Calling a function from a SO breaks in libc

classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view
|

Calling a function from a SO breaks in libc

Gleb Popov-2
When debugging software with GDB on FreeBSD 12+ I often get into following
situation:

(gdb) next
(anonymous namespace)::D13AsmPrinter::EmitInstruction (this=0x8056b4b80,
MI=0x805765690) at
/home/arr/projects/d13/lib/Target/D13/D13AsmPrinter.cpp:152
152           EmitToStreamer(*OutStreamer, TmpInst);
(gdb) step

Now instead of entering into EmitStreamer(), I see:

_thr_rtld_set_flag (mask=1) at /usr/src/lib/libthr/thread/thr_rtld.c:171
171     {

Trying to get out of there:

(gdb) finish
Run till exit from #0  _thr_rtld_set_flag (mask=1) at
/usr/src/lib/libthr/thread/thr_rtld.c:171
0x000000080027669b in thread_mask_set (mask=<optimized out>) at
/usr/src/libexec/rtld-elf/rtld_lock.c:177
177             return lockinfo.thread_set_flag(mask);
Value returned is $3 = 0
(gdb) finish
Run till exit from #0  0x000000080027669b in thread_mask_set
(mask=<optimized out>) at /usr/src/libexec/rtld-elf/rtld_lock.c:177
rlock_acquire (lock=0x800287ba0 <rtld_locks>, lockstate=0x7fffffff9608) at
/usr/src/libexec/rtld-elf/rtld_lock.c:203
203             if (thread_mask_set(lock->mask) & lock->mask) {
(gdb) finish
Run till exit from #0  rlock_acquire (lock=0x800287ba0 <rtld_locks>,
lockstate=0x7fffffff9608) at /usr/src/libexec/rtld-elf/rtld_lock.c:203
_rtld_bind (obj=0x80028b400, reloff=625152) at
/usr/src/libexec/rtld-elf/rtld.c:808
808         if (sigsetjmp(lockstate.env, 0) != 0)
(gdb) finish
Run till exit from #0  _rtld_bind (obj=0x80028b400, reloff=625152) at
/usr/src/libexec/rtld-elf/rtld.c:808
_rtld_bind_start () at /usr/src/libexec/rtld-elf/amd64/rtld_start.S:124
124             movq    %rax,0x60(%rsp)         # Store target over reloff
argument
Value returned is $4 = 34402603952
(gdb) finish
Run till exit from #0  _rtld_bind_start () at
/usr/src/libexec/rtld-elf/amd64/rtld_start.S:124

Program received signal SIGTRAP, Trace/breakpoint trap.
_rtld_bind_start () at /usr/src/libexec/rtld-elf/amd64/rtld_start.S:155
155             ret                             # "Return" to target address
(gdb) finish
Run till exit from #0  _rtld_bind_start () at
/usr/src/libexec/rtld-elf/amd64/rtld_start.S:155

Program received signal SIGTRAP, Trace/breakpoint trap.
0x00000008028e13b4 in llvm::AsmPrinter::EmitToStreamer (this=0x80028b400,
S=..., Inst=...) at
/home/arr/projects/d13/lib/CodeGen/AsmPrinter/AsmPrinter.cpp:228

After this. the debugging is impossible, because every "cont/step/next"
command yields "Program received signal SIGTRAP, Trace/breakpoint trap".
Trying to "print" anything causes GDB to crash.

It should be noted, that I'm linking to LLVM-7.so library. Using static
linkage for the executable I'm debugging makes this bug go away.

Anyone have an idea what's going on and how to fix that?
_______________________________________________
[hidden email] mailing list
https://lists.freebsd.org/mailman/listinfo/freebsd-hackers
To unsubscribe, send any mail to "[hidden email]"