1 .. SPDX-License-Identifier: GPL-2.0+
2 .. Copyright (c) 2016 Google, Inc
7 Firmware often consists of several components which must be packaged together.
8 For example, we may have SPL, U-Boot, a device tree and an environment area
9 grouped together and placed in MMC flash. When the system starts, it must be
10 able to find these pieces.
12 Building firmware should be separate from packaging it. Many of the complexities
13 of modern firmware build systems come from trying to do both at once. With
14 binman, you build all the pieces that are needed, using whatever assortment of
15 projects and build systems are needed, then use binman to stitch everything
22 Binman reads your board's device tree and finds a node which describes the
23 required image layout. It uses this to work out what to place where.
25 Binman provides a mechanism for building images, from simple SPL + U-Boot
26 combinations, to more complex arrangements with many parts. It also allows
27 users to inspect images, extract and replace binaries within them, repacking if
34 Apart from basic padding, alignment and positioning features, Binman supports
35 hierarchical images, compression, hashing and dealing with the binary blobs
36 which are a sad trend in open-source firmware at present.
38 Executable binaries can access the location of other binaries in an image by
39 using special linker symbols (zero-overhead but somewhat limited) or by reading
40 the devicetree description of the image.
42 Binman is designed primarily for use with U-Boot and associated binaries such
43 as ARM Trusted Firmware, but it is suitable for use with other projects, such
44 as Zephyr. Binman also provides facilities useful in Chromium OS, such as CBFS,
45 vblocks and and the like.
47 Binman provides a way to process binaries before they are included, by adding a
50 Binman is intended for use with U-Boot but is designed to be general enough
51 to be useful in other image-packaging situations.
57 As mentioned above, packaging of firmware is quite a different task from
58 building the various parts. In many cases the various binaries which go into
59 the image come from separate build systems. For example, ARM Trusted Firmware
60 is used on ARMv8 devices but is not built in the U-Boot tree. If a Linux kernel
61 is included in the firmware image, it is built elsewhere.
63 It is of course possible to add more and more build rules to the U-Boot
64 build system to cover these cases. It can shell out to other Makefiles and
65 build scripts. But it seems better to create a clear divide between building
66 software and packaging it.
68 At present this is handled by manual instructions, different for each board,
69 on how to create images that will boot. By turning these instructions into a
70 standard format, we can support making valid images for any board without
71 manual effort, lots of READMEs, etc.
75 - Each binary can have its own build system and tool chain without creating
76 any dependencies between them
77 - Avoids the need for a single-shot build: individual parts can be updated
78 and brought in as needed
79 - Provides for a standard image description available in the build and at
81 - SoC-specific image-signing tools can be accommodated
82 - Avoids cluttering the U-Boot build system with image-building code
83 - The image description is automatically available at run-time in U-Boot,
84 SPL. It can be made available to other software also
85 - The image description is easily readable (it's a text file in device-tree
86 format) and permits flexible packing of binaries
92 Binman uses the following terms:
94 - image - an output file containing a firmware image
95 - binary - an input binary that goes into the image
101 FIT is U-Boot's official image format. It supports multiple binaries with
102 load / execution addresses, compression. It also supports verification
103 through hashing and RSA signatures.
105 FIT was originally designed to support booting a Linux kernel (with an
106 optional ramdisk) and device tree chosen from various options in the FIT.
107 Now that U-Boot supports configuration via device tree, it is possible to
108 load U-Boot from a FIT, with the device tree chosen by SPL.
110 Binman considers FIT to be one of the binaries it can place in the image.
112 Where possible it is best to put as much as possible in the FIT, with binman
113 used to deal with cases not covered by FIT. Examples include initial
114 execution (since FIT itself does not have an executable header) and dealing
115 with device boundaries, such as the read-only/read-write separation in SPI
118 For U-Boot, binman should not be used to create ad-hoc images in place of
122 Relationship to mkimage
123 -----------------------
125 The mkimage tool provides a means to create a FIT. Traditionally it has
126 needed an image description file: a device tree, like binman, but in a
127 different format. More recently it has started to support a '-f auto' mode
128 which can generate that automatically.
130 More relevant to binman, mkimage also permits creation of many SoC-specific
131 image types. These can be listed by running 'mkimage -T list'. Examples
132 include 'rksd', the Rockchip SD/MMC boot format. The mkimage tool is often
133 called from the U-Boot build system for this reason.
135 Binman considers the output files created by mkimage to be binary blobs
136 which it can place in an image. Binman does not replace the mkimage tool or
137 this purpose. It would be possible in some situations to create a new entry
138 type for the images in mkimage, but this would not add functionality. It
139 seems better to use the mkimage tool to generate binaries and avoid blurring
140 the boundaries between building input files (mkimage) and packaging then
141 into a final image (binman).
147 Example use of binman in U-Boot
148 -------------------------------
150 Binman aims to replace some of the ad-hoc image creation in the U-Boot
153 Consider sunxi. It has the following steps:
155 #. It uses a custom mksunxiboot tool to build an SPL image called
156 sunxi-spl.bin. This should probably move into mkimage.
158 #. It uses mkimage to package U-Boot into a legacy image file (so that it can
159 hold the load and execution address) called u-boot.img.
161 #. It builds a final output image called u-boot-sunxi-with-spl.bin which
162 consists of sunxi-spl.bin, some padding and u-boot.img.
