This is the mail archive of the gcc-bugs@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

[Bug target/84521] [8 Regression] aarch64: Frame-pointer corruption with setjmp/longjmp and -fomit-frame-pointer


https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84521

--- Comment #1 from David Malcolm <dmalcolm at gcc dot gnu.org> ---
Created attachment 43489
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=43489&action=edit
Reproducer

When compiled with:
  gcc -DDUMP -g -O0 -fstack-protector-strong -Wall test.c
this runs to completion, and the x29 values show the function calls/returns:
x29 = 0x7ff2977910 : main : start of main
x29 = 0x7ff29778d0 : test_2 : start of test_2
x29 = 0x7ff29776a0 : test_1 : start of test_1
x29 = 0x7ff29776a0 : test_1 : zero return
x29 = 0x7ff2977690 : uses_longjmp : in uses_longjmp
x29 = 0x7ff2977740 : test_1 : non-zero return
x29 = 0x7ff2977690 : after_longjmp : after raise
x29 = 0x7ff2977740 : test_1 : end of test_1
x29 = 0x7ff29778d0 : test_2 : end of test_2
x29 = 0x7ff2977910 : main : end of main

On adding -fomit-frame-pointer, it crashes, and the x29 values show a
corruption after "uses_longjmp" which becomes a crash when the x29 value is
later used:

x29 = 0x7fff4709c0 : main : start of main
x29 = 0x7fff470960 : test_2 : start of test_2
x29 = 0x7fff470960 : test_1 : start of test_1
x29 = 0x7fff470960 : test_1 : zero return
x29 = 0x7fff470720 : uses_longjmp : in uses_longjmp
x29 = 0x7fff4707d0 : test_1 : non-zero return
x29 = 0x7fff4707d0 : after_longjmp : after raise
x29 = 0x7fff4707d0 : test_1 : end of test_1
x29 = 0x7fff4707d0 : test_2 : end of test_2
*** stack smashing detected ***: ./test-O0-omit-fp terminated

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]