Here is a very simple assembly program, just return 12 after executed.
$ cat a.asm
global _start
section .text
_start: mov rax, 60 ; system call for exit
mov rdi, 12 ; exit code 12
syscall
It can be built and executed correctly:
$ nasm -f elf64 a.asm && ld a.o && ./a.out || echo $?
12
But the size of a.out is big, it is more than 4k:
$ wc -c a.out
4664 a.out
I try to understand it by reading elf content:
$ readelf -l a.out
Elf file type is EXEC (Executable file)
Entry point 0x401000
There are 2 program headers, starting at offset 64
Program Headers:
Type Offset VirtAddr PhysAddr
FileSiz MemSiz Flags Align
LOAD 0x0000000000000000 0x0000000000400000 0x0000000000400000
0x00000000000000b0 0x00000000000000b0 R 0x1000
LOAD 0x0000000000001000 0x0000000000401000 0x0000000000401000
0x000000000000000c 0x000000000000000c R E 0x1000
Section to Segment mapping:
Segment Sections...
00
01 .text
it is strange, segment 00 is aligned by 0x1000, I think it means such segment at least will occupy 4096 bytes.
My question is what is this segment 00?
(nasm version 2.14.02, ld version 2.34, os is Ubuntu 20.04.1)
Since it starts at file offset zero, it is probably a "padding" segment introduced to make the loading of the ELF more efficient.
The .text segment will, in fact, be already aligned in the file as it should be in memory.
You can force ld not to align sections both in memory and in the file with -n. You can also strip the symbols with -s.
This will reduce the size to about 352 bytes.
Now the ELF contains:
The ELF header (Needed)
The program header table (Needed)
The code (Needed)
The string table (Possibly unneeded)
The section table (Possibly unneeded)
The string table can be removed, but apparently strips can't do that.
I've removed the .shstrtab section data and all the section headers manually to shrink the size down to 144 bytes.
Consider that 64 bytes come from the ELF header, 60 from the single program header and 12 from your code; for a total of 136 bytes.
The extra 8 bytes are padding, 4 bytes at the end of the code section (easy to remove), and one at the end of the program header (which requires a bit of patching).
Related
For a university course, I like to compare code-sizes of functionally similar programs if written and compiled using gcc/clang versus assembly. In the process of re-evaluating how to further shrink the size of some executables, I couldn't trust my eyes when the very same assembly code I assembled/linked 2 years ago now has grown >10x in size after building it again (which true for multiple programs, not only helloworld):
$ make
as -32 -o helloworld-asm-2020.o helloworld-asm-2020.s
ld -melf_i386 -o helloworld-asm-2020 helloworld-asm-2020.o
$ ls -l
-rwxr-xr-x 1 xxx users 708 Jul 18 2018 helloworld-asm-2018*
-rwxr-xr-x 1 xxx users 8704 Nov 25 15:00 helloworld-asm-2020*
-rwxr-xr-x 1 xxx users 4724 Nov 25 15:00 helloworld-asm-2020-n*
-rwxr-xr-x 1 xxx users 4228 Nov 25 15:00 helloworld-asm-2020-n-sstripped*
-rwxr-xr-x 1 xxx users 604 Nov 25 15:00 helloworld-asm-2020.o*
-rw-r--r-- 1 xxx users 498 Nov 25 14:44 helloworld-asm-2020.s
The assembly code is:
.code32
.section .data
msg: .ascii "Hello, world!\n"
len = . - msg
.section .text
.globl _start
_start:
movl $len, %edx # EDX = message length
movl $msg, %ecx # ECX = address of message
movl $1, %ebx # EBX = file descriptor (1 = stdout)
movl $4, %eax # EAX = syscall number (4 = write)
int $0x80 # call kernel by interrupt
# and exit
movl $0, %ebx # return code is zero
movl $1, %eax # exit syscall number (1 = exit)
int $0x80 # call kernel again
The same hello world program, compiled using GNU as and GNU ld (always using 32-bit assembly) was 708 bytes then, and has grown to 8.5K now. Even when telling the linker to turn off page alignment (ld -n), it still has almost 4.2K. stripping/sstripping doesn't pay off either.
readelf tells me that the start of section headers is much later in the code (byte 468 vs 8464), but I have no idea why. It's running on the same arch system as in 2018, the Makefile is the same and I'm not linking against any libraries (especially not libc). I guess something regarding ld has changed due to the fact that the object file is still quite small, but what and why?