164 Binman is intended to replace the last step. The U-Boot build system builds
165 u-boot.bin and sunxi-spl.bin. Binman can then take over creation of
166 sunxi-spl.bin (by calling mksunxiboot, or hopefully one day mkimage). In any
167 case, it would then create the image from the component parts.
169 This simplifies the U-Boot Makefile somewhat, since various pieces of logic
170 can be replaced by a call to binman.
173 Example use of binman for x86
174 -----------------------------
176 In most cases x86 images have a lot of binary blobs, 'black-box' code
177 provided by Intel which must be run for the platform to work. Typically
178 these blobs are not relocatable and must be placed at fixed areas in the
181 Currently this is handled by ifdtool, which places microcode, FSP, MRC, VGA
182 BIOS, reference code and Intel ME binaries into a u-boot.rom file.
184 Binman is intended to replace all of this, with ifdtool left to handle only
185 the configuration of the Intel-format descriptor.
191 First install prerequisites, e.g::
193 sudo apt-get install python-pyelftools python3-pyelftools lzma-alone \
196 You can run binman directly if you put it on your PATH. But if you want to
197 install into your `~/.local` Python directory, use::
199 pip install tools/patman tools/dtoc tools/binman
201 Note that binman makes use of libraries from patman and dtoc, which is why these
202 need to be installed. Also you need `libfdt` and `pylibfdt` which can be
203 installed like this::
205 git clone git://git.kernel.org/pub/scm/utils/dtc/dtc.git
208 make NO_PYTHON=1 install
210 This installs the `libfdt.so` library into `~/lib` so you can use
211 `LD_LIBRARY_PATH=~/lib` when running binman. If you want to install it in the
212 system-library directory, replace the last line with::
214 make NO_PYTHON=1 PREFIX=/ install
221 binman build -b <board_name>
223 to build an image for a board. The board name is the same name used when
224 configuring U-Boot (e.g. for sandbox_defconfig the board name is 'sandbox').
225 Binman assumes that the input files for the build are in ../b/<board_name>.
227 Or you can specify this explicitly::
229 binman build -I <build_path>
231 where <build_path> is the build directory containing the output of the U-Boot
234 (Future work will make this more configurable)
236 In either case, binman picks up the device tree file (u-boot.dtb) and looks
237 for its instructions in the 'binman' node.
239 Binman has a few other options which you can see by running 'binman -h'.
242 Enabling binman for a board
243 ---------------------------
245 At present binman is invoked from a rule in the main Makefile. You should be
246 able to enable CONFIG_BINMAN to enable this rule.
248 The output file is typically named image.bin and is located in the output
249 directory. If input files are needed to you add these to INPUTS-y either in the
250 main Makefile or in a config.mk file in your arch subdirectory.
252 Once binman is executed it will pick up its instructions from a device-tree
253 file, typically <soc>-u-boot.dtsi, where <soc> is your CONFIG_SYS_SOC value.
254 You can use other, more specific CONFIG options - see 'Automatic .dtsi
258 Access to binman entry offsets at run time (symbols)
259 ----------------------------------------------------
261 Binman assembles images and determines where each entry is placed in the image.
262 This information may be useful to U-Boot at run time. For example, in SPL it
263 is useful to be able to find the location of U-Boot so that it can be executed
264 when SPL is finished.
266 Binman allows you to declare symbols in the SPL image which are filled in
267 with their correct values during the build. For example::
269 binman_sym_declare(ulong, u_boot_any, image_pos);
271 declares a ulong value which will be assigned to the image-pos of any U-Boot
272 image (u-boot.bin, u-boot.img, u-boot-nodtb.bin) that is present in the image.
273 You can access this value with something like::
275 ulong u_boot_offset = binman_sym(ulong, u_boot_any, image_pos);
277 Thus u_boot_offset will be set to the image-pos of U-Boot in memory, assuming
278 that the whole image has been loaded, or is available in flash. You can then
279 jump to that address to start U-Boot.
281 At present this feature is only supported in SPL and TPL. In principle it is
282 possible to fill in such symbols in U-Boot proper, as well, but a future C
283 library is planned for this instead, to read from the device tree.
285 As well as image-pos, it is possible to read the size of an entry and its
286 offset (which is the start position of the entry within its parent).
288 A small technical note: Binman automatically adds the base address of the image
289 (i.e. __image_copy_start) to the value of the image-pos symbol, so that when the
290 image is loaded to its linked address, the value will be correct and actually
291 point into the image.
293 For example, say SPL is at the start of the image and linked to start at address
294 80108000. If U-Boot's image-pos is 0x8000 then binman will write an image-pos
295 for U-Boot of 80110000 into the SPL binary, since it assumes the image is loaded
296 to 80108000, with SPL at 80108000 and U-Boot at 80110000.
298 For x86 devices (with the end-at-4gb property) this base address is not added
299 since it is assumed that images are XIP and the offsets already include the
303 Access to binman entry offsets at run time (fdt)
304 ------------------------------------------------
306 Binman can update the U-Boot FDT to include the final position and size of
307 each entry in the images it processes. The option to enable this is -u and it
308 causes binman to make sure that the 'offset', 'image-pos' and 'size' properties
309 are set correctly for every entry. Since it is not necessary to specify these in
310 the image definition, binman calculates the final values and writes these to
311 the device tree. These can be used by U-Boot at run-time to find the location
314 Alternatively, an FDT map entry can be used to add a special FDT containing
315 just the information about the image. This is preceded by a magic string so can
316 be located anywhere in the image. An image header (typically at the start or end
317 of the image) can be used to point to the FDT map. See fdtmap and image-header
318 entries for more information.
