I'm using valgrind/callgrind to profile my server code for some optimization. The two most used calls that callgrind is reporting to me (using kcachegrind to view) are _dl_lookup_symbol_x and do_lookup_x. However I have no idea what either of these are and can't seem to find any documentation about them.
Could anyone please tell me where these two functions are used and what they do?
_dl_lookup_symbol_x is an internal function inside the glibc C runtime library. If you browse the source for glibc, you'll find this comment above the _dl_lookup_symbol_x definition:
/* Search loaded objects' symbol tables for a definition of the symbol
UNDEF_NAME, perhaps with a requested version for the symbol.
do_lookup_x is merely a helper function called within the _dl_lookup_symbol_x function.
I'm no expert on the internals of glibc, but from what I can gather, _dl_lookup_symbol_x looks for a symbol (such as a function) inside shared libraries loaded by your program.
I don't know why these functions are called so often in your profiling, but at least now you have some clue as to what they do. Your profiling should tell you what functions are responsible for calling _dl_lookup_symbol_x so often.
Note that it would be normal for _dl_lookup_symbol_x to be called many times when the program first starts, as the runtime figures out the addresses of shared library functions with a given name. If you're profiling a very short-lived program, then it's not surprising that you'd see that most of the time is spent in internal "housekeeping" functions rather than your own code.
dl_lookup_symbol is called from _dl_fixup which is called from _dl_runtime_resolve_sse
In gdb if you break on the first function call, any call, then in lazy linking, which is default when using gcc, the function's address will not actually be resolved until after that call pretty much returns to main. One of the functions that the process of uses to resolve the address of the function is dl_lookup_symbol, so you should expect to see it once for each function that is called by every program that uses lazy linking, pretty much all of them.
Source code:
#include <stdio.h>
#include <string.h>
int main (void)
{
size_t a = strlen("Hello");
printf("%s is %d letters\n", "Hello", a);
if (a != sizeof("Hello") - 1)
{
perror("What?");
}
size_t b = strlen("Bye");
printf("%s is %d letters\n", "Bye", b);
if (b != sizeof("Bye") - 1)
{
perror("What?");
}
return 0;
}
Going to gdb:
(gdb) disas
Dump of assembler code for function main:
0x0000555555554700 <+0>: push rbp
0x0000555555554701 <+1>: mov rbp,rsp
0x0000555555554704 <+4>: sub rsp,0x10
0x0000555555554708 <+8>: mov QWORD PTR [rbp-0x8],0x5
0x0000555555554710 <+16>: mov rax,QWORD PTR [rbp-0x8]
0x0000555555554714 <+20>: mov rdx,rax
0x0000555555554717 <+23>: lea rsi,[rip+0xf6] # 0x555555554814
0x000055555555471e <+30>: lea rdi,[rip+0xf5] # 0x55555555481a
0x0000555555554725 <+37>: mov eax,0x0
=> 0x000055555555472a <+42>: call 0x5555555545a0 <printf@plt>
0x000055555555472f <+47>: cmp QWORD PTR [rbp-0x8],0x5
0x0000555555554734 <+52>: je 0x555555554742 <main+66>
0x0000555555554736 <+54>: lea rdi,[rip+0xef] # 0x55555555482c
0x000055555555473d <+61>: call 0x5555555545b0 <perror@plt>
0x0000555555554742 <+66>: mov QWORD PTR [rbp-0x10],0x3
0x000055555555474a <+74>: mov rax,QWORD PTR [rbp-0x10]
0x000055555555474e <+78>: mov rdx,rax
0x0000555555554751 <+81>: lea rsi,[rip+0xda] # 0x555555554832
0x0000555555554758 <+88>: lea rdi,[rip+0xbb] # 0x55555555481a
0x000055555555475f <+95>: mov eax,0x0
0x0000555555554764 <+100>: call 0x5555555545a0 <printf@plt>
0x0000555555554769 <+105>: cmp QWORD PTR [rbp-0x10],0x3
0x000055555555476e <+110>: je 0x55555555477c <main+124>
0x0000555555554770 <+112>: lea rdi,[rip+0xb5] # 0x55555555482c
0x0000555555554777 <+119>: call 0x5555555545b0 <perror@plt>
0x000055555555477c <+124>: mov eax,0x0
0x0000555555554781 <+129>: leave
0x0000555555554782 <+130>: ret
The following, after stepping into printf, is where lazy linking's magic results show up, after it's done. Here I'm just showing that as I step through this, the program walks through, but once printf is resolved, if I stepi at
0x00005555555545a0 <+0>: jmp QWORD PTR [rip+0x200a72]
the program will go straight to printf. So, next time the function is called jmp QWORD PTR [rip+0x200a72] will take us straight to printf. All code after this point and until the function returns to main is just the resolving of printf's address and only happens the first time that the function is called.