Disclaimer: I'm building 32-bit executables on an x86-64 machine.
Edit: I'm using GNU binutils (as & ld) version 2.35.1 Here is a base64-encoded archive which includes the source and both executables (small old one, large new one) :
cat << EOF | base64 -d | tar xj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EOF
Update:
When using ld.gold instead of ld.bfd (to which /usr/bin/ld is symlinked to by default), the executable size becomes as small as expected:
$ cat Makefile
TARGET=helloworld
all:
as -32 -o ${TARGET}-asm.o ${TARGET}-asm.s
ld.bfd -melf_i386 -o ${TARGET}-asm-bfd ${TARGET}-asm.o
ld.gold -melf_i386 -o ${TARGET}-asm-gold ${TARGET}-asm.o
rm ${TARGET}-asm.o
$ make -q
$ ls -l
total 68
-rw-r--r-- 1 eso eso 200 Dec 1 13:57 Makefile
-rwxrwxr-x 1 eso eso 8700 Dec 1 13:57 helloworld-asm-bfd
-rwxrwxr-x 1 eso eso 732 Dec 1 13:57 helloworld-asm-gold
-rw-r--r-- 1 eso eso 498 Dec 1 13:44 helloworld-asm.s
Maybe I just used gold previously without being aware.
It's not 10x in general, it's page-alignment of a couple sections as Jester says, per changes to ld's default linker script for security reasons:
First change: Making sure data from .data isn't present in any of the mapping of .text, so none of that static data is available for ROP / Spectre gadgets in an executable page. (In older ld, that meant the program-headers mapped the same disk-block twice, also into a RW-without-exec segment for the actual .data section. The executable mapping was still read-only.)
More recent change: Separate .rodata from .text into separate segments, again so static data isn't mapped into an executable page. Previously, const char code[]= {...} could be cast to a function pointer and called, without needing mprotect or gcc -z execstack or other tricks, if you wanted to test shellcode that way. (A separate Linux kernel change made -z execstack only apply to the actual stack, not READ_IMPLIES_EXEC.)
See Why an ELF executable could have 4 LOAD segments? for this history, including the strange fact that .rodata is in a separate segment from the read-only mapping for access to the ELF metadata.
That extra space is just 00 padding and will compress well in a .tar.gz or whatever.
So it has a worst-case upper bound of about 2x 4k extra pages of padding, and tiny executables are close to that worst case.
gcc -Wl,--nmagic will turn off page-alignment of sections if you want that for some reason. (see the ld(1) man page) I don't know why that doesn't pack everything down to the old size. Perhaps checking the default linker script would shed some light, but it's pretty long. Run ld --verbose to see it.
stripping won't help for padding that's part of a section; I think it can only remove whole sections.
ld -z noseparate-code uses the old layout, only 2 total segments to cover the .text and .rodata sections, and the .data and .bss sections. (And the ELF metadata that dynamic linking wants access to.)
Related:
Linking with gcc instead of ld
This question is about ld, but note that if you're using gcc -nostdlib, that used to also default to making a static executable. But modern Linux distros config GCC with -pie as the default, and GCC won't make a static-pie by default even if there aren't any shared libraries being linked. Unlike with -no-pie mode where it will simply make a static executable in that case. (A static-pie still needs startup code to apply relocations for any absolute addresses.)
So the equivalent of ld directly is gcc -nostdlib -static (which implies -no-pie). Or gcc -nostdlib -no-pie should let it default to -static when there are no shared libs being linked. You can combine this with -Wl,--nmagic and/or -Wl,-z -Wl,noseparate-code.
Also:
A Whirlwind Tutorial on Creating Really Teensy ELF Executables for Linux - eventually making a 45 byte executable, with the machine code for an _exit syscall stuffed into the ELF program header itself.