324 The -m option causes binman to output a .map file for each image that it
325 generates. This shows the offset and size of each entry. For example::
328 00000000 00000028 main-section
329 00000000 00000010 section@0
330 00000000 00000004 u-boot
331 00000010 00000010 section@1
332 00000000 00000004 u-boot
334 This shows a hierarchical image with two sections, each with a single entry. The
335 offsets of the sections are absolute hex byte offsets within the image. The
336 offsets of the entries are relative to their respective sections. The size of
337 each entry is also shown, in bytes (hex). The indentation shows the entries
338 nested inside their sections.
341 Passing command-line arguments to entries
342 -----------------------------------------
344 Sometimes it is useful to pass binman the value of an entry property from the
345 command line. For example some entries need access to files and it is not
346 always convenient to put these filenames in the image definition (device tree).
348 The -a option supports this::
354 <prop> is the property to set
355 <value> is the value to set it to
357 Not all properties can be provided this way. Only some entries support it,
358 typically for filenames.
361 Image description format
362 ========================
364 The binman node is called 'binman'. An example image description is shown
368 filename = "u-boot-sunxi-with-spl.bin";
371 filename = "spl/sunxi-spl.bin";
374 offset = <CONFIG_SPL_PAD_TO>;
379 This requests binman to create an image file called u-boot-sunxi-with-spl.bin
380 consisting of a specially formatted SPL (spl/sunxi-spl.bin, built by the
381 normal U-Boot Makefile), some 0xff padding, and a U-Boot legacy image. The
382 padding comes from the fact that the second binary is placed at
383 CONFIG_SPL_PAD_TO. If that line were omitted then the U-Boot binary would
384 immediately follow the SPL binary.
386 The binman node describes an image. The sub-nodes describe entries in the
387 image. Each entry represents a region within the overall image. The name of
388 the entry (blob, u-boot) tells binman what to put there. For 'blob' we must
389 provide a filename. For 'u-boot', binman knows that this means 'u-boot.bin'.
391 Entries are normally placed into the image sequentially, one after the other.
392 The image size is the total size of all entries. As you can see, you can
393 specify the start offset of an entry using the 'offset' property.
395 Note that due to a device tree requirement, all entries must have a unique
396 name. If you want to put the same binary in the image multiple times, you can
397 use any unique name, with the 'type' property providing the type.
399 The attributes supported for entries are described below.
402 This sets the offset of an entry within the image or section containing
403 it. The first byte of the image is normally at offset 0. If 'offset' is
404 not provided, binman sets it to the end of the previous region, or the
405 start of the image's entry area (normally 0) if there is no previous
409 This sets the alignment of the entry. The entry offset is adjusted
410 so that the entry starts on an aligned boundary within the containing
411 section or image. For example 'align = <16>' means that the entry will
412 start on a 16-byte boundary. This may mean that padding is added before
413 the entry. The padding is part of the containing section but is not
414 included in the entry, meaning that an empty space may be created before
415 the entry starts. Alignment should be a power of 2. If 'align' is not
416 provided, no alignment is performed.
419 This sets the size of the entry. The contents will be padded out to
420 this size. If this is not provided, it will be set to the size of the
424 Padding before the contents of the entry. Normally this is 0, meaning
425 that the contents start at the beginning of the entry. This can be used
426 to offset the entry contents a little. While this does not affect the
427 contents of the entry within binman itself (the padding is performed
428 only when its parent section is assembled), the end result will be that
429 the entry starts with the padding bytes, so may grow. Defaults to 0.
432 Padding after the contents of the entry. Normally this is 0, meaning
433 that the entry ends at the last byte of content (unless adjusted by
434 other properties). This allows room to be created in the image for
435 this entry to expand later. While this does not affect the contents of
436 the entry within binman itself (the padding is performed only when its
437 parent section is assembled), the end result will be that the entry ends
438 with the padding bytes, so may grow. Defaults to 0.
441 This sets the alignment of the entry size. For example, to ensure
442 that the size of an entry is a multiple of 64 bytes, set this to 64.
443 While this does not affect the contents of the entry within binman
444 itself (the padding is performed only when its parent section is
445 assembled), the end result is that the entry ends with the padding
446 bytes, so may grow. If 'align-size' is not provided, no alignment is
450 This sets the alignment of the end of an entry with respect to the
451 containing section. Some entries require that they end on an alignment
452 boundary, regardless of where they start. This does not move the start
453 of the entry, so the contents of the entry will still start at the
454 beginning. But there may be padding at the end. While this does not
455 affect the contents of the entry within binman itself (the padding is
456 performed only when its parent section is assembled), the end result
457 is that the entry ends with the padding bytes, so may grow.
458 If 'align-end' is not provided, no alignment is performed.
461 For 'blob' types this provides the filename containing the binary to
462 put into the entry. If binman knows about the entry type (like
463 u-boot-bin), then there is no need to specify this.