(gdb) disas
Dump of assembler code for function printf@plt:
=> 0x00005555555545a0 <+0>: jmp QWORD PTR [rip+0x200a72] # 0x555555755018
0x00005555555545a6 <+6>: push 0x0
0x00005555555545ab <+11>: jmp 0x555555554590
(gdb) disas
Dump of assembler code for function printf@plt:
=> 0x00005555555545a0 <+0>: jmp QWORD PTR [rip+0x200a72] # 0x555555755018
0x00005555555545a6 <+6>: push 0x0
0x00005555555545ab <+11>: jmp 0x555555554590
End of assembler dump.
(gdb) stepi
0x00005555555545a6 in printf@plt ()
(gdb) disas
Dump of assembler code for function printf@plt:
0x00005555555545a0 <+0>: jmp QWORD PTR [rip+0x200a72] # 0x555555755018
=> 0x00005555555545a6 <+6>: push 0x0
0x00005555555545ab <+11>: jmp 0x555555554590
End of assembler dump.
(gdb) stepi
0x00005555555545ab in printf@plt ()
(gdb) disas
Dump of assembler code for function printf@plt:
0x00005555555545a0 <+0>: jmp QWORD PTR [rip+0x200a72] # 0x555555755018
0x00005555555545a6 <+6>: push 0x0
=> 0x00005555555545ab <+11>: jmp 0x555555554590
Stepping in. Not sure why gdb doesn't pick these up, they are just instructions in the .plt table. And yes the addresses aren't the same, my system adds 0x555555554000 to everything before we even hit _start
(gdb) stepi
0x0000555555554590 in ?? ()
(gdb) disas
No function contains program counter for selected frame.
(gdb) stepi
0x0000555555554596 in ?? ()
(gdb) disas
No function contains program counter for selected frame.
, proof:
$ gdb -q ./simpletest
Reading symbols from ./simpletest...done.
(gdb) break _start
Breakpoint 1 at 0x5d0
(gdb) run
Starting program: /flerb/flerb/flerb/simpletest
Breakpoint 1, 0x00005555555545d0 in _start ()
(gdb)
From objdump:
Disassembly of section .plt:
0000000000000590 <.plt>:
590: ff 35 72 0a 20 00 push QWORD PTR [rip+0x200a72] # 201008 <_GLOBAL_OFFSET_TABLE_+0x8>
596: ff 25 74 0a 20 00 jmp QWORD PTR [rip+0x200a74] # 201010 <_GLOBAL_OFFSET_TABLE_+0x10>
59c: 0f 1f 40 00 nop DWORD PTR [rax+0x0]
gdb
(gdb) stepi
_dl_runtime_resolve_sse () at ../sysdeps/x86_64/dl-trampoline.h:164
164 ../sysdeps/x86_64/dl-trampoline.h: No such file or directory.
Hey, _dl_runtime_resolve_sse, we're getting closer.The following might look ugly. It's not. It's just saving registers so that the call 0x00007ffff7def11a <+154>: call 0x7ffff7de7970 <_dl_fixup> halfway through can use whatever registers it needs to, and that's truly where the magic happens. At the end 0x00007ffff7def1b0 <+304>: bnd jmp r11 will take us to printf just this one time that the program goes through this code for this particular function, because it's the only time that the program goes through this code for this particular function.