FASM can make quite small executables, using its mode where it outputs a static executable (not object file) directly with no ELF section metadata, just program headers. (It's a pain to debug with GDB or disassemble with objdump; most tools assume there will be section headers, even though they're not needed to run static executables.)
What is a reasonable minimum number of assembly instructions for a small C program including setup?
What's the difference between "statically linked" and "not a dynamic executable" from Linux ldd? (static vs. static-pie vs. (dynamic) PIE that happens to have no shared libraries.)
For a university course, I like to compare code-sizes of functionally similar programs if written and compiled using gcc/clang versus assembly. In the process of re-evaluating how to further shrink the size of some executables, I couldn't trust my eyes when the very same assembly code I assembled/linked 2 years ago now has grown >10x in size after building it again (which true for multiple programs, not only helloworld):
$ make
as -32 -o helloworld-asm-2020.o helloworld-asm-2020.s
ld -melf_i386 -o helloworld-asm-2020 helloworld-asm-2020.o
$ ls -l
-rwxr-xr-x 1 xxx users 708 Jul 18 2018 helloworld-asm-2018*
-rwxr-xr-x 1 xxx users 8704 Nov 25 15:00 helloworld-asm-2020*
-rwxr-xr-x 1 xxx users 4724 Nov 25 15:00 helloworld-asm-2020-n*
-rwxr-xr-x 1 xxx users 4228 Nov 25 15:00 helloworld-asm-2020-n-sstripped*
-rwxr-xr-x 1 xxx users 604 Nov 25 15:00 helloworld-asm-2020.o*
-rw-r--r-- 1 xxx users 498 Nov 25 14:44 helloworld-asm-2020.s
The assembly code is:
.code32
.section .data
msg: .ascii "Hello, world!\n"
len = . - msg
.section .text
.globl _start
_start:
movl $len, %edx # EDX = message length
movl $msg, %ecx # ECX = address of message
movl $1, %ebx # EBX = file descriptor (1 = stdout)
movl $4, %eax # EAX = syscall number (4 = write)
int $0x80 # call kernel by interrupt
# and exit
movl $0, %ebx # return code is zero
movl $1, %eax # exit syscall number (1 = exit)
int $0x80 # call kernel again
The same hello world program, compiled using GNU as and GNU ld (always using 32-bit assembly) was 708 bytes then, and has grown to 8.5K now. Even when telling the linker to turn off page alignment (ld -n), it still has almost 4.2K. stripping/sstripping doesn't pay off either.
readelf tells me that the start of section headers is much later in the code (byte 468 vs 8464), but I have no idea why. It's running on the same arch system as in 2018, the Makefile is the same and I'm not linking against any libraries (especially not libc). I guess something regarding ld has changed due to the fact that the object file is still quite small, but what and why?
Disclaimer: I'm building 32-bit executables on an x86-64 machine.
Edit: I'm using GNU binutils (as & ld) version 2.35.1 Here is a base64-encoded archive which includes the source and both executables (small old one, large new one) :
cat << EOF | base64 -d | tar xj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EOF
Update:
When using ld.gold instead of ld.bfd (to which /usr/bin/ld is symlinked to by default), the executable size becomes as small as expected:
$ cat Makefile
TARGET=helloworld
all:
as -32 -o ${TARGET}-asm.o ${TARGET}-asm.s
ld.bfd -melf_i386 -o ${TARGET}-asm-bfd ${TARGET}-asm.o
ld.gold -melf_i386 -o ${TARGET}-asm-gold ${TARGET}-asm.o
rm ${TARGET}-asm.o
$ make -q
$ ls -l
total 68
-rw-r--r-- 1 eso eso 200 Dec 1 13:57 Makefile
-rwxrwxr-x 1 eso eso 8700 Dec 1 13:57 helloworld-asm-bfd
-rwxrwxr-x 1 eso eso 732 Dec 1 13:57 helloworld-asm-gold
-rw-r--r-- 1 eso eso 498 Dec 1 13:44 helloworld-asm.s
Maybe I just used gold previously without being aware.
It's not 10x in general, it's page-alignment of a couple sections as Jester says, per changes to ld's default linker script for security reasons:
First change: Making sure data from .data isn't present in any of the mapping of .text, so none of that static data is available for ROP / Spectre gadgets in an executable page. (In older ld, that meant the program-headers mapped the same disk-block twice, also into a RW-without-exec segment for the actual .data section. The executable mapping was still read-only.)
