Re: dlopen interception on libasan

[Nick Hudson <skrll%netbsd.org@localhost>]

On 11/30/16 15:08, Christos Zoulas wrote:
> In article <583EE75A.8050205%netbsd.org@localhost>,
> Nick Hudson  <skrll%netbsd.org@localhost> wrote:
> > On 11/30/16 14:37, Christos Zoulas wrote:
> > > On Nov 30,  7:34am, skrll%netbsd.org@localhost (Nick Hudson) wrote:
> > [incorrect test deleted]
> >
> > > You did not define dlopen() in your own shared object; libasan needs to
> > > do that to override the default definition. This is not the right test.
> > OK, so... add something like
> >
> > https://nxr.netbsd.org/xref/src/libexec/ld.elf_so/symbol.c#565
> >
> > to
> >
> >    https://nxr.netbsd.org/xref/src/libexec/ld.elf_so/rtld.c#1125
> You mean changing the behavior of RTLD_NEXT, to look at the dynamic linker
> first? I think I like your idea better since it hides the dlopen from ld.so
> implementation detail.


More this.. (not compile tested)

i.e. should the dynamic linker  be searched last and only if we found a 
weak symbol

Nick

Index: rtld.c
===================================================================
RCS file: /cvsroot/src/libexec/ld.elf_so/rtld.c,v
retrieving revision 1.178
diff -u -p -r1.178 rtld.c
--- rtld.c	24 May 2016 20:32:33 -0000	1.178
+++ rtld.c	30 Nov 2016 15:28:47 -0000
@@ -1134,6 +1134,22 @@ do_dlsym(void *handle, const char *name,
 					break;
 				}
 			}
+			/*
+			 * Search the dynamic linker itself, and possibly resolve the
+			 * symbol from there.  This is how the application links to
+			 * dynamic linker services such as dlopen.  Only the values listed
+			 * in the "_rtld_exports" array can be resolved from the dynamic
+			 * linker.
+			 */
+			if (def == NULL || ELF_ST_BIND(def->st_info) == STB_WEAK) {
+				rdbg(("Search the dynamic linker itself."));
+				symp = _rtld_symlook_obj(name, hash, &_rtld_objself, flags,
+				    ventry);
+				if (symp != NULL && _rtld_is_exported(symp)) {
+					def = symp;
+					defobj = &_rtld_objself;
+				}
+			}
 			break;
 
 		case (intptr_t)RTLD_DEFAULT: