1 -------------------------------------------------
2 Building EFI Applications Using the GNU Toolchain
3 -------------------------------------------------
5 David Mosberger <davidm@hpl.hp.com>
10 Copyright (c) 1999-2007 Hewlett-Packard Co.
11 Copyright (c) 2006-2010 Intel Co.
13 Last update: 04/09/2007
17 This document has two parts: the first part describes how to develop
18 EFI applications for IA-64,x86 and x86_64 using the GNU toolchain and the EFI
19 development environment contained in this directory. The second part
20 describes some of the more subtle aspects of how this development
25 * Part 1: Developing EFI Applications
30 To develop x86 and x86_64 EFI applications, the following tools are needed:
32 - gcc-3.0 or newer (gcc 2.7.2 is NOT sufficient!)
33 As of gnu-efi-3.0b, the Redhat 8.0 toolchain is known to work,
34 but the Redhat 9.0 toolchain is not currently supported.
36 - A version of "objcopy" that supports EFI applications. To
37 check if your version includes EFI support, issue the
42 and verify that the line "supported targets" contains the
43 string "efi-app-ia32" and "efi-app-x86_64". The binutils release
44 binutils-2.17.50.0.14 supports Intel64 EFI.
46 - For debugging purposes, it's useful to have a version of
47 "objdump" that supports EFI applications as well. This
48 allows inspect and disassemble EFI binaries.
50 To develop IA-64 EFI applications, the following tools are needed:
52 - A version of gcc newer than July 30th 1999 (older versions
53 had problems with generating position independent code).
54 As of gnu-efi-3.0b, gcc-3.1 is known to work well.
56 - A version of "objcopy" that supports EFI applications. To
57 check if your version includes EFI support, issue the
62 and verify that the line "supported targets" contains the
63 string "efi-app-ia64".
65 - For debugging purposes, it's useful to have a version of
66 "objdump" that supports EFI applications as well. This
67 allows inspect and disassemble EFI binaries.
70 ** Directory Structure
72 This EFI development environment contains the following
75 inc: This directory contains the EFI-related include files. The
76 files are taken from Intel's EFI source distribution, except
77 that various fixes were applied to make it compile with the
80 lib: This directory contains the source code for Intel's EFI library.
81 Again, the files are taken from Intel's EFI source
82 distribution, with changes to make them compile with the GNU
85 gnuefi: This directory contains the glue necessary to convert ELF64
86 binaries to EFI binaries. Various runtime code bits, such as
87 a self-relocator are included as well. This code has been
88 contributed by the Hewlett-Packard Company and is distributed
91 apps: This directory contains a few simple EFI test apps.
95 It is necessary to edit the Makefile in the directory containing this
96 README file before EFI applications can be built. Specifically, you
97 should verify that macros CC, AS, LD, AR, RANLIB, and OBJCOPY point to
98 the appropriate compiler, assembler, linker, ar, and ranlib binaries,
101 If you're working in a cross-development environment, be sure to set
102 macro ARCH to the desired target architecture ("ia32" for x86, "x86_64" for
103 x86_64 and "ia64" for IA-64). For convenience, this can also be done from
104 the make command line (e.g., "make ARCH=ia64").
109 To build the sample EFI applications provided in subdirectory "apps",
110 simply invoke "make" in the toplevel directory (the directory
111 containing this README file). This should build lib/libefi.a and
112 gnuefi/libgnuefi.a first and then all the EFI applications such as a
118 Just copy the EFI application (e.g., apps/t6.efi) to the EFI
119 filesystem, boot EFI, and then select "Invoke EFI application" to run
120 the application you want to test. Alternatively, you can invoke the
121 Intel-provided "nshell" application and then invoke your test binary
122 via the command line interface that "nshell" provides.