More recent change: Separate .rodata from .text into separate segments, again so static data isn't mapped into an executable page. Previously, const char code[]= {...} could be cast to a function pointer and called, without needing mprotect or gcc -z execstack or other tricks, if you wanted to test shellcode that way. (A separate Linux kernel change made -z execstack only apply to the actual stack, not READ_IMPLIES_EXEC.)
See Why an ELF executable could have 4 LOAD segments? for this history, including the strange fact that .rodata is in a separate segment from the read-only mapping for access to the ELF metadata.
That extra space is just 00 padding and will compress well in a .tar.gz or whatever.
So it has a worst-case upper bound of about 2x 4k extra pages of padding, and tiny executables are close to that worst case.
gcc -Wl,--nmagic will turn off page-alignment of sections if you want that for some reason. (see the ld(1) man page) I don't know why that doesn't pack everything down to the old size. Perhaps checking the default linker script would shed some light, but it's pretty long. Run ld --verbose to see it.
stripping won't help for padding that's part of a section; I think it can only remove whole sections.
ld -z noseparate-code uses the old layout, only 2 total segments to cover the .text and .rodata sections, and the .data and .bss sections. (And the ELF metadata that dynamic linking wants access to.)
Related:
Linking with gcc instead of ld
This question is about ld, but note that if you're using gcc -nostdlib, that used to also default to making a static executable. But modern Linux distros config GCC with -pie as the default, and GCC won't make a static-pie by default even if there aren't any shared libraries being linked. Unlike with -no-pie mode where it will simply make a static executable in that case. (A static-pie still needs startup code to apply relocations for any absolute addresses.)
So the equivalent of ld directly is gcc -nostdlib -static (which implies -no-pie). Or gcc -nostdlib -no-pie should let it default to -static when there are no shared libs being linked. You can combine this with -Wl,--nmagic and/or -Wl,-z -Wl,noseparate-code.
Also:
A Whirlwind Tutorial on Creating Really Teensy ELF Executables for Linux - eventually making a 45 byte executable, with the machine code for an _exit syscall stuffed into the ELF program header itself.
FASM can make quite small executables, using its mode where it outputs a static executable (not object file) directly with no ELF section metadata, just program headers. (It's a pain to debug with GDB or disassemble with objdump; most tools assume there will be section headers, even though they're not needed to run static executables.)
What is a reasonable minimum number of assembly instructions for a small C program including setup?
What's the difference between "statically linked" and "not a dynamic executable" from Linux ldd? (static vs. static-pie vs. (dynamic) PIE that happens to have no shared libraries.)
Take a simple hello world program and compile it as follows:
> g++ --version
g++ 6.3.0
Copyright (C) 2016 Free Software Foundation, Inc.
This is free software; see the source for copying conditions. There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
> g++ -fuse-ld=gold test.cpp -o test
Inspecting the binary produced:
> readelf -l ./test
Elf file type is EXEC (Executable file)
Entry point 0x400750
There are 9 program headers, starting at offset 64
Program Headers:
Type Offset VirtAddr PhysAddr
FileSiz MemSiz Flags Align
PHDR 0x0000000000000040 0x0000000000400040 0x0000000000400040
0x00000000000001f8 0x00000000000001f8 R 8
INTERP 0x0000000000000238 0x0000000000400238 0x0000000000400238
0x000000000000001c 0x000000000000001c R 1
[Requesting program interpreter: /lib64/ld-linux-x86-64.so.2]
LOAD 0x0000000000000000 0x0000000000400000 0x0000000000400000
0x0000000000000ac8 0x0000000000000ac8 R E 1000
LOAD 0x0000000000000dc0 0x0000000000401dc0 0x0000000000401dc0
0x0000000000000288 0x00000000000003d0 RW 1000
DYNAMIC 0x0000000000000de0 0x0000000000401de0 0x0000000000401de0
0x0000000000000200 0x0000000000000200 RW 8
NOTE 0x0000000000000254 0x0000000000400254 0x0000000000400254
0x0000000000000044 0x0000000000000044 R 4
GNU_EH_FRAME 0x0000000000000a8c 0x0000000000400a8c 0x0000000000400a8c
0x000000000000003c 0x000000000000003c R 4
GNU_STACK 0x0000000000000000 0x0000000000000000 0x0000000000000000
0x0000000000000000 0x0000000000000000 RW 0
GNU_RELRO 0x0000000000000dc0 0x0000000000401dc0 0x0000000000401dc0
0x0000000000000240 0x0000000000000240 RW 8
Section to Segment mapping:
Segment Sections...
00
01 .interp
02 .interp .note.ABI-tag .note.gnu.build-id .dynsym .dynstr .gnu.hash .gnu.version .gnu.version_r .rela.dyn .rela.plt .init .plt .text .fini .rodata .eh_frame .eh_frame_hdr
03 .jcr .fini_array .init_array .dynamic .got .got.plt .data .bss
04 .dynamic
05 .note.ABI-tag .note.gnu.build-id
06 .eh_frame_hdr
07
08 .jcr .fini_array .init_array .dynamic .got
Notice that the interpreter used is ld. Whilst the program happens to work, I've not been able to find any information on whether this is safe. For all I know, gold interprets the ELF specification in a subtly different and incompatible way that requires a different interpreter.
I've done my best to research this, but have been unable to find anything that answers my question. The closest I've found is that gold struggles to link the Linux kernel (or struggled, since time has past and it may have been fixed).
You're falling into the naming trap — gold is a link editor, while ld.so is a dynamic loader. Although at different times, they are called linkers (the latter often referred to also as runtime linker.)
Their scope and usage is very different, the first one generates the final executable you'll eventually run, while the latter takes the generated file, finds its dependencies, and resolve (links) the undefined symbols between those.
Indeed, gold and ld (precisely, bfd-ld), the link editors, are provided by binutils (or alternative toolchain packages such as clang and so on), while ld.so is provided by the C library package, usually glibc on Linux distributions, but alternatively uclibc or musl.
Combining this with Martin Rosenau's comment...
Looking at the content of /usr/bin/gold, you can see that the string /lib64/ld-linux-x86-64.so.2 is stored inside the gold executable. This means that the gold linker itself "decides" using that runtime interpreter. For this reason I doubt that there are incompatibilities.
... ld.so should be compatible with the gold linker.
Consider this AMD64 assembly program:
.globl _start
_start:
xorl %edi, %edi
movl $60, %eax
syscall
If I compile that with gcc -nostdlib and run ldd a.out, I get this:
statically linked
If I instead compile that with gcc -static -nostdlib and run ldd a.out, I get this:
not a dynamic executable
What's the difference between statically linked and not a dynamic executable? And if my binary was already statically linked, why does adding -static affect anything?
There are two separate things here:
Requesting an ELF interpreter (ld.so) or not.
Like #!/bin/sh but for binaries, runs before your _start.
This is the difference between a static vs. dynamic executable.
The list of dynamically linked libraries for ld.so to load happens to be empty.
This is apparently what ldd calls "statically linked", i.e. that any libraries you might have linked at build time were static libraries.
Other tools like file and readelf give more information and use terminology that matches what you'd expect.
Your GCC is configured so -pie is the default, and gcc doesn't make a static-pie for the special case of no dynamic libraries.
gcc -nostdlib just makes a PIE that happens not to link to any libraries but is otherwise identical to a normal PIE, specifying an ELF interpreter.
ldd confusingly calls this "statically linked".
file : ELF 64-bit LSB pie executable, x86-64, version 1 (SYSV), dynamically linked, interpreter /lib64/ld-linux-x86-64.so.2 ...
gcc -nostdlib -static overrides the -pie default and makes a true static executable.
file : ELF 64-bit LSB executable, x86-64, version 1 (SYSV), statically linked ...
gcc -nostdlib -no-pie also chooses to make a static executable as an optimization for the case where there are no dynamic libraries at all. Since a non-PIE executable couldn't have been ASLRed anyway, this makes sense. Byte-for-byte identical to the -static case.
gcc -nostdlib -static-pie makes an ASLRable executable that doesn't need an ELF interpreter. GCC doesn't do this by default for gcc -pie -nostdlib, unlike the no-pie case where it chooses to sidestep ld.so when no dynamically-linked libraries are involved.
file : ELF 64-bit LSB pie executable, x86-64, version 1 (SYSV), statically linked ...
-static-pie is obscure, rarely used, and older file doesn't identify it as statically linked.
-nostdlib doesn't imply -no-pie or -static, and -static-pie has to be explicitly specified to get that.
gcc -static-pie invokes ld -static -pie, so ld has to know what that means. Unlike with the non-PIE case where you don't have to ask for a dynamic executable explicitly, you just get one if you pass ld any .so libraries. I think that's why you happen to get a static executable from gcc -nostdlib -no-pie - GCC doesn't have to do anything special, it's just ld doing that optimization.
But ld doesn't enable -static implicitly when -pie is specified, even when there are no shared libraries to link.
Details
Examples generated with gcc --version gcc (Arch Linux 9.3.0-1) 9.3.0
ld --version GNU ld (GNU Binutils) 2.34 (also readelf is binutils)
ldd --version ldd (GNU libc) 2.31
file --version file-5.38 - note that static-pie detection has changed in recent patches, with Ubuntu cherry-picking an unreleased patch. (Thanks #Joseph for the detective work) - this in 2019 detected dynamic = having a PT_INTERP to handle static-pie, but it was reverted to detect based on PT_DYNAMIC so shared libraries count as dynamic. debian bug #948269. static-pie is an obscure rarely-used feature.
GCC ends up running ld -pie exit.o with a dynamic linker path specified, and no libraries. (And a boatload of other options to support possible LTO link-time optimization, but the keys here are -dynamic-linker /lib64/ld-linux-x86-64.so.2 -pie. collect2 is just a wrapper around ld.)
$ gcc -nostdlib exit.s -v # output manually line wrapped with \ for readability
...
COLLECT_GCC_OPTIONS='-nostdlib' '-v' '-mtune=generic' '-march=x86-64'
/usr/lib/gcc/x86_64-pc-linux-gnu/9.3.0/collect2 \
-plugin /usr/lib/gcc/x86_64-pc-linux-gnu/9.3.0/liblto_plugin.so \
-plugin-opt=/usr/lib/gcc/x86_64-pc-linux-gnu/9.3.0/lto-wrapper \
-plugin-opt=-fresolution=/tmp/ccoNx1IR.res \
--build-id --eh-frame-hdr --hash-style=gnu \
-m elf_x86_64 -dynamic-linker /lib64/ld-linux-x86-64.so.2 -pie \
-L/usr/lib/gcc/x86_64-pc-linux-gnu/9.3.0 \
-L/usr/lib/gcc/x86_64-pc-linux-gnu/9.3.0/../../../../lib -L/lib/../lib \
-L/usr/lib/../lib \
-L/usr/lib/gcc/x86_64-pc-linux-gnu/9.3.0/../../.. \
/tmp/cctm2fSS.o
You get a dynamic PIE with no dependencies on other libraries. Running it still invokes the "ELF interpreter" /lib64/ld-linux-x86-64.so.2 on it which runs before jumping to your _start. (Although the kernel has already mapped the executable's ELF segments to ASLRed virtual addresses, along with ld.so's text / data / bss).
file and readelf are more descriptive.
PIE non-static executable from gcc -nostdlib
$ gcc -nostdlib exit.s -o exit-default
$ ls -l exit-default
-rwxr-xr-x 1 peter peter 13536 May 2 02:15 exit-default
$ ldd exit-default
statically linked
$ file exit-default
exit-default: ELF 64-bit LSB pie executable, x86-64, version 1 (SYSV), dynamically linked, interpreter /lib64/ld-linux-x86-64.so.2, BuildID[sha1]=05a4d1bdbc94d6f91cca1c9c26314e1aa227a3a5, not stripped
$ readelf -a exit-default
...
Type: DYN (Shared object file)
Machine: Advanced Micro Devices X86-64
Version: 0x1
Entry point address: 0x1000
...
Program Headers:
Type Offset VirtAddr PhysAddr
FileSiz MemSiz Flags Align
PHDR 0x0000000000000040 0x0000000000000040 0x0000000000000040
0x00000000000001f8 0x00000000000001f8 R 0x8
INTERP 0x0000000000000238 0x0000000000000238 0x0000000000000238
0x000000000000001c 0x000000000000001c R 0x1
[Requesting program interpreter: /lib64/ld-linux-x86-64.so.2]
LOAD 0x0000000000000000 0x0000000000000000 0x0000000000000000
0x00000000000002b1 0x00000000000002b1 R 0x1000
LOAD 0x0000000000001000 0x0000000000001000 0x0000000000001000
0x0000000000000009 0x0000000000000009 R E 0x1000
... (the Read+Exec segment to be mapped at virt addr 0x1000 is where your text section was linked.)
If you strace it you can also see the differences:
$ gcc -nostdlib exit.s -o exit-default
$ strace ./exit-default
execve("./exit-default", ["./exit-default"], 0x7ffe1f526040 /* 51 vars */) = 0
brk(NULL) = 0x5617eb1e4000
arch_prctl(0x3001 /* ARCH_??? */, 0x7ffcea703380) = -1 EINVAL (Invalid argument)
access("/etc/ld.so.preload", R_OK) = -1 ENOENT (No such file or directory)
mmap(NULL, 8192, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x7f9ff5b3e000
arch_prctl(ARCH_SET_FS, 0x7f9ff5b3ea80) = 0
mprotect(0x5617eabac000, 4096, PROT_READ) = 0
exit(0) = ?
+++ exited with 0 +++
vs. -static and -static-pie the first instruction executed in user-space is your _start (which you can also check with GDB using starti).
$ strace ./exit-static-pie
execve("./exit-static-pie", ["./exit-static-pie"], 0x7ffcdac96dd0 /* 51 vars */) = 0
exit(0) = ?
+++ exited with 0 +++
gcc -nostdlib -static-pie
$ gcc -nostdlib -static-pie exit.s -o exit-static-pie
$ ls -l exit-static-pie
-rwxr-xr-x 1 peter peter 13440 May 2 02:18 exit-static-pie
peter#volta:/tmp$ ldd exit-static-pie
statically linked
peter#volta:/tmp$ file exit-static-pie
exit-static-pie: ELF 64-bit LSB pie executable, x86-64, version 1 (SYSV), statically linked, BuildID[sha1]=daeb4a8f11bec1bb1aaa13cd48d24b5795af638e, not stripped
$ readelf -a exit-static-pie
...
Type: DYN (Shared object file)
Machine: Advanced Micro Devices X86-64
Version: 0x1
Entry point address: 0x1000
...
Program Headers:
Type Offset VirtAddr PhysAddr
FileSiz MemSiz Flags Align
LOAD 0x0000000000000000 0x0000000000000000 0x0000000000000000
0x0000000000000229 0x0000000000000229 R 0x1000
LOAD 0x0000000000001000 0x0000000000001000 0x0000000000001000
0x0000000000000009 0x0000000000000009 R E 0x1000
... (no Interp header, but still a read+exec text segment)
Notice that the addresses are still relative to the image base, leaving ASLR up to the kernel.
Surprisingly, ldd doesn't say that it's not a dynamic executable. That might be a bug, or a side effect of some implementation detail.
gcc -nostdlib -static traditional non-PIE old-school static executable
$ gcc -nostdlib -static exit.s -o exit-static
$ ls -l exit-static
-rwxr-xr-x 1 peter peter 4744 May 2 02:26 exit-static
peter#volta:/tmp$ ldd exit-static
not a dynamic executable
peter#volta:/tmp$ file exit-static
exit-static: ELF 64-bit LSB executable, x86-64, version 1 (SYSV), statically linked, BuildID[sha1]=1b03e3d05709b7288fe3006b4696fd0c11fb1cb2, not stripped
peter#volta:/tmp$ readelf -a exit-static
ELF Header:
...
Type: EXEC (Executable file)
Machine: Advanced Micro Devices X86-64
Version: 0x1
Entry point address: 0x401000
... (Note the absolute entry-point address nailed down at link time)
(And that the ELF type is EXEC, not DYN)
Program Headers:
Type Offset VirtAddr PhysAddr
FileSiz MemSiz Flags Align
LOAD 0x0000000000000000 0x0000000000400000 0x0000000000400000
0x000000000000010c 0x000000000000010c R 0x1000
LOAD 0x0000000000001000 0x0000000000401000 0x0000000000401000
0x0000000000000009 0x0000000000000009 R E 0x1000
NOTE 0x00000000000000e8 0x00000000004000e8 0x00000000004000e8
0x0000000000000024 0x0000000000000024 R 0x4
Section to Segment mapping:
Segment Sections...
00 .note.gnu.build-id
01 .text
02 .note.gnu.build-id
...
Those are all the program headers; unlike pie / static-pie I'm not leaving any out, just other whole parts of the readelf -a output.
Also note the absolute virtual addresses in the program headers that don't give the kernel a choice where in virtual address space to map the file. This is the difference between EXEC and DYN types of ELF objects. PIE executables are shared objects with an entry point, allowing us to get ASLR for the main executable. Actual EXEC executables have a link-time-chosen memory layout.
ldd apparently only reports "not a dynamic executable" when both:
no ELF interpreter (dynamic linker) path
ELF type = EXEC
I need to disassemble /proc/kcore file in Linux and I need to obtain virtual addresses of some special instructions to put kprobes later on it. According to this document /proc/kcore is an image of physical memory, but in this question someone answered that it is kernel's virtual memory (exactly what I am looking for).
When I use objdump tool to disassemble it, it starts with address something like f7c0b000, but udis86 starts with 0x0 (and totally different instruction). When I try to grep some specific instruction, let's say mov 0xf7c1d60c,%edx, I got:
objdump
f7c0b022 mov 0xf7c1d60c,%edx
udis86
290ec02a mov 0xf7c1d60c,%edx
It looks like the offset between udis86 and objdump is always 0xbffff000. Why so strange offset? How can I obtain virtual address of specific instruction? Somewhere I've read, that kernel is statically mapped at virtual address 0xc0000000 + 0x100000. If /proc/kcore is really physical image, is it correct only to add 0x100000 to addresses returned by objdump and I will get virtual address?
objdump understands ELF format files (such as /proc/kcore). It is able to extract the executable sections of the file while ignoring non-executable content (such as .note sections).
You can see the structure of an ELF exectuable using the -h flag, for example:
# objdump -h /proc/kcore
/proc/kcore: file format elf64-x86-64
Sections:
Idx Name Size VMA LMA File off Algn
0 note0 00001944 0000000000000000 0000000000000000 000002a8 2**0
CONTENTS, READONLY
1 .reg/0 000000d8 0000000000000000 0000000000000000 0000032c 2**2
CONTENTS
2 .reg 000000d8 0000000000000000 0000000000000000 0000032c 2**2
CONTENTS
3 load1 00800000 ffffffffff600000 0000000000000000 7fffff602000 2**12
CONTENTS, ALLOC, LOAD, CODE
(...)
It looks like the udcli tool from udis86 probably starts disassembling things from the beginning of the file, which suggests that your output will probably start with a bunch of irrelevant output and it's up to you to figure out where execution starts.
UPDATE
Here's the verification. We use this answer to extract the first load section from /proc/kcore, like this:
# dd if=/proc/kcore of=mysection bs=1 skip=$[0x7fffff602000] count=$[0x00800000]
And now if we view that with udcli:
# udcli mysection
0000000000000000 48 dec eax
0000000000000001 c7c060000000 mov eax, 0x60
0000000000000007 0f05 syscall
0000000000000009 c3 ret
000000000000000a cc int3
000000000000000b cc int3
We see that it looks almost identical to the output of objdump -d /proc/kcore:
# objdump -d /proc/kcore
/proc/kcore: file format elf64-x86-64
Disassembly of section load1:
ffffffffff600000 <load1>:
ffffffffff600000: 48 c7 c0 60 00 00 00 mov $0x60,%rax
ffffffffff600007: 0f 05 syscall
ffffffffff600009: c3 retq
ffffffffff60000a: cc int3
ffffffffff60000b: cc int3