466 Sets the type of an entry. This defaults to the entry name, but it is
467 possible to use any name, and then add (for example) 'type = "u-boot"'
471 Indicates that the offset of this entry should not be set by placing
472 it immediately after the entry before. Instead, is set by another
473 entry which knows where this entry should go. When this boolean
474 property is present, binman will give an error if another entry does
475 not set the offset (with the GetOffsets() method).
478 This cannot be set on entry (or at least it is ignored if it is), but
479 with the -u option, binman will set it to the absolute image position
480 for each entry. This makes it easy to find out exactly where the entry
481 ended up in the image, regardless of parent sections, etc.
484 Expand the size of this entry to fit available space. This space is only
485 limited by the size of the image/section and the position of the next
489 Sets the compression algortihm to use (for blobs only). See the entry
490 documentation for details.
493 Sets the tag of the message to show if this entry is missing. This is
494 used for external blobs. When they are missing it is helpful to show
495 information about what needs to be fixed. See missing-blob-help for the
496 message for each tag.
499 By default binman substitutes entries with expanded versions if available,
500 so that a `u-boot` entry type turns into `u-boot-expanded`, for example. The
501 `--no-expanded` command-line option disables this globally. The
502 `no-expanded` property disables this just for a single entry. Put the
503 `no-expanded` boolean property in the node to select this behaviour.
505 The attributes supported for images and sections are described below. Several
506 are similar to those for entries.
509 Sets the image size in bytes, for example 'size = <0x100000>' for a
513 This is similar to 'offset' in entries, setting the offset of a section
514 within the image or section containing it. The first byte of the section
515 is normally at offset 0. If 'offset' is not provided, binman sets it to
516 the end of the previous region, or the start of the image's entry area
517 (normally 0) if there is no previous region.
520 This sets the alignment of the image size. For example, to ensure
521 that the image ends on a 512-byte boundary, use 'align-size = <512>'.
522 If 'align-size' is not provided, no alignment is performed.
525 This sets the padding before the image entries. The first entry will
526 be positioned after the padding. This defaults to 0.
529 This sets the padding after the image entries. The padding will be
530 placed after the last entry. This defaults to 0.
533 This specifies the pad byte to use when padding in the image. It
534 defaults to 0. To use 0xff, you would add 'pad-byte = <0xff>'.
537 This specifies the image filename. It defaults to 'image.bin'.
540 This causes binman to reorder the entries as needed to make sure they
541 are in increasing positional order. This can be used when your entry
542 order may not match the positional order. A common situation is where
543 the 'offset' properties are set by CONFIG options, so their ordering is
546 This is a boolean property so needs no value. To enable it, add a
547 line 'sort-by-offset;' to your description.
550 Normally only a single image is generated. To create more than one
551 image, put this property in the binman node. For example, this will
552 create image1.bin containing u-boot.bin, and image2.bin containing
553 both spl/u-boot-spl.bin and u-boot.bin::
571 For x86 machines the ROM offsets start just before 4GB and extend
572 up so that the image finished at the 4GB boundary. This boolean
573 option can be enabled to support this. The image size must be
574 provided so that binman knows when the image should start. For an
575 8MB ROM, the offset of the first entry would be 0xfff80000 with
576 this option, instead of 0 without this option.
579 This property specifies the entry offset of the first entry.
581 For PowerPC mpc85xx based CPU, CONFIG_SYS_TEXT_BASE is the entry
582 offset of the first entry. It can be 0xeff40000 or 0xfff40000 for
583 nor flash boot, 0x201000 for sd boot etc.
585 'end-at-4gb' property is not applicable where CONFIG_SYS_TEXT_BASE +
589 Specifies the default alignment for entries in this section, if they do
590 not specify an alignment. Note that this only applies to top-level entries
591 in the section (direct subentries), not any subentries of those entries.
592 This means that each section must specify its own default alignment, if
595 Examples of the above options can be found in the tests. See the
596 tools/binman/test directory.
598 It is possible to have the same binary appear multiple times in the image,
599 either by using a unit number suffix (u-boot@0, u-boot@1) or by using a
600 different name for each and specifying the type with the 'type' attribute.
603 Sections and hierachical images
604 -------------------------------
606 Sometimes it is convenient to split an image into several pieces, each of which
607 contains its own set of binaries. An example is a flash device where part of
608 the image is read-only and part is read-write. We can set up sections for each
609 of these, and place binaries in them independently. The image is still produced
610 as a single output file.
612 This feature provides a way of creating hierarchical images. For example here
613 is an example image with two copies of U-Boot. One is read-only (ro), intended
614 to be written only in the factory. Another is read-write (rw), so that it can be
615 upgraded in the field. The sizes are fixed so that the ro/rw boundary is known
616 and can be programmed::
634 This image could be placed into a SPI flash chip, with the protection boundary
637 A few special properties are provided for sections:
640 Indicates that this section is read-only. This has no impact on binman's
641 operation, but his property can be read at run time.
644 This string is prepended to all the names of the binaries in the
645 section. In the example above, the 'u-boot' binaries which actually be
646 renamed to 'ro-u-boot' and 'rw-u-boot'. This can be useful to
647 distinguish binaries with otherwise identical names.
653 Image nodes act like sections but also have a few extra properties:
656 Output filename for the image. This defaults to image.bin (or in the
657 case of multiple images <nodename>.bin where <nodename> is the name of
661 Create an image that can be repacked. With this option it is possible
662 to change anything in the image after it is created, including updating
663 the position and size of image components. By default this is not
664 permitted since it is not possibly to know whether this might violate a
665 constraint in the image description. For example, if a section has to
666 increase in size to hold a larger binary, that might cause the section
667 to fall out of its allow region (e.g. read-only portion of flash).
669 Adding this property causes the original offset and size values in the
670 image description to be stored in the FDT and fdtmap.
676 It is possible to ask binman to hash the contents of an entry and write that
677 value back to the device-tree node. For example::
687 Here, a new 'value' property will be written to the 'hash' node containing
688 the hash of the 'u-boot' entry. Only SHA256 is supported at present. Whole
689 sections can be hased if desired, by adding the 'hash' node to the section.
691 The has value can be chcked at runtime by hashing the data actually read and
692 comparing this has to the value in the device tree.
698 Binman automatically replaces 'u-boot' with an expanded version of that, i.e.
699 'u-boot-expanded'. This means that when you write::
707 type = "u-boot-expanded';
710 which in turn expands to::
722 U-Boot's various phase binaries actually comprise two or three pieces.
723 For example, u-boot.bin has the executable followed by a devicetree.
725 With binman we want to be able to update that devicetree with full image
726 information so that it is accessible to the executable. This is tricky
727 if it is not clear where the devicetree starts.
729 The above feature ensures that the devicetree is clearly separated from the
730 U-Boot executable and can be updated separately by binman as needed. It can be
731 disabled with the --no-expanded flag if required.
733 The same applies for u-boot-spl and u-boot-spl. In those cases, the expansion
734 includes the BSS padding, so for example::
743 type = "u-boot-expanded';
746 which in turn expands to::
761 Of course we should not expand SPL if it has no devicetree. Also if the BSS
762 padding is not needed (because BSS is in RAM as with CONFIG_SPL_SEPARATE_BSS),
763 the 'u-boot-spl-bss-pad' subnode should not be created. The use of the expaned
764 entry type is controlled by the UseExpanded() method. In the SPL case it checks
765 the 'spl-dtb' entry arg, which is 'y' or '1' if SPL has a devicetree.
767 For the BSS case, a 'spl-bss-pad' entry arg controls whether it is present. All
768 entry args are provided by the U-Boot Makefile.
774 Binman support compression for 'blob' entries (those of type 'blob' and
775 derivatives). To enable this for an entry, add a 'compress' property::
778 filename = "datafile";
782 The entry will then contain the compressed data, using the 'lz4' compression
783 algorithm. Currently this is the only one that is supported. The uncompressed
784 size is written to the node in an 'uncomp-size' property, if -u is used.
786 Compression is also supported for sections. In that case the entire section is
787 compressed in one block, including all its contents. This means that accessing
788 an entry from the section required decompressing the entire section. Also, the
789 size of a section indicates the space that it consumes in its parent section
790 (and typically the image). With compression, the section may contain more data,
791 and the uncomp-size property indicates that, as above. The contents of the
792 section is compressed first, before any padding is added. This ensures that the
793 padding itself is not compressed, which would be a waste of time.
796 Automatic .dtsi inclusion
797 -------------------------
799 It is sometimes inconvenient to add a 'binman' node to the .dts file for each
800 board. This can be done by using #include to bring in a common file. Another
801 approach supported by the U-Boot build system is to automatically include
802 a common header. You can then put the binman node (and anything else that is
803 specific to U-Boot, such as u-boot,dm-pre-reloc properies) in that header
806 Binman will search for the following files in arch/<arch>/dts::
808 <dts>-u-boot.dtsi where <dts> is the base name of the .dts file
809 <CONFIG_SYS_SOC>-u-boot.dtsi
810 <CONFIG_SYS_CPU>-u-boot.dtsi
811 <CONFIG_SYS_VENDOR>-u-boot.dtsi
814 U-Boot will only use the first one that it finds. If you need to include a
815 more general file you can do that from the more specific file using #include.
816 If you are having trouble figuring out what is going on, you can use
817 `DEVICE_TREE_DEBUG=1` with your build::
819 make DEVICE_TREE_DEBUG=1
820 scripts/Makefile.lib:334: Automatic .dtsi inclusion: options:
821 arch/arm/dts/juno-r2-u-boot.dtsi arch/arm/dts/-u-boot.dtsi
822 arch/arm/dts/armv8-u-boot.dtsi arch/arm/dts/armltd-u-boot.dtsi
823 arch/arm/dts/u-boot.dtsi ... found: "arch/arm/dts/juno-r2-u-boot.dtsi"
829 For the EFI app, where U-Boot is loaded from UEFI and runs as an app, there is
830 no way to update the devicetree after U-Boot is built. Normally this works by
831 creating a new u-boot.dtb.out with he updated devicetree, which is automatically
832 built into the output image. With ELF this is not possible since the ELF is
833 not part of an image, just a stand-along file. We must create an updated ELF
834 file with the new devicetree.
836 This is handled by the --update-fdt-in-elf option. It takes four arguments,
839 infile - filename of input ELF file, e.g. 'u-boot's
840 outfile - filename of output ELF file, e.g. 'u-boot.out'
841 begin_sym - symbol at the start of the embedded devicetree, e.g.
