Comments (14)
namhyung
commented on August 29, 2024
Is the test program (t-abc) is a same binary for both case (or is it also rebuilt with GCC7)? I need to check GCC changes about PLT handling..
from uftrace.
namhyung
commented on August 29, 2024
If the binaries are different (IOW rebuit), could you please chek there's any difference in the disassembly output?
from uftrace.
Taeung
commented on August 29, 2024
@namhyung I think there aren't big difference about PLT handling..
But if checking disassembly code of test program (t-abc) built by gcc 5,
Disassembly of section .plt:
0000000000400510 <getpid@plt-0x10>:
400510: ff 35 f2 0a 20 00 pushq 0x200af2(%rip) # 601008 <_GLOBAL_OFFSET_TABLE_+0x8>
400516: ff 25 f4 0a 20 00 jmpq *0x200af4(%rip) # 601010 <_GLOBAL_OFFSET_TABLE_+0x10>
40051c: 0f 1f 40 00 nopl 0x0(%rax)
0000000000400520 <getpid@plt>:
400520: ff 25 f2 0a 20 00 jmpq *0x200af2(%rip) # 601018 <_GLOBAL_OFFSET_TABLE_+0x18>
400526: 68 00 00 00 00 pushq $0x0
40052b: e9 e0 ff ff ff jmpq 400510 <_init+0x20>
0000000000400530 <__libc_start_main@plt>:
400530: ff 25 ea 0a 20 00 jmpq *0x200aea(%rip) # 601020 <_GLOBAL_OFFSET_TABLE_+0x20>
400536: 68 01 00 00 00 pushq $0x1
40053b: e9 d0 ff ff ff jmpq 400510 <_init+0x20>
0000000000400540 <__monstartup@plt>:
400540: ff 25 e2 0a 20 00 jmpq *0x200ae2(%rip) # 601028 <_GLOBAL_OFFSET_TABLE_+0x28>
400546: 68 02 00 00 00 pushq $0x2
40054b: e9 c0 ff ff ff jmpq 400510 <_init+0x20>
0000000000400550 <mcount@plt>:
400550: ff 25 da 0a 20 00 jmpq *0x200ada(%rip) # 601030 <_GLOBAL_OFFSET_TABLE_+0x30>
400556: 68 03 00 00 00 pushq $0x3
40055b: e9 b0 ff ff ff jmpq 400510 <_init+0x20>
0000000000400560 <__cxa_atexit@plt>:
400560: ff 25 d2 0a 20 00 jmpq *0x200ad2(%rip) # 601038 <_GLOBAL_OFFSET_TABLE_+0x38>
400566: 68 04 00 00 00 pushq $0x4
40056b: e9 a0 ff ff ff jmpq 400510 <_init+0x20>
0000000000400570 <atoi@plt>:
400570: ff 25 ca 0a 20 00 jmpq *0x200aca(%rip) # 601040 <_GLOBAL_OFFSET_TABLE_+0x40>
400576: 68 05 00 00 00 pushq $0x5
40057b: e9 90 ff ff ff jmpq 400510 <_init+0x20>
...