(gdb) disas
Dump of assembler code for function _dl_runtime_resolve_sse:
=> 0x00007ffff7def080 <+0>: push rbx
0x00007ffff7def081 <+1>: mov rbx,rsp
0x00007ffff7def084 <+4>: and rsp,0xfffffffffffffff0
0x00007ffff7def088 <+8>: sub rsp,0x100
0x00007ffff7def08f <+15>: mov QWORD PTR [rsp+0xc0],rax
0x00007ffff7def097 <+23>: mov QWORD PTR [rsp+0xc8],rcx
0x00007ffff7def09f <+31>: mov QWORD PTR [rsp+0xd0],rdx
0x00007ffff7def0a7 <+39>: mov QWORD PTR [rsp+0xd8],rsi
0x00007ffff7def0af <+47>: mov QWORD PTR [rsp+0xe0],rdi
0x00007ffff7def0b7 <+55>: mov QWORD PTR [rsp+0xe8],r8
0x00007ffff7def0bf <+63>: mov QWORD PTR [rsp+0xf0],r9
0x00007ffff7def0c7 <+71>: movaps XMMWORD PTR [rsp],xmm0
0x00007ffff7def0cb <+75>: movaps XMMWORD PTR [rsp+0x10],xmm1
0x00007ffff7def0d0 <+80>: movaps XMMWORD PTR [rsp+0x20],xmm2
0x00007ffff7def0d5 <+85>: movaps XMMWORD PTR [rsp+0x30],xmm3
0x00007ffff7def0da <+90>: movaps XMMWORD PTR [rsp+0x40],xmm4
0x00007ffff7def0df <+95>: movaps XMMWORD PTR [rsp+0x50],xmm5
0x00007ffff7def0e4 <+100>: movaps XMMWORD PTR [rsp+0x60],xmm6
0x00007ffff7def0e9 <+105>: movaps XMMWORD PTR [rsp+0x70],xmm7
0x00007ffff7def0ee <+110>: bndmov [rsp+0x80],bnd0
0x00007ffff7def0f7 <+119>: bndmov [rsp+0x90],bnd1
0x00007ffff7def100 <+128>: bndmov [rsp+0xa0],bnd2
0x00007ffff7def109 <+137>: bndmov [rsp+0xb0],bnd3
0x00007ffff7def112 <+146>: mov rsi,QWORD PTR [rbx+0x10]
0x00007ffff7def116 <+150>: mov rdi,QWORD PTR [rbx+0x8]
0x00007ffff7def11a <+154>: call 0x7ffff7de7970 <_dl_fixup>
0x00007ffff7def11f <+159>: mov r11,rax
0x00007ffff7def122 <+162>: bndmov bnd3,[rsp+0xb0]
0x00007ffff7def12b <+171>: bndmov bnd2,[rsp+0xa0]
0x00007ffff7def134 <+180>: bndmov bnd1,[rsp+0x90]
0x00007ffff7def13d <+189>: bndmov bnd0,[rsp+0x80]
0x00007ffff7def146 <+198>: mov r9,QWORD PTR [rsp+0xf0]
0x00007ffff7def14e <+206>: mov r8,QWORD PTR [rsp+0xe8]
0x00007ffff7def156 <+214>: mov rdi,QWORD PTR [rsp+0xe0]
0x00007ffff7def15e <+222>: mov rsi,QWORD PTR [rsp+0xd8]
0x00007ffff7def166 <+230>: mov rdx,QWORD PTR [rsp+0xd0]
0x00007ffff7def16e <+238>: mov rcx,QWORD PTR [rsp+0xc8]
0x00007ffff7def176 <+246>: mov rax,QWORD PTR [rsp+0xc0]
0x00007ffff7def17e <+254>: movaps xmm0,XMMWORD PTR [rsp]
0x00007ffff7def182 <+258>: movaps xmm1,XMMWORD PTR [rsp+0x10]
0x00007ffff7def187 <+263>: movaps xmm2,XMMWORD PTR [rsp+0x20]
0x00007ffff7def18c <+268>: movaps xmm3,XMMWORD PTR [rsp+0x30]
0x00007ffff7def191 <+273>: movaps xmm4,XMMWORD PTR [rsp+0x40]
0x00007ffff7def196 <+278>: movaps xmm5,XMMWORD PTR [rsp+0x50]
0x00007ffff7def19b <+283>: movaps xmm6,XMMWORD PTR [rsp+0x60]
0x00007ffff7def1a0 <+288>: movaps xmm7,XMMWORD PTR [rsp+0x70]
0x00007ffff7def1a5 <+293>: mov rsp,rbx
0x00007ffff7def1a8 <+296>: mov rbx,QWORD PTR [rsp]
0x00007ffff7def1ac <+300>: add rsp,0x18
0x00007ffff7def1b0 <+304>: bnd jmp r11
End of assembler dump.
that bnd jmp r11 at the end will take us to printf.
Lastly we step into _dl_fixup, where I singled out the call to _dl_lookup_symbol
(gdb) break *0x7ffff7def11a
Note: breakpoint 5 also set at pc 0x7ffff7def11a.
Breakpoint 8 at 0x7ffff7def11a: file ../sysdeps/x86_64/dl-trampoline.h, line 212.
(gdb) cont
Continuing.
Breakpoint 5, _dl_runtime_resolve_sse () at ../sysdeps/x86_64/dl-trampoline.h:212
212 in ../sysdeps/x86_64/dl-trampoline.h
(gdb) stepi
_dl_fixup (l=0x7ffff7ffe168, reloc_arg=0) at ../elf/dl-runtime.c:66
66 ../elf/dl-runtime.c: No such file or directory.
(gdb) disas
Dump of assembler code for function _dl_fixup:
=> 0x00007ffff7de7970 <+0>: push rbx
0x00007ffff7de7971 <+1>: mov r10,rdi
0x00007ffff7de7974 <+4>: mov esi,esi
0x00007ffff7de7976 <+6>: lea rdx,[rsi+rsi*2]
0x00007ffff7de797a <+10>: sub rsp,0x10
0x00007ffff7de797e <+14>: mov rax,QWORD PTR [rdi+0x68]
0x00007ffff7de7982 <+18>: mov rdi,QWORD PTR [rax+0x8]
0x00007ffff7de7986 <+22>: mov rax,QWORD PTR [r10+0xf8]
0x00007ffff7de798d <+29>: mov rax,QWORD PTR [rax+0x8]
0x00007ffff7de7991 <+33>: lea r8,[rax+rdx*8]
0x00007ffff7de7995 <+37>: mov rax,QWORD PTR [r10+0x70]
0x00007ffff7de7999 <+41>: mov rcx,QWORD PTR [r8+0x8]
0x00007ffff7de799d <+45>: mov rax,QWORD PTR [rax+0x8]
0x00007ffff7de79a1 <+49>: mov rdx,rcx
0x00007ffff7de79a4 <+52>: shr rdx,0x20
0x00007ffff7de79a8 <+56>: lea rsi,[rdx+rdx*2]
0x00007ffff7de79ac <+60>: lea rsi,[rax+rsi*8]
0x00007ffff7de79b0 <+64>: mov rax,QWORD PTR [r10]
0x00007ffff7de79b3 <+67>: mov QWORD PTR [rsp+0x8],rsi
0x00007ffff7de79b8 <+72>: mov rbx,rax
0x00007ffff7de79bb <+75>: add rbx,QWORD PTR [r8]
0x00007ffff7de79be <+78>: cmp ecx,0x7
0x00007ffff7de79c1 <+81>: jne 0x7ffff7de7b17 <_dl_fixup+423>
0x00007ffff7de79c7 <+87>: test BYTE PTR [rsi+0x5],0x3
0x00007ffff7de79cb <+91>: jne 0x7ffff7de7a67 <_dl_fixup+247>
0x00007ffff7de79d1 <+97>: mov rax,QWORD PTR [r10+0x1c8]
0x00007ffff7de79d8 <+104>: test rax,rax
0x00007ffff7de79db <+107>: je 0x7ffff7de7a90 <_dl_fixup+288>
0x00007ffff7de79e1 <+113>: mov rax,QWORD PTR [rax+0x8]
0x00007ffff7de79e5 <+117>: movzx eax,WORD PTR [rax+rdx*2]
0x00007ffff7de79e9 <+121>: and eax,0x7fff
0x00007ffff7de79ee <+126>: lea rdx,[rax+rax*2]
0x00007ffff7de79f2 <+130>: mov rax,QWORD PTR [r10+0x2e0]
0x00007ffff7de79f9 <+137>: lea r8,[rax+rdx*8]