125 ** Writing Your Own EFI Application
127 Suppose you have your own EFI application in a file called
128 "apps/myefiapp.c". To get this application built by the GNU EFI build
129 environment, simply add "myefiapp.efi" to macro TARGETS in
130 apps/Makefile. Once this is done, invoke "make" in the top level
131 directory. This should result in EFI application apps/myefiapp.efi,
134 The GNU EFI build environment allows to write EFI applications as
135 described in Intel's EFI documentation, except for two differences:
137 - The EFI application's entry point is always called "efi_main". The
138 declaration of this routine is:
140 EFI_STATUS efi_main (EFI_HANDLE image, EFI_SYSTEM_TABLE *systab);
142 - UNICODE string literals must be written as W2U(L"Sample String")
143 instead of just L"Sample String". The W2U() macro is defined in
144 <efilib.h>. This header file also declares the function W2UCpy()
145 which allows to convert a wide string into a UNICODE string and
146 store the result in a programmer-supplied buffer.
148 - Calls to EFI services should be made via uefi_call_wrapper(). This
149 ensures appropriate parameter passing for the architecture.
152 * Part 2: Inner Workings
154 WARNING: This part contains all the gory detail of how the GNU EFI
155 toolchain works. Normal users do not have to worry about such
156 details. Reading this part incurs a definite risk of inducing severe
157 headaches or other maladies.
159 The basic idea behind the GNU EFI build environment is to use the GNU
160 toolchain to build a normal ELF binary that, at the end, is converted
161 to an EFI binary. EFI binaries are really just PE32+ binaries. PE
162 stands for "Portable Executable" and is the object file format
163 Microsoft is using on its Windows platforms. PE is basically the COFF
164 object file format with an MS-DOS2.0 compatible header slapped on in
165 front of it. The "32" in PE32+ stands for 32 bits, meaning that PE32
166 is a 32-bit object file format. The plus in "PE32+" indicates that
167 this format has been hacked to allow loading a 4GB binary anywhere in
168 a 64-bit address space (unlike ELF64, however, this is not a full
169 64-bit object file format because the entire binary cannot span more
170 than 4GB of address space). EFI binaries are plain PE32+ binaries
171 except that the "subsystem id" differs from normal Windows binaries.
172 There are two flavors of EFI binaries: "applications" and "drivers"
173 and each has there own subsystem id and are identical otherwise. At
174 present, the GNU EFI build environment supports the building of EFI
175 applications only, though it would be trivial to generate drivers, as
176 the only difference is the subsystem id. For more details on PE32+,
179 http://msdn.microsoft.com/library/specs/msdn_pecoff.htm.
181 In theory, converting a suitable ELF64 binary to PE32+ is easy and
182 could be accomplished with the "objcopy" utility by specifying option
183 --target=efi-app-ia32 (x86) or --target=efi-app-ia64 (IA-64). But
184 life never is that easy, so here some complicating factors:
186 (1) COFF sections are very different from ELF sections.
188 ELF binaries distinguish between program headers and sections.
189 The program headers describe the memory segments that need to
190 be loaded/initialized, whereas the sections describe what
191 constitutes those segments. In COFF (and therefore PE32+) no
192 such distinction is made. Thus, COFF sections need to be page
193 aligned and have a size that is a multiple of the page size
194 (4KB for EFI), whereas ELF allows sections at arbitrary
195 addresses and with arbitrary sizes.
197 (2) EFI binaries should be relocatable.
199 Since EFI binaries are executed in physical mode, EFI cannot
200 guarantee that a given binary can be loaded at its preferred
201 address. EFI does _try_ to load a binary at it's preferred
202 address, but if it can't do so, it will load it at another
203 address and then relocate the binary using the contents of the
206 (3) On IA-64, the EFI entry point needs to point to a function
207 descriptor, not to the code address of the entry point.
209 (4) The EFI specification assumes that wide characters use UNICODE
212 ANSI C does not specify the size or encoding that a wide
213 character uses. These choices are "implementation defined".
214 On most UNIX systems, the GNU toolchain uses a wchar_t that is
215 4 bytes in size. The encoding used for such characters is
218 In the following sections, we address how the GNU EFI build
219 environment addresses each of these issues.
