d/freebsd-nq
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
4b330699f8
freebsd-nq/libexec/rtld-elf/rtld.c

5514 lines
146 KiB
C
Raw Normal View History

Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
/*-
Revamp the mechanism for enumerating and calling shared objects' init and fini functions. Now the code is very careful to hold no locks when calling these functions. Thus the dynamic linker cannot be re-entered with a lock already held. Remove the tolerance for recursive locking that I added in revision 1.2 of dllockinit.c. Recursive locking shouldn't happen any more. Mozilla and JDK users: I'd appreciate confirmation that things still work right (or at least the same) with these changes.
2000-01-09 21:13:48 +00:00
* Copyright 1996, 1997, 1998, 1999, 2000 John D. Polstra.
Implement dlinfo() function. Introdice RTLD_SELF special handle and properly process it within dlsym() and dlinfo() functions. The intention is to improve our compatibility with Solaris and to make a Java port easier. Partially submitted by: phantom
2003-02-13 17:47:44 +00:00
* Copyright 2003 Alexander Kabaev <kan@FreeBSD.ORG>.
Update my copyright, note The FreeBSD Foundation involvement. While tweaking copyright block, switch to use __FBSDID for tag. Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2017-05-18 09:34:26 +00:00
* Copyright 2009-2013 Konstantin Belousov <kib@FreeBSD.ORG>.
Split the symlook_obj1 into a loop iterating over the ELF object symbol hash elements, and a helper matched_symbol() which match the given hash entry and request, performing needed type and version checks. Based on dragonflybsd support for GNU hash by John Marino <draco marino st> Reviewed by: kan Tested by: bapt MFC after: 2 weeks
2012-04-30 13:29:21 +00:00
* Copyright 2012 John Marino <draco@marino.st>.
Update my copyright, note The FreeBSD Foundation involvement. While tweaking copyright block, switch to use __FBSDID for tag. Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2017-05-18 09:34:26 +00:00
* Copyright 2014-2017 The FreeBSD Foundation
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
* All rights reserved.
*
Update my copyright, note The FreeBSD Foundation involvement. While tweaking copyright block, switch to use __FBSDID for tag. Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2017-05-18 09:34:26 +00:00
* Portions of this software were developed by Konstantin Belousov
* under sponsorship from the FreeBSD Foundation.
*
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
* IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
* OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
* IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
* INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
* NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
* THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
/*
* Dynamic linker for ELF.
*
* John Polstra <jdp@polstra.com>.
*/
Update my copyright, note The FreeBSD Foundation involvement. While tweaking copyright block, switch to use __FBSDID for tag. Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2017-05-18 09:34:26 +00:00
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
#include <sys/param.h>
If "dangerous" environment variables (LD_PRELOAD, LD_LIBMAP, LD_LIBMAP_DISABLE, LD_LIBRARY_PATH) are used, then make sure the libraries being loaded aren't on a noexec-mounted filesystem. This is a compromise position: I'm assuming that nobody will be silly enough to set the noexec mount flag on part of the default library path, in order to avoid adding extra overhead into the common case (where those environment variables aren't used). Discussed with: csjp, secteam MFC after: 1 week
2005-03-24 10:12:29 +00:00
#include <sys/mount.h>
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
#include <sys/mman.h>
When checking to see if a shared object is already loaded, look for a device/inode match if no pathname match is found.
1999-08-30 01:50:41 +00:00
#include <sys/stat.h>
Use the newly provided aux vectors to get pagesize and osreldate information. Use local version of getpagesize(), rtld_getpagesize() in private allocator. Override the __getosreldate() previously fetched from libc_pic.a with local version that uses aux value if present. Note that __getosreldate() is used by rtld indirectly, by mmap(2) libc wrapper. To be able to utilize aux, split digest_dynamic() for use by init_rtld() into two parts, where the first one does not call malloc(), and the second part uses it. init_rtld() is able to initialize global variables before digest_dynamic2() calls. In particular, pagesize and osreldate are set up from the aux values. Now, rtld avoids (two) sysctl calls in startup. Tested by: marius (sparc64) MFC after: 1 month
2010-08-17 09:05:39 +00:00
#include <sys/sysctl.h>
Add various utrace's for use with ktrace to the ELF runtime linker. To activate the traces, set the LD_UTRACE (or LD_32_UTRACE) environment variable. This also includes code in kdump(8) to parse the traces. Reviewed by: kan, jdp MFC after: 2 weeks
2007-01-09 17:50:05 +00:00
#include <sys/uio.h>
Implement the dynamic string token substitution in the rpath and soneeded pathes. The $ORIGIN, $OSNAME, $OSREL and $PLATFORM tokens are supported. Enabling the substitution requires DF_ORIGIN flag in DT_FLAGS or DF_1_ORIGIN if DF_FLAGS_1, that may be set with -z origin gnu ld flag. Translation is unconditionally disabled for setuid/setgid processes. The $ORIGIN translation relies on the AT_EXECPATH auxinfo supplied by kernel. Requested by: maho Tested by: maho, pho Reviewed by: kan
2009-03-18 13:40:37 +00:00
#include <sys/utsname.h>
Add various utrace's for use with ktrace to the ELF runtime linker. To activate the traces, set the LD_UTRACE (or LD_32_UTRACE) environment variable. This also includes code in kdump(8) to parse the traces. Reviewed by: kan, jdp MFC after: 2 weeks
2007-01-09 17:50:05 +00:00
#include <sys/ktrace.h>
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
#include <dlfcn.h>
#include <err.h>
#include <errno.h>
#include <fcntl.h>
#include <stdarg.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <unistd.h>
#include "debug.h"
#include "rtld.h"
Dynamic object dependency mapping: libmap. This is an optional feature, disabled by default. This will be useful to people testing the various POSIX threading libraries under -CURRENT but can easily serve other needs.
2003-04-07 16:21:26 +00:00
#include "libmap.h"
Move all the paths into a new path.h to centralize them.
2015-10-31 04:39:28 +00:00
#include "paths.h"
Add support for Thread Local Storage.
2004-08-03 08:51:00 +00:00
#include "rtld_tls.h"
Rtld links with the specially built pic static libc library to get some C runtime services, like printf(). Unfortunately, the multithread-safeness measures in the libc do not work in rtld environment. Rip the kernel printf() implementation and use it in the rtld instead of libc version. This printf does not require any shared global data and thus is mt-safe. Systematically use rtld_printf() and related functions, remove the calls to err(3). Note that stdio is still pulled from libc due to libmap implementaion using fopen(). This is safe but unoptimal, and can be