0x00007ffff7de79fd <+141>: mov eax,0x0
0x00007ffff7de7a02 <+146>: mov r9d,DWORD PTR [r8+0x8]
0x00007ffff7de7a06 <+150>: test r9d,r9d
0x00007ffff7de7a09 <+153>: cmove r8,rax
0x00007ffff7de7a0d <+157>: mov edx,DWORD PTR fs:0x18
0x00007ffff7de7a15 <+165>: test edx,edx
0x00007ffff7de7a17 <+167>: mov eax,0x1
0x00007ffff7de7a1c <+172>: jne 0x7ffff7de7ab0 <_dl_fixup+320>
0x00007ffff7de7a22 <+178>: mov esi,DWORD PTR [rsi]
0x00007ffff7de7a24 <+180>: mov rcx,QWORD PTR [r10+0x380]
0x00007ffff7de7a2b <+187>: lea rdx,[rsp+0x8]
0x00007ffff7de7a30 <+192>: push 0x0
0x00007ffff7de7a32 <+194>: push rax
0x00007ffff7de7a33 <+195>: mov r9d,0x1
0x00007ffff7de7a39 <+201>: add rdi,rsi
0x00007ffff7de7a3c <+204>: mov rsi,r10
0x00007ffff7de7a3f <+207>: call 0x7ffff7de2d10 <_dl_lookup_symbol_x>
0x00007ffff7de7a44 <+212>: mov r8,rax
0x00007ffff7de7a47 <+215>: mov eax,DWORD PTR fs:0x18
0x00007ffff7de7a4f <+223>: test eax,eax
0x00007ffff7de7a51 <+225>: pop rcx
0x00007ffff7de7a52 <+226>: pop rsi
0x00007ffff7de7a53 <+227>: jne 0x7ffff7de7ad0 <_dl_fixup+352>
0x00007ffff7de7a55 <+229>: mov rsi,QWORD PTR [rsp+0x8]
0x00007ffff7de7a5a <+234>: test rsi,rsi
0x00007ffff7de7a5d <+237>: je 0x7ffff7de7aa0 <_dl_fixup+304>
0x00007ffff7de7a5f <+239>: test r8,r8
0x00007ffff7de7a62 <+242>: je 0x7ffff7de7aa8 <_dl_fixup+312>
0x00007ffff7de7a64 <+244>: mov rax,QWORD PTR [r8]
0x00007ffff7de7a67 <+247>: movzx edx,BYTE PTR [rsi+0x4]
0x00007ffff7de7a6b <+251>: add rax,QWORD PTR [rsi+0x8]
0x00007ffff7de7a6f <+255>: and edx,0xf
0x00007ffff7de7a72 <+258>: cmp dl,0xa
0x00007ffff7de7a75 <+261>: je 0x7ffff7de7b10 <_dl_fixup+416>
0x00007ffff7de7a7b <+267>: mov edx,DWORD PTR [rip+0x215267] # 0x7ffff7ffcce8 <_rtld_global_ro+72>
0x00007ffff7de7a81 <+273>: test edx,edx
0x00007ffff7de7a83 <+275>: jne 0x7ffff7de7a88 <_dl_fixup+280>
0x00007ffff7de7a85 <+277>: mov QWORD PTR [rbx],rax
0x00007ffff7de7a88 <+280>: add rsp,0x10
0x00007ffff7de7a8c <+284>: pop rbx
0x00007ffff7de7a8d <+285>: ret
0x00007ffff7de7a8e <+286>: xchg ax,ax
0x00007ffff7de7a90 <+288>: xor r8d,r8d
0x00007ffff7de7a93 <+291>: jmp 0x7ffff7de7a0d <_dl_fixup+157>
0x00007ffff7de7a98 <+296>: nop DWORD PTR [rax+rax*1+0x0]
0x00007ffff7de7aa0 <+304>: xor eax,eax
0x00007ffff7de7aa2 <+306>: jmp 0x7ffff7de7a7b <_dl_fixup+267>
0x00007ffff7de7aa4 <+308>: nop DWORD PTR [rax+0x0]