843 end_sym - symbol at the start of the embedded devicetree, e.g.
846 When this flag is used, U-Boot does all the normal packaging, but as an
847 additional step, it creates a new ELF file with the new devicetree embedded in
850 If logging is enabled you will see a message like this::
852 Updating file 'u-boot' with data length 0x400a (16394) between symbols
853 '__dtb_dt_begin' and '__dtb_dt_end'
855 There must be enough space for the updated devicetree. If not, an error like
856 the following is produced::
858 ValueError: Not enough space in 'u-boot' for data length 0x400a (16394);
859 size is 0x1744 (5956)
865 For details on the various entry types supported by binman and how to use them,
866 see entries.rst which is generated from the source code using:
868 binman entry-docs >tools/binman/entries.rst
882 It is possible to list the entries in an existing firmware image created by
883 binman, provided that there is an 'fdtmap' entry in the image. For example::
885 $ binman ls -i image.bin
886 Name Image-pos Size Entry-type Offset Uncomp-size
887 ----------------------------------------------------------------------
888 main-section c00 section 0
890 section 5fc section 4
892 u-boot 138 4 u-boot 38
893 u-boot-dtb 180 108 u-boot-dtb 80 3b5
894 u-boot-dtb 500 1ff u-boot-dtb 400 3b5
895 fdtmap 6fc 381 fdtmap 6fc
896 image-header bf8 8 image-header bf8
898 This shows the hierarchy of the image, the position, size and type of each
899 entry, the offset of each entry within its parent and the uncompressed size if
900 the entry is compressed.
902 It is also possible to list just some files in an image, e.g.::
904 $ binman ls -i image.bin section/cbfs
905 Name Image-pos Size Entry-type Offset Uncomp-size
906 --------------------------------------------------------------------
908 u-boot 138 4 u-boot 38
909 u-boot-dtb 180 108 u-boot-dtb 80 3b5
913 $ binman ls -i image.bin "*cb*" "*head*"
914 Name Image-pos Size Entry-type Offset Uncomp-size
915 ----------------------------------------------------------------------
917 u-boot 138 4 u-boot 38
918 u-boot-dtb 180 108 u-boot-dtb 80 3b5
919 image-header bf8 8 image-header bf8
921 If an older version of binman is used to list images created by a newer one, it
922 is possible that it will contain entry types that are not supported. These still
923 show with the correct type, but binman just sees them as blobs (plain binary
924 data). Any special features of that etype are not supported by the old binman.
927 Extracting files from images
928 ----------------------------
930 You can extract files from an existing firmware image created by binman,
931 provided that there is an 'fdtmap' entry in the image. For example::
933 $ binman extract -i image.bin section/cbfs/u-boot
935 which will write the uncompressed contents of that entry to the file 'u-boot' in
936 the current directory. You can also extract to a particular file, in this case
939 $ binman extract -i image.bin section/cbfs/u-boot -f u-boot.bin
941 It is possible to extract all files into a destination directory, which will
942 put files in subdirectories matching the entry hierarchy::
944 $ binman extract -i image.bin -O outdir
946 or just a selection::
948 $ binman extract -i image.bin "*u-boot*" -O outdir
950 Some entry types have alternative formats, for example fdtmap which allows
951 extracted just the devicetree binary without the fdtmap header::
953 $ binman extract -i /tmp/b/odroid-c4/image.bin -f out.dtb -F fdt fdtmap
957 // totalsize: 0x8ab (2219)
958 // off_dt_struct: 0x38
959 // off_dt_strings: 0x82c
960 // off_mem_rsvmap: 0x28
962 // last_comp_version: 2
963 // boot_cpuid_phys: 0x0
964 // size_dt_strings: 0x7f
965 // size_dt_struct: 0x7f4
968 image-node = "binman";
969 image-pos = <0x00000000>;
973 Use `-F list` to see what alternative formats are available::
975 $ binman extract -i /tmp/b/odroid-c4/image.bin -F list
976 Flag (-F) Entry type Description
977 fdt fdtmap Extract the devicetree blob from the fdtmap
980 Replacing files in an image
981 ---------------------------
983 You can replace files in an existing firmware image created by binman, provided
984 that there is an 'fdtmap' entry in the image. For example::
986 $ binman replace -i image.bin section/cbfs/u-boot
988 which will write the contents of the file 'u-boot' from the current directory
989 to the that entry, compressing if necessary. If the entry size changes, you must
990 add the 'allow-repack' property to the original image before generating it (see
991 above), otherwise you will get an error.
993 You can also use a particular file, in this case u-boot.bin::
995 $ binman replace -i image.bin section/cbfs/u-boot -f u-boot.bin
997 It is possible to replace all files from a source directory which uses the same
998 hierarchy as the entries::
1000 $ binman replace -i image.bin -I indir
1002 Files that are missing will generate a warning.
1004 You can also replace just a selection of entries::
1006 $ binman replace -i image.bin "*u-boot*" -I indir
1012 Binman normally operates silently unless there is an error, in which case it
1013 just displays the error. The -D/--debug option can be used to create a full
1014 backtrace when errors occur. You can use BINMAN_DEBUG=1 when building to select
1017 Internally binman logs some output while it is running. This can be displayed
1018 by increasing the -v/--verbosity from the default of 1:
1022 2: notices (important messages)
1023 3: info about major operations
1024 4: detailed information about each operation
1025 5: debug (all output)
1027 You can use BINMAN_VERBOSE=5 (for example) when building to select this.
1033 Order of image creation
1034 -----------------------
1036 Image creation proceeds in the following order, for each entry in the image.
1038 1. AddMissingProperties() - binman can add calculated values to the device
1039 tree as part of its processing, for example the offset and size of each
1040 entry. This method adds any properties associated with this, expanding the
1041 device tree as needed. These properties can have placeholder values which are
1042 set later by SetCalculatedProperties(). By that stage the size of sections
1043 cannot be changed (since it would cause the images to need to be repacked),
1044 but the correct values can be inserted.
