From: Jo-Philipp Wich <jow@openwrt.org>
Date: Fri, 5 Nov 2010 21:38:38 +0000 (+0000)
Subject: ld.so: ldd crashes when __LDSO_SEARCH_INTERP_PATH__ is not #defined Since b65c7b2c79d... 
X-Git-Url: http://git.lede-project.org./?a=commitdiff_plain;h=12b122f31fda49cca1659805570dd8b90cf8240c;p=openwrt%2Fstaging%2Frobimarko.git

ld.so: ldd crashes when __LDSO_SEARCH_INTERP_PATH__ is not #defined Since b65c7b2c79debcb9017e31913e01eeaa280106fb, the implicit search path can be disabled by not #defining __LDSO_SEARCH_INTERP_PATH__. This causes _dl_ldsopath to never be set, so it remains NULL. _dl_ldsopath is still used when __LDSO_LDD_SUPPORT__ is #defined, to strip the path off of the beginning of the absolute path to the ld.so interpreter in use for printing. The _dl_strlen will crash with a NULL argument.

Rather than relying on _dl_ldsopath, this change causes ldd to compute
the interpreter's basename directly.

glibc ld.so seems to print the full path to the interpreter without
any computed basename or =>. I personally prefer glibc's behavior, but
to preserve backwards compatibility with uClibc ld.so, the existing
format with the computed basename, =>, and full path is used here. This
enables simpler (and unchanged) text processing in a pipeline.

Signed-off-by: Mark Mentovai <mark at moxienet.com>

SVN-Revision: 23892
---

diff --git a/toolchain/uClibc/patches-0.9.31/130-ldso-fix-__dl_parse_dynamic_info-segfault.patch b/toolchain/uClibc/patches-0.9.31/130-ldso-fix-__dl_parse_dynamic_info-segfault.patch
new file mode 100644
index 0000000000..f853d2ed80
--- /dev/null
+++ b/toolchain/uClibc/patches-0.9.31/130-ldso-fix-__dl_parse_dynamic_info-segfault.patch
@@ -0,0 +1,43 @@
+[PATCH] ld.so: ldd crashes when __LDSO_SEARCH_INTERP_PATH__ is not #defined
+Since b65c7b2c79debcb9017e31913e01eeaa280106fb, the implicit search path
+can be disabled by not #defining __LDSO_SEARCH_INTERP_PATH__. This
+causes _dl_ldsopath to never be set, so it remains NULL. _dl_ldsopath is
+still used when __LDSO_LDD_SUPPORT__ is #defined, to strip the path off
+of the beginning of the absolute path to the ld.so interpreter in use
+for printing. The _dl_strlen will crash with a NULL argument.
+
+Rather than relying on _dl_ldsopath, this change causes ldd to compute
+the interpreter's basename directly.
+
+glibc ld.so seems to print the full path to the interpreter without
+any computed basename or =>. I personally prefer glibc's behavior, but
+to preserve backwards compatibility with uClibc ld.so, the existing
+format with the computed basename, =>, and full path is used here. This
+enables simpler (and unchanged) text processing in a pipeline.
+
+Signed-off-by: Mark Mentovai <mark at moxienet.com>
+---
+ldso/ldso/ldso.c |   12 +++++++++---
+1 files changed, 9 insertions(+), 3 deletions(-)
+
+--- a/ldso/ldso/ldso.c
++++ b/ldso/ldso/ldso.c
+@@ -923,9 +923,15 @@ void _dl_get_ready_to_run(struct elf_res
+ #ifdef __LDSO_LDD_SUPPORT__
+ 	/* End of the line for ldd.... */
+ 	if (trace_loaded_objects) {
+-		_dl_dprintf(1, "\t%s => %s (%x)\n",
+-			    rpnt->dyn->libname + _dl_strlen(_dl_ldsopath) + 1,
+-			    rpnt->dyn->libname, DL_LOADADDR_BASE(rpnt->dyn->loadaddr));
++		/* glibc ld.so/ldd would just do
++		 * _dl_dprintf(1, "\t%s (%x)\n", rpnt->dyn->libname,
++		 *	       DL_LOADADDR_BASE(rpnt->dyn->loadaddr));
++		 * but uClibc has always used the => format. */
++		char *ptmp = _dl_strrchr(rpnt->dyn->libname, '/');
++		if (ptmp != rpnt->dyn->libname)
++			++ptmp;
++		_dl_dprintf(1, "\t%s => %s (%x)\n", ptmp, rpnt->dyn->libname,
++			    DL_LOADADDR_BASE(rpnt->dyn->loadaddr));
+ 		_dl_exit(0);
+ 	}
+ #endif