
543 Additionally, all targets except binary and ihex can have -freebsd as a
545 suffix.
546
548 If binary is used as the value for --input-target, the input file will
549 be embedded as a data section in an ELF relocatable object, with sym‐
550 bols _binary_<file_name>_start, _binary_<file_name>_end, and _bi‐
551 nary_<file_name>_size representing the start, end and size of the data,
552 where <file_name> is the path of the input file as specified on the
553 command line with non-alphanumeric characters converted to _.
554 If binary is used as the value for --output-target, the output file
556 will be a raw binary file, containing the memory image of the input
557 file. Symbols and relocation information will be discarded. The image
558 will start at the address of the first loadable section in the output.
559
561 llvm-objcopy exits with a non-zero exit code if there is an error.
562 Otherwise, it exits with code 0.
563
565 To report bugs, please visit <‐
566 https://github.com/llvm/llvm-project/labels/tools:llvm-objcopy/strip/>.
567 There is a known issue with --input-target and --target causing only
569 binary and ihex formats to have any effect. Other values will be ig‐
570 nored and llvm-objcopy will attempt to guess the input format.
571
573 llvm-strip(1)
574
576 Maintained by the LLVM Team (https://llvm.org/).
577
579 2003-2023, LLVM Project
58015 2023-07-20 LLVM-OBJCOPY(1)