1046 2. ProcessFdt() - process the device tree information as required by the
1047 particular entry. This may involve adding or deleting properties. If the
1048 processing is complete, this method should return True. If the processing
1049 cannot complete because it needs the ProcessFdt() method of another entry to
1050 run first, this method should return False, in which case it will be called
1053 3. GetEntryContents() - the contents of each entry are obtained, normally by
1054 reading from a file. This calls the Entry.ObtainContents() to read the
1055 contents. The default version of Entry.ObtainContents() calls
1056 Entry.GetDefaultFilename() and then reads that file. So a common mechanism
1057 to select a file to read is to override that function in the subclass. The
1058 functions must return True when they have read the contents. Binman will
1059 retry calling the functions a few times if False is returned, allowing
1060 dependencies between the contents of different entries.
1062 4. GetEntryOffsets() - calls Entry.GetOffsets() for each entry. This can
1063 return a dict containing entries that need updating. The key should be the
1064 entry name and the value is a tuple (offset, size). This allows an entry to
1065 provide the offset and size for other entries. The default implementation
1066 of GetEntryOffsets() returns {}.
1068 5. PackEntries() - calls Entry.Pack() which figures out the offset and
1069 size of an entry. The 'current' image offset is passed in, and the function
1070 returns the offset immediately after the entry being packed. The default
1071 implementation of Pack() is usually sufficient.
1073 Note: for sections, this also checks that the entries do not overlap, nor extend
1074 outside the section. If the section does not have a defined size, the size is
1075 set large enough to hold all the entries.
1077 6. SetImagePos() - sets the image position of every entry. This is the absolute
1078 position 'image-pos', as opposed to 'offset' which is relative to the containing
1079 section. This must be done after all offsets are known, which is why it is quite
1080 late in the ordering.
1082 7. SetCalculatedProperties() - update any calculated properties in the device
1083 tree. This sets the correct 'offset' and 'size' vaues, for example.
1085 8. ProcessEntryContents() - this calls Entry.ProcessContents() on each entry.
1086 The default implementatoin does nothing. This can be overriden to adjust the
1087 contents of an entry in some way. For example, it would be possible to create
1088 an entry containing a hash of the contents of some other entries. At this
1089 stage the offset and size of entries should not be adjusted unless absolutely
1090 necessary, since it requires a repack (going back to PackEntries()).
1092 9. ResetForPack() - if the ProcessEntryContents() step failed, in that an entry
1093 has changed its size, then there is no alternative but to go back to step 5 and
1094 try again, repacking the entries with the updated size. ResetForPack() removes
1095 the fixed offset/size values added by binman, so that the packing can start from
1098 10. WriteSymbols() - write the value of symbols into the U-Boot SPL binary.
1099 See 'Access to binman entry offsets at run time' below for a description of
1100 what happens in this stage.
1102 11. BuildImage() - builds the image and writes it to a file
1104 12. WriteMap() - writes a text file containing a map of the image. This is the
1111 Binman can make use of external command-line tools to handle processing of
1112 entry contents or to generate entry contents. These tools are executed using
1113 the 'tools' module's Run() method. The tools generally must exist on the PATH,
1114 but the --toolpath option can be used to specify additional search paths to
1115 use. This option can be specified multiple times to add more than one path.
1117 For some compile tools binman will use the versions specified by commonly-used
1118 environment variables like CC and HOSTCC for the C compiler, based on whether
1119 the tool's output will be used for the target or for the host machine. If those
1120 aren't given, it will also try to derive target-specific versions from the
1121 CROSS_COMPILE environment variable during a cross-compilation.
1123 If the tool is not available in the path you can use BINMAN_TOOLPATHS to specify
1124 a space-separated list of paths to search, e.g.::
1126 BINMAN_TOOLPATHS="/tools/g12a /tools/tegra" binman ...
1132 Binary blobs, even if the source code is available, complicate building
1133 firmware. The instructions can involve multiple steps and the binaries may be
1134 hard to build or obtain. Binman at least provides a unified description of how
1135 to build the final image, no matter what steps are needed to get there.
1137 Binman also provides a `blob-ext` entry type that pulls in a binary blob from an
1138 external file. If the file is missing, binman can optionally complete the build
1139 and just report a warning. Use the `-M/--allow-missing` option to enble this.
1140 This is useful in CI systems which want to check that everything is correct but
1141 don't have access to the blobs.
1143 If the blobs are in a different directory, you can specify this with the `-I`
1146 For U-Boot, you can use set the BINMAN_INDIRS environment variable to provide a
1147 space-separated list of directories to search for binary blobs::
1149 BINMAN_INDIRS="odroid-c4/fip/g12a \
1150 odroid-c4/build/board/hardkernel/odroidc4/firmware \
1151 odroid-c4/build/scp_task" binman ...
1156 Binman is a critical tool and is designed to be very testable. Entry
1157 implementations target 100% test coverage. Run 'binman test -T' to check this.
1159 To enable Python test coverage on Debian-type distributions (e.g. Ubuntu)::
1161 $ sudo apt-get install python-coverage python3-coverage python-pytest
1167 Binman tries to run tests concurrently. This means that the tests make use of
1168 all available CPUs to run.
1172 $ sudo apt-get install python-subunit python3-subunit
1174 Use '-P 1' to disable this. It is automatically disabled when code coverage is
1175 being used (-T) since they are incompatible.
1181 Sometimes when debugging tests it is useful to keep the input and output
1182 directories so they can be examined later. Use -X or --test-preserve-dirs for
1186 Running tests on non-x86 architectures
1187 --------------------------------------
1189 Binman's tests have been written under the assumption that they'll be run on a
1190 x86-like host and there hasn't been an attempt to make them portable yet.
1191 However, it's possible to run the tests by cross-compiling to x86.
1193 To install an x86 cross-compiler on Debian-type distributions (e.g. Ubuntu)::
1195 $ sudo apt-get install gcc-x86-64-linux-gnu
1197 Then, you can run the tests under cross-compilation::
1199 $ CROSS_COMPILE=x86_64-linux-gnu- binman test -T
1201 You can also use gcc-i686-linux-gnu similar to the above.
1204 Writing new entries and debugging
1205 ---------------------------------
1207 The behaviour of entries is defined by the Entry class. All other entries are
1208 a subclass of this. An important subclass is Entry_blob which takes binary
1209 data from a file and places it in the entry. In fact most entry types are
1210 subclasses of Entry_blob.
1212 Each entry type is a separate file in the tools/binman/etype directory. Each
1213 file contains a class called Entry_<type> where <type> is the entry type.
1214 New entry types can be supported by adding new files in that directory.
1215 These will automatically be detected by binman when needed.
1217 Entry properties are documented in entry.py. The entry subclasses are free
1218 to change the values of properties to support special behaviour. For example,
1219 when Entry_blob loads a file, it sets content_size to the size of the file.
1220 Entry classes can adjust other entries. For example, an entry that knows
1221 where other entries should be positioned can set up those entries' offsets
1222 so they don't need to be set in the binman decription. It can also adjust
1225 Most of the time such essoteric behaviour is not needed, but it can be
1226 essential for complex images.
1228 If you need to specify a particular device-tree compiler to use, you can define
1229 the DTC environment variable. This can be useful when the system dtc is too
1232 To enable a full backtrace and other debugging features in binman, pass
1233 BINMAN_DEBUG=1 to your build::
1235 make qemu-x86_defconfig
1238 To enable verbose logging from binman, base BINMAN_VERBOSE to your build, which
1239 adds a -v<level> option to the call to binman::
1241 make qemu-x86_defconfig
1242 make BINMAN_VERBOSE=5
1245 Building sections in parallel
1246 -----------------------------
1248 By default binman uses multiprocessing to speed up compilation of large images.
1249 This works at a section level, with one thread for each entry in the section.
1250 This can speed things up if the entries are large and use compression.
1252 This feature can be disabled with the '-T' flag, which defaults to a suitable
1253 value for your machine. This depends on the Python version, e.g on v3.8 it uses
1254 12 threads on an 8-core machine. See ConcurrentFutures_ for more details.
1256 The special value -T0 selects single-threaded mode, useful for debugging during
1257 development, since dealing with exceptions and problems in threads is more
1258 difficult. This avoids any use of ThreadPoolExecutor.
1264 Binman takes a lot of inspiration from a Chrome OS tool called
1265 'cros_bundle_firmware', which I wrote some years ago. That tool was based on
1266 a reasonably simple and sound design but has expanded greatly over the
1267 years. In particular its handling of x86 images is convoluted.
1269 Quite a few lessons have been learned which are hopefully applied here.
1275 On the face of it, a tool to create firmware images should be fairly simple:
1276 just find all the input binaries and place them at the right place in the
1277 image. The difficulty comes from the wide variety of input types (simple
1278 flat binaries containing code, packaged data with various headers), packing
1279 requirments (alignment, spacing, device boundaries) and other required
1280 features such as hierarchical images.
1282 The design challenge is to make it easy to create simple images, while
1283 allowing the more complex cases to be supported. For example, for most
1284 images we don't much care exactly where each binary ends up, so we should
1285 not have to specify that unnecessarily.
1287 New entry types should aim to provide simple usage where possible. If new
1288 core features are needed, they can be added in the Entry base class.
1296 - Use of-platdata to make the information available to code that is unable
1297 to use device tree (such as a very small SPL image). For now, limited info is
1298 available via linker symbols
1299 - Allow easy building of images by specifying just the board name
1300 - Support building an image for a board (-b) more completely, with a
1301 configurable build directory
1302 - Detect invalid properties in nodes
1303 - Sort the fdtmap by offset
1304 - Output temporary files to a different directory
1307 Simon Glass <sjg@chromium.org>
1310 .. _ConcurrentFutures: https://docs.python.org/3/library/concurrent.futures.html#concurrent.futures.ThreadPoolExecutor