Bug #2663
closedStrange behaviour when loading a library twice.
0%
Description
Hello. This is an output of uname:
$ uname -a
DragonFly cyberspace.cyberspace 3.6-RELEASE DragonFly
v3.6.2.2.g063e0-RELEASE #7: Sun Apr 13 09:50:36 MSK 2014
vasily@cyberspace.cyberspace:/usr/obj/usr/src/sys/MYKERNEL
x86_64
Trying to port common lisp IOLib library on DragonFly, I have encountered a
strange behaviour (of RTLD as I think). Assume you have a library which
mmap's a fixed (with flag MAP_TRYFIXED) region of memory in its
constructor. When your program opens it with dlopen, then closes it with
dlcose and again opens with dlopen, the second call to mmap from the
constructor will fail. Is it OK?
I attach a testing program (test.c). Try to compile it and run with
libthread_xu.so:
$ cc -o test test.c
$ ./test /usr/lib/thread/libthread_xu.so.
You will get the following output:
testing /usr/lib/thread/libthread_xu.so
Fatal error 'Cannot allocate red zone for initial thread' at line 275 in
file /usr/src/lib/libthread_xu/thread/thr_in
it.c (errno =
12)
Abort (core dumped)
You can try any other library which is linked against -lpthread or you can
try my own library testlib.c
Compile it as follows:
$ cc -shared -fPIC -o testlib.so testlib.c
And run the test:
$ ./test ./testlib.so
You will get:
testing ./libtest.so
0
mmap OK
0
Failed to allocate some space with initialized == 0
In FreeBSD, for example, mmap succeeds two times.
Files
Updated by shamaz over 10 years ago
UPD: Please ignore my own library, as MAP_FIXED flag in FreeBSD just overwrites existing mapping. Try any library which was linked against pthreads.
For some reason the constructor of libthread_xu.so will be called twice which will give you an error:
Fatal error 'Cannot allocate red zone for initial thread' at line 275 in file /usr/src/lib/libthread_xu/thread/thr_init.c (errno = 12)
Updated by shamaz over 10 years ago
- Status changed from New to Closed
Fix committed: 0dc5e56d98ab5fc55416a204886f8b132cc9b7f1