0x00007ffff7de7aa8 <+312>: xor eax,eax
0x00007ffff7de7aaa <+314>: jmp 0x7ffff7de7a67 <_dl_fixup+247>
0x00007ffff7de7aac <+316>: nop DWORD PTR [rax+0x0]
0x00007ffff7de7ab0 <+320>: mov DWORD PTR fs:0x1c,0x1
0x00007ffff7de7abc <+332>: mov eax,0x5
0x00007ffff7de7ac1 <+337>: jmp 0x7ffff7de7a22 <_dl_fixup+178>
0x00007ffff7de7ac6 <+342>: nop WORD PTR cs:[rax+rax*1+0x0]
0x00007ffff7de7ad0 <+352>: xor eax,eax
0x00007ffff7de7ad2 <+354>: xchg DWORD PTR fs:0x1c,eax
0x00007ffff7de7ada <+362>: cmp eax,0x2
0x00007ffff7de7add <+365>: jne 0x7ffff7de7a55 <_dl_fixup+229>
0x00007ffff7de7ae3 <+371>: mov rdi,QWORD PTR fs:0x10
0x00007ffff7de7aec <+380>: xor r10d,r10d
0x00007ffff7de7aef <+383>: add rdi,0x1c
0x00007ffff7de7af3 <+387>: mov edx,0x1
0x00007ffff7de7af8 <+392>: mov esi,0x81
0x00007ffff7de7afd <+397>: mov eax,0xca
0x00007ffff7de7b02 <+402>: syscall
0x00007ffff7de7b04 <+404>: jmp 0x7ffff7de7a55 <_dl_fixup+229>
0x00007ffff7de7b09 <+409>: nop DWORD PTR [rax+0x0]
0x00007ffff7de7b10 <+416>: call rax
0x00007ffff7de7b12 <+418>: jmp 0x7ffff7de7a7b <_dl_fixup+267>
0x00007ffff7de7b17 <+423>: lea rcx,[rip+0x10862] # 0x7ffff7df8380 <__PRETTY_FUNCTION__.10645>
0x00007ffff7de7b1e <+430>: lea rsi,[rip+0xea8f] # 0x7ffff7df65b4
0x00007ffff7de7b25 <+437>: lea rdi,[rip+0x1081c] # 0x7ffff7df8348
0x00007ffff7de7b2c <+444>: mov edx,0x4f
0x00007ffff7de7b31 <+449>: call 0x7ffff7df1210 <__GI___assert_fail>
End of assembler dump.
This is as far as this journey goes. the dl_lookup_symbol function goes to a land of parsing elf files and working with symbols, that's where the pfm happens.
{var%20f='http://v.t.sina.com.cn/share/share.php?appkey=1515056452',u=z||d.location,p=['&url=',e(u),'&title=',e(t||d.title),'&source=',e(r),'&sourceUrl=',e(l),'&content=',c||'gb2312','&pic=',e(p||'')].join('');function%20a(){if(!window.open([f,p].join(''),'mb',['toolbar=0,status=0,resizable=1,width=440,height=430,left=',(s.width-440)/2,',top=',(s.height-430)/2].join('')))u.href=[f,p].join('');};if(/Firefox/.test(navigator.userAgent))setTimeout(a,0);else%20a();})(screen,document,encodeURIComponent,'','','https://img.manongdao.com/sparkler-677774__340.jpg', '推荐 男人必须洒脱 的文章《What is _dl_lookup_symbol_x ? - C++ Profiling》','https://www.manongdao.com/article-952163.html','页面编码gb2312|utf-8默认gb2312'));){kind=link}