00000000004006c6 <a>:
4006c6: 55 push %rbp
4006c7: 48 89 e5 mov %rsp,%rbp
4006ca: e8 81 fe ff ff callq 400550 <mcount@plt>
4006cf: e8 05 00 00 00 callq 4006d9 <b>
4006d4: 83 e8 01 sub $0x1,%eax
4006d7: 5d pop %rbp
4006d8: c3 retq
00000000004006d9 <b>:
4006d9: 55 push %rbp
4006da: 48 89 e5 mov %rsp,%rbp
4006dd: e8 6e fe ff ff callq 400550 <mcount@plt>
4006e2: e8 05 00 00 00 callq 4006ec <c>
4006e7: 83 c0 01 add $0x1,%eax
4006ea: 5d pop %rbp
4006eb: c3 retq
00000000004006ec <c>:
4006ec: 55 push %rbp
4006ed: 48 89 e5 mov %rsp,%rbp
4006f0: e8 5b fe ff ff callq 400550 <mcount@plt>
4006f5: e8 26 fe ff ff callq 400520 <getpid@plt>
4006fa: 89 c1 mov %eax,%ecx
4006fc: ba 89 b5 f8 14 mov $0x14f8b589,%edx
400701: 89 c8 mov %ecx,%eax
400703: f7 ea imul %edx
400705: c1 fa 0d sar $0xd,%edx
400708: 89 c8 mov %ecx,%eax
40070a: c1 f8 1f sar $0x1f,%eax
40070d: 29 c2 sub %eax,%edx
40070f: 89 d0 mov %edx,%eax
400711: 69 c0 a0 86 01 00 imul $0x186a0,%eax,%eax
400717: 29 c1 sub %eax,%ecx
400719: 89 c8 mov %ecx,%eax
40071b: 5d pop %rbp
40071c: c3 retq
000000000040071d <main>:
40071d: 55 push %rbp
40071e: 48 89 e5 mov %rsp,%rbp
400721: 48 83 ec 20 sub $0x20,%rsp
400725: e8 26 fe ff ff callq 400550 <mcount@plt>
40072a: 89 7d ec mov %edi,-0x14(%rbp)
40072d: 48 89 75 e0 mov %rsi,-0x20(%rbp)
400731: c7 45 fc 00 00 00 00 movl $0x0,-0x4(%rbp)
400738: 83 7d ec 01 cmpl $0x1,-0x14(%rbp)
40073c: 7e 16 jle 400754 <main+0x37>
40073e: 48 8b 45 e0 mov -0x20(%rbp),%rax
400742: 48 83 c0 08 add $0x8,%rax
400746: 48 8b 00 mov (%rax),%rax
400749: 48 89 c7 mov %rax,%rdi
40074c: e8 1f fe ff ff callq 400570 <atoi@plt>
400751: 89 45 fc mov %eax,-0x4(%rbp)
400754: e8 6d ff ff ff callq 4006c6 <a>
400759: 01 45 fc add %eax,-0x4(%rbp)
40075c: 83 7d fc 00 cmpl $0x0,-0x4(%rbp)
400760: 0f 94 c0 sete %al
400763: 0f b6 c0 movzbl %al,%eax
400766: c9 leaveq
400767: c3 retq
400768: 0f 1f 84 00 00 00 00 nopl 0x0(%rax,%rax,1)
40076f: 00
If checking the case of gcc 7,
Disassembly of section .plt:
00000000004004d0 <getpid@plt-0x10>:
4004d0: ff 35 d2 06 20 00 pushq 0x2006d2(%rip) # 600ba8 <_GLOBAL_OFFSET_TABLE_+0x8>
4004d6: ff 25 d4 06 20 00 jmpq *0x2006d4(%rip) # 600bb0 <_GLOBAL_OFFSET_TABLE_+0x10>
4004dc: 0f 1f 40 00 nopl 0x0(%rax)
00000000004004e0 <getpid@plt>:
4004e0: ff 25 d2 06 20 00 jmpq *0x2006d2(%rip) # 600bb8 <_GLOBAL_OFFSET_TABLE_+0x18>
4004e6: 68 00 00 00 00 pushq $0x0
4004eb: e9 e0 ff ff ff jmpq 4004d0 <_init+0x28>
00000000004004f0 <__libc_start_main@plt>:
4004f0: ff 25 ca 06 20 00 jmpq *0x2006ca(%rip) # 600bc0 <_GLOBAL_OFFSET_TABLE_+0x20>
4004f6: 68 01 00 00 00 pushq $0x1
4004fb: e9 d0 ff ff ff jmpq 4004d0 <_init+0x28>
0000000000400500 <__monstartup@plt>:
400500: ff 25 c2 06 20 00 jmpq *0x2006c2(%rip) # 600bc8 <_GLOBAL_OFFSET_TABLE_+0x28>
400506: 68 02 00 00 00 pushq $0x2
40050b: e9 c0 ff ff ff jmpq 4004d0 <_init+0x28>
0000000000400510 <mcount@plt>:
400510: ff 25 ba 06 20 00 jmpq *0x2006ba(%rip) # 600bd0 <_GLOBAL_OFFSET_TABLE_+0x30>
400516: 68 03 00 00 00 pushq $0x3
40051b: e9 b0 ff ff ff jmpq 4004d0 <_init+0x28>
0000000000400520 <__cxa_atexit@plt>:
400520: ff 25 b2 06 20 00 jmpq *0x2006b2(%rip) # 600bd8 <_GLOBAL_OFFSET_TABLE_+0x38>
400526: 68 04 00 00 00 pushq $0x4
40052b: e9 a0 ff ff ff jmpq 4004d0 <_init+0x28>
0000000000400530 <atoi@plt>:
400530: ff 25 aa 06 20 00 jmpq *0x2006aa(%rip) # 600be0 <_GLOBAL_OFFSET_TABLE_+0x40>
400536: 68 05 00 00 00 pushq $0x5
40053b: e9 90 ff ff ff jmpq 4004d0 <_init+0x28>
...