222 ** (1) Accommodating COFF Sections
224 In order to satisfy the COFF constraint of page-sized and page-aligned
225 sections, the GNU EFI build environment uses the special linker script
226 in gnuefi/elf_$(ARCH)_efi.lds where $(ARCH) is the target architecture
227 ("ia32" for x86, "x86_64" for x86_64 and "ia64" for IA-64).
228 This script is set up to create only eight COFF section, each page aligned
229 and page sized.These eight sections are used to group together the much
230 greater number of sections that are typically present in ELF object files.
234 Collects the ELF .hash info (this section _must_ be the first
235 section in order to build a shared object file; the section is
236 not actually loaded or used at runtime).
239 Collects all sections containing executable code.
242 Collects read-only and read-write data, literal string data,
243 global offset tables, the uninitialized data segment (bss) and
244 various other sections containing data.
246 The reason read-only data is placed here instead of the in
247 .text is to make it possible to disassemble the .text section
248 without getting garbage due to read-only data. Besides, since
249 EFI binaries execute in physical mode, differences in page
250 protection do not matter.
252 The reason the uninitialized data is placed in this section is
253 that the EFI loader appears to be unable to handle sections
254 that are allocated but not loaded from the binary.
256 .dynamic, .dynsym, .rela, .rel, .reloc
257 These sections contains the dynamic information necessary to
258 self-relocate the binary (see below).
260 A couple of more points worth noting about the linker script:
262 o On IA-64, the global pointer symbol (__gp) needs to be placed such
263 that the _entire_ EFI binary can be addressed using the signed
264 22-bit offset that the "addl" instruction affords. Specifically,
265 this means that __gp should be placed at ImageBase + 0x200000.
266 Strictly speaking, only a couple of symbols need to be addressable
267 in this fashion, so with some care it should be possible to build
268 binaries much larger than 4MB. To get a list of symbols that need
269 to be addressable in this fashion, grep the assembly files in
270 directory gnuefi for the string "@gprel".
272 o The link address (ImageBase) of the binary is (arbitrarily) set to
273 zero. This could be set to something larger to increase the chance
274 of EFI being able to load the binary without requiring relocation.
275 However, a start address of 0 makes debugging a wee bit easier
276 (great for those of us who can add, but not subtract... ;-).
278 o The relocation related sections (.dynamic, .rel, .rela, .reloc)
279 cannot be placed inside .data because some tools in the GNU
280 toolchain rely on the existence of these sections.
282 o Some sections in the ELF binary intentionally get dropped when
283 building the EFI binary. Particularly noteworthy are the dynamic
284 relocation sections for the .plabel and .reloc sections. It would
285 be _wrong_ to include these sections in the EFI binary because it
286 would result in .reloc and .plabel being relocated twice (once by
287 the EFI loader and once by the self-relocator; see below for a
288 description of the latter). Specifically, only the sections
289 mentioned with the -j option in the final "objcopy" command are
290 retained in the EFI binary (see apps/Makefile).
293 ** (2) Building Relocatable Binaries
295 ELF binaries are normally linked for a fixed load address and are thus
296 not relocatable. The only kind of ELF object that is relocatable are
297 shared objects ("shared libraries"). However, even those objects are
298 usually not completely position independent and therefore require
299 runtime relocation by the dynamic loader. For example, IA-64 binaries
300 normally require relocation of the global offset table.
302 The approach to building relocatable binaries in the GNU EFI build
305 (a) build an ELF shared object
307 (b) link it together with a self-relocator that takes care of
308 applying the dynamic relocations that may be present in the
311 (c) convert the resulting image to an EFI binary
313 The self-relocator is of course architecture dependent. The x86
314 version can be found in gnuefi/reloc_ia32.c, the x86_64 version
315 can be found in gnuefi/reloc_x86_64.c and the IA-64 version can be
316 found in gnuefi/reloc_ia64.S.
