dlopen interception on libasan

[Christos Zoulas]

Well, the problem in PR/51670 is that with -fsanitizer=address the
dlopen interception is broken. libasan.so provides its own dlopen()
and then tries to find the real dlopen using dlsym(RTLD_NEXT,
"dlopen"). That does not work for NetBSD because dlopen in libc is
just a stub (that perhaps should abort), and the real dlopen is in
ld.elf_so itself. Unfortunately we can't get to it anymore easily
because RTLD_DEFAULT, RTLD_SELF, RTLD_NEXT, NULL all return the
wrong dlopen(). A simple fix is to provide RTLD_DL that returns
symbols defined by the dynamic linker. Here's a patch that does
this. Unless I hear otherwise I am planning to commit this soon.

christos

Index: include/dlfcn.h
===
RCS file: /cvsroot/src/include/dlfcn.h,v
retrieving revision 1.24
diff -u -u -r1.24 dlfcn.h
--- include/dlfcn.h 16 Feb 2012 23:00:39 -  1.24
+++ include/dlfcn.h 30 Nov 2016 01:28:01 -
@@ -80,6 +80,7 @@
 #defineRTLD_NEXT   ((void *) -1)   /* Search subsequent objects. */
 #defineRTLD_DEFAULT((void *) -2)   /* Use default search 
algorithm. */
 #defineRTLD_SELF   ((void *) -3)   /* Search the caller itself. */
+#defineRTLD_DL ((void *) -4)   /* Search only the dynamic 
linker */
 
 /*
  * dlctl() commands
Index: external/gpl3/gcc/dist/libsanitizer/interception/interception_linux.cc
===
RCS file: 
/cvsroot/src/external/gpl3/gcc/dist/libsanitizer/interception/interception_linux.cc,v
retrieving revision 1.4
diff -u -u -r1.4 interception_linux.cc
--- external/gpl3/gcc/dist/libsanitizer/interception/interception_linux.cc  
24 Jan 2016 09:43:38 -  1.4
+++ external/gpl3/gcc/dist/libsanitizer/interception/interception_linux.cc  
30 Nov 2016 01:28:01 -
@@ -28,12 +28,16 @@
 namespace __interception {
 bool GetRealFunctionAddress(const char *func_name, uptr *func_addr,
 uptr real, uptr wrapper) {
+  void *w = RTLD_NEXT;
 #ifdef __NetBSD__
   // XXX: Until I come up with something better to deal with renames.
   if (mystrcmp(func_name, "sigaction") == 0)
 func_name = "__sigaction14";
+  if (mystrcmp(func_name, "dlopen") == 0 || mystrcmp(func_name, "dlclose") == 
0)
+w = RTLD_DL;
+
 #endif
-  *func_addr = (uptr)dlsym(RTLD_NEXT, func_name);
+  *func_addr = (uptr)dlsym(w, func_name);
   return real == wrapper;
 }
 
Index: libexec/ld.elf_so/rtld.c
===
RCS file: /cvsroot/src/libexec/ld.elf_so/rtld.c,v
retrieving revision 1.178
diff -u -u -r1.178 rtld.c
--- libexec/ld.elf_so/rtld.c24 May 2016 20:32:33 -  1.178
+++ libexec/ld.elf_so/rtld.c30 Nov 2016 01:28:02 -
@@ -1106,6 +1106,9 @@
defobj = NULL;
 
switch ((intptr_t)handle) {
+   case (intptr_t)RTLD_DL:
+   obj = &_rtld_objself;
+   goto gotobj;
case (intptr_t)NULL:
case (intptr_t)RTLD_NEXT:
case (intptr_t)RTLD_DEFAULT:
@@ -1115,8 +1118,9 @@
lookup_mutex_exit();
return NULL;
}
-
+gotobj:
switch ((intptr_t)handle) {
+   case (intptr_t)RTLD_DL:
case (intptr_t)NULL: /* Just the caller's shared object. */
def = _rtld_symlook_obj(name, hash, obj, flags, ventry);
defobj = obj;

Re: curses: Home key sends 0x162 in tmux

[Brett Lymn]

On Tue, Nov 29, 2016 at 05:48:22PM +, Christos Zoulas wrote:
> In article 
> <1877343761.1128907.1480410438073.javamail.ngm...@webmail19.arcor-online.net>,
> Carsten Kunze   wrote:
> >Hello,
> >
> >the Home key sends 0x162 instead of 0x106 in tmux.  A small test program
> >to be compiled with
> 
> Hmm, it works for me (gives 0x106). Also I could not reproduce your other
> ASAN/vs/no-ASAN issue. Is that current on amd64? What terminal?
> 

What does the output of infocmp look like?  Curses pulls the definitions
of the keys from the terminfo entry.  If that has mapped HOME to mean
BEGIN (which doesn't sound unreasonable) then that is what curses will
do.

-- 
Brett Lymn
Let go, or be dragged - Zen proverb.

curses: Home key sends 0x162 in tmux

[Carsten Kunze]

Hello,

the Home key sends 0x162 instead of 0x106 in tmux.  A small test program to be 
compiled with

cc main.c -lcurses

and quit with CTRL-c shows the hex key codes:

$ cat main.c

   
#include 

int
main()
{
initscr();
noecho();
cbreak();
keypad(stdscr, TRUE);

while (1) {
printw("0x%x\n", getch());
refresh();
}

return 0;
}