0000000000400667 <a>:
400667: 55 push %rbp
400668: 48 89 e5 mov %rsp,%rbp
40066b: e8 a0 fe ff ff callq 400510 <mcount@plt>
400670: e8 05 00 00 00 callq 40067a <b>
400675: 83 e8 01 sub $0x1,%eax
400678: 5d pop %rbp
400679: c3 retq
000000000040067a <b>:
40067a: 55 push %rbp
40067b: 48 89 e5 mov %rsp,%rbp
40067e: e8 8d fe ff ff callq 400510 <mcount@plt>
400683: e8 05 00 00 00 callq 40068d <c>
400688: 83 c0 01 add $0x1,%eax
40068b: 5d pop %rbp
40068c: c3 retq
000000000040068d <c>:
40068d: 55 push %rbp
40068e: 48 89 e5 mov %rsp,%rbp
400691: e8 7a fe ff ff callq 400510 <mcount@plt>
400696: e8 45 fe ff ff callq 4004e0 <getpid@plt>
40069b: 89 c1 mov %eax,%ecx
40069d: ba 89 b5 f8 14 mov $0x14f8b589,%edx
4006a2: 89 c8 mov %ecx,%eax
4006a4: f7 ea imul %edx
4006a6: c1 fa 0d sar $0xd,%edx
4006a9: 89 c8 mov %ecx,%eax
4006ab: c1 f8 1f sar $0x1f,%eax
4006ae: 29 c2 sub %eax,%edx
4006b0: 89 d0 mov %edx,%eax
4006b2: 69 c0 a0 86 01 00 imul $0x186a0,%eax,%eax
4006b8: 29 c1 sub %eax,%ecx
4006ba: 89 c8 mov %ecx,%eax
4006bc: 5d pop %rbp
4006bd: c3 retq
00000000004006be <main>:
4006be: 55 push %rbp
4006bf: 48 89 e5 mov %rsp,%rbp
4006c2: 48 83 ec 20 sub $0x20,%rsp
4006c6: e8 45 fe ff ff callq 400510 <mcount@plt>
4006cb: 89 7d ec mov %edi,-0x14(%rbp)
4006ce: 48 89 75 e0 mov %rsi,-0x20(%rbp)
4006d2: c7 45 fc 00 00 00 00 movl $0x0,-0x4(%rbp)
4006d9: 83 7d ec 01 cmpl $0x1,-0x14(%rbp)
4006dd: 7e 16 jle 4006f5 <main+0x37>
4006df: 48 8b 45 e0 mov -0x20(%rbp),%rax
4006e3: 48 83 c0 08 add $0x8,%rax
4006e7: 48 8b 00 mov (%rax),%rax
4006ea: 48 89 c7 mov %rax,%rdi
4006ed: e8 3e fe ff ff callq 400530 <atoi@plt>
4006f2: 89 45 fc mov %eax,-0x4(%rbp)
4006f5: e8 6d ff ff ff callq 400667 <a>
4006fa: 01 45 fc add %eax,-0x4(%rbp)
4006fd: 83 7d fc 00 cmpl $0x0,-0x4(%rbp)
400701: 0f 94 c0 sete %al
400704: 0f b6 c0 movzbl %al,%eax
400707: c9 leaveq
400708: c3 retq
400709: 0f 1f 80 00 00 00 00 nopl 0x0(%rax)
from uftrace.
Taeung
commented on August 29, 2024
@namhyung And let me show all cases about gcc 5 and gcc 7.
Case A) with uftrace built by gcc 5
- A-1) test program(t-abc) built by
gcc 5
$ gcc -dumpversion
5.4.0
$ ./runtest.py basic
Test case pg finstrument-fu
------------------------: O0 O1 O2 O3 Os O0 O1 O2 O3 Os
001 basic : OK OK OK OK OK OK OK OK OK OK
...
- A-2) test program(t-abc) built by
gcc 7
$ gcc -dumpversion
7.0.1
$ ./runtest.py basic
Test case pg finstrument-fu
------------------------: O0 O1 O2 O3 Os O0 O1 O2 O3 Os
001 basic : OK OK OK OK OK OK OK OK OK OK
...
Case B) with uftrace built by gcc 7
- B-1) test program(t-abc) built by
gcc 5
$ gcc -dumpversion
5.4.0
$ ./runtest.py basic
Test case pg finstrument-fu
------------------------: O0 O1 O2 O3 Os O0 O1 O2 O3 Os
001 basic : SG SG SG SG SG SG SG SG SG SG
child terminated by signal: 11: Segmentation fault
child terminated by signal: 11: Segmentation fault
child terminated by signal: 11: Segmentation fault
child terminated by signal: 11: Segmentation fault
child terminated by signal: 11: Segmentation fault
child terminated by signal: 11: Segmentation fault
child terminated by signal: 11: Segmentation fault
child terminated by signal: 11: Segmentation fault
child terminated by signal: 11: Segmentation fault
child terminated by signal: 11: Segmentation fault
...
- B-2) test program(t-abc) built by
gcc 7
$ gcc -dumpversion
7.0.1
$ ./runtest.py basic
Test case pg finstrument-fu
------------------------: O0 O1 O2 O3 Os O0 O1 O2 O3 Os
001 basic : SG SG SG SG SG SG SG SG SG SG
child terminated by signal: 11: Segmentation fault
child terminated by signal: 11: Segmentation fault
child terminated by signal: 11: Segmentation fault
child terminated by signal: 11: Segmentation fault
child terminated by signal: 11: Segmentation fault
child terminated by signal: 11: Segmentation fault
child terminated by signal: 11: Segmentation fault
child terminated by signal: 11: Segmentation fault
child terminated by signal: 11: Segmentation fault
child terminated by signal: 11: Segmentation fault
...
from uftrace.
namhyung
commented on August 29, 2024
OK. Many thanks. Could you provide a link for libmcount-fast.so built with GCC7 so that I can take a look at the binary?
from uftrace.
Taeung
commented on August 29, 2024
@namhyung Okey, you want to download libmcount-fast.so built by gcc 7. Right ?
from uftrace.
namhyung
commented on August 29, 2024
Yes, appreciated!
from uftrace.
Taeung
commented on August 29, 2024
Hum.. My build way is wrong ?..
$ readelf -p .comment /usr/local/lib/libmcount-fast.so
String dump of section '.comment':
[ 0] GCC: (GNU) 7.0.1 20170405 (experimental)
[ 29] GCC: (Ubuntu 5.4.0-6ubuntu1~16.04.4) 5.4.0 20160609
from uftrace.
namhyung
commented on August 29, 2024
Not sure.. maybe it contains some object files built with GCC5. But I can reproduce the segfault so it's ok for me.
from uftrace.
namhyung
commented on August 29, 2024
And the problem might be due to stack alignment required by "movaps" instruction. Could you please test below patch?
diff --git a/arch/x86_64/plthook.S b/arch/x86_64/plthook.S
index baa6b52..fe0ff44 100644
--- a/arch/x86_64/plthook.S
+++ b/arch/x86_64/plthook.S
@@ -3,8 +3,10 @@ plt_hooker:
.cfi_startproc
/* PLT code already pushed symbol and module indices */
.cfi_adjust_cfa_offset 16
- sub $48, %rsp
- .cfi_adjust_cfa_offset 48
+ sub $56, %rsp
+ .cfi_adjust_cfa_offset 56
+ movq %rax, 48(%rsp)
+ .cfi_offset rdi, -32
movq %rdi, 40(%rsp)
.cfi_offset rdi, -32
movq %rsi, 32(%rsp)
@@ -19,11 +21,11 @@ plt_hooker:
.cfi_offset r9, -72
/* child idx */
- movq 56(%rsp), %rsi
+ movq 64(%rsp), %rsi
/* address of parent ip */
- lea 64(%rsp), %rdi
+ lea 72(%rsp), %rdi
/* module id */
- movq 48(%rsp), %rdx
+ movq 56(%rsp), %rdx
/* mcount_args */
movq %rsp, %rcx
@@ -35,8 +37,9 @@ plt_hooker:
movq 24(%rsp), %rdx
movq 32(%rsp), %rsi
movq 40(%rsp), %rdi
- add $48, %rsp
- .cfi_adjust_cfa_offset -48
+ /* ignore rax */
+ add $56, %rsp
+ .cfi_adjust_cfa_offset -56
cmpq $0, %rax
cmovz plthook_resolver_addr(%rip), %rax
from uftrace.
Taeung
commented on August 29, 2024
Sure, after I tested this, will tell you the result ! 😄
from uftrace.
Taeung
commented on August 29, 2024
@namhyung The segfault is gone ! with your patch.
But I don't understand exactly what +8 and alignment for movaps instruction means..
I think I seem to need the domain knowledge and Google.. 😄
from uftrace.
namhyung
commented on August 29, 2024
FYI
http://x86.renejeschke.de/html/file_module_x86_id_180.html
(gdb) disas
Dump of assembler code for function plthook_entry:
0x00007f09d8db5300 <+0>: push %r15
0x00007f09d8db5302 <+2>: push %r14
0x00007f09d8db5304 <+4>: push %r13
0x00007f09d8db5306 <+6>: push %r12
0x00007f09d8db5308 <+8>: push %rbp
0x00007f09d8db5309 <+9>: push %rbx
0x00007f09d8db530a <+10>: pxor %xmm0,%xmm0
0x00007f09d8db530e <+14>: sub $0x58,%rsp
0x00007f09d8db5312 <+18>: lea 0x2169c8(%rip),%rax # 0x7f09d8fcbce1 <mcount_setup_done>
0x00007f09d8db5319 <+25>: movq $0x0,0x40(%rsp)
=> 0x00007f09d8db5322 <+34>: movaps %xmm0,0x20(%rsp)
(gdb) p $rsp
$1 = (void *) 0x7fffa19971f8
from uftrace.
Taeung
commented on August 29, 2024
Thank you !
from uftrace.
Related Issues (20)
- test: Run test suite on multiple environments HOT 1
- kernel tests failures HOT 1
- cmds: static `tmp_dirname` used stack after return HOT 1
- Give monotonic time offset from real time. HOT 4
- Various runtime issues running numerical simulation HOT 26
- Fix close file descriptor in setup_dwarf_info function to release acquired resource
- Generate uftrace.data in rotation HOT 1
- Add new `hist` command HOT 2
- how to trace a dynamic lib HOT 7
- Build failures with LTO HOT 2
- Segmentation fault with uftrace dump --flame-graph HOT 8
- Typo in README: Redundant 'the' in documentation
- Improve Documentation about Flame Graph (with --sample-time examples) HOT 1
- Compile warning on `‘sizeof’ in the earlier argument and not in the later argument` HOT 2
- Compile warning caused by Python3.12 header from `-Wdeclaration-after-statement` HOT 2
- what should i do when i want to trace a member function? HOT 1
- demangling failure with __device_stub_ prefix HOT 5
- "./misc/demangler --unknown" triggers segfault
- How to get uftrace data before the process terminates? HOT 3
- Segmentation fault HOT 8
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from uftrace.