318 The self-relocator operates as follows: the startup code invokes it
319 right after EFI has handed off control to the EFI binary at symbol
320 "_start". Upon activation, the self-relocator searches the .dynamic
321 section (whose starting address is given by symbol _DYNAMIC) for the
322 dynamic relocation information, which can be found in the DT_REL,
323 DT_RELSZ, and DT_RELENT entries of the dynamic table (DT_RELA,
324 DT_RELASZ, and DT_RELAENT in the case of rela relocations, as is the
325 case for IA-64). The dynamic relocation information points to the ELF
326 relocation table. Once this table is found, the self-relocator walks
327 through it, applying each relocation one by one. Since the EFI
328 binaries are fully resolved shared objects, only a subset of all
329 possible relocations need to be supported. Specifically, on x86 only
330 the R_386_RELATIVE relocation is needed. On IA-64, the relocations
331 R_IA64_DIR64LSB, R_IA64_REL64LSB, and R_IA64_FPTR64LSB are needed.
332 Note that the R_IA64_FPTR64LSB relocation requires access to the
333 dynamic symbol table. This is why the .dynsym section is included in
334 the EFI binary. Another complication is that this relocation requires
335 memory to hold the function descriptors (aka "procedure labels" or
336 "plabels"). Each function descriptor uses 16 bytes of memory. The
337 IA-64 self-relocator currently reserves a static memory area that can
338 hold 100 of these descriptors. If the self-relocator runs out of
339 space, it causes the EFI binary to fail with error code 5
340 (EFI_BUFFER_TOO_SMALL). When this happens, the manifest constant
341 MAX_FUNCTION_DESCRIPTORS in gnuefi/reloc_ia64.S should be increased
342 and the application recompiled. An easy way to count the number of
343 function descriptors required by an EFI application is to run the
346 objdump --dynamic-reloc example.so | fgrep FPTR64 | wc -l
348 assuming "example" is the name of the desired EFI application.
351 ** (3) Creating the Function Descriptor for the IA-64 EFI Binaries
353 As mentioned above, the IA-64 PE32+ format assumes that the entry
354 point of the binary is a function descriptor. A function descriptors
355 consists of two double words: the first one is the code entry point
356 and the second is the global pointer that should be loaded before
357 calling the entry point. Since the ELF toolchain doesn't know how to
358 generate a function descriptor for the entry point, the startup code
359 in gnuefi/crt0-efi-ia64.S crafts one manually by with the code:
361 .section .plabel, "a"
366 this places the procedure label for entry point _start in a section
367 called ".plabel". Now, the only problem is that _start and __gp need
368 to be relocated _before_ EFI hands control over to the EFI binary.
369 Fortunately, PE32+ defines a section called ".reloc" that can achieve
370 this. Thus, in addition to manually crafting the function descriptor,
371 the startup code also crafts a ".reloc" section that has will cause
372 the EFI loader to relocate the function descriptor before handing over
373 control to the EFI binary (again, see the PECOFF spec mentioned above
376 A final question may be why .plabel and .reloc need to go in their own
377 COFF sections. The answer is simply: we need to be able to discard
378 the relocation entries that are generated for these sections. By
379 placing them in these sections, the relocations end up in sections
380 ".rela.plabel" and ".rela.reloc" which makes it easy to filter them
381 out in the filter script. Also, the ".reloc" section needs to be in
382 its own section so that the objcopy program can recognize it and can
383 create the correct directory entries in the PE32+ binary.
386 ** (4) Convenient and Portable Generation of UNICODE String Literals
388 As of gnu-efi-3.0, we make use (and somewhat abuse) the gcc option
389 that forces wide characters (WCHAR_T) to use short integers (2 bytes)
390 instead of integers (4 bytes). This way we match the Unicode character
391 size. By abuse, we mean that we rely on the fact that the regular ASCII
392 characters are encoded the same way between (short) wide characters
393 and Unicode and basically only use the first byte. This allows us
394 to just use them interchangeably.
396 The gcc option to force short wide characters is : -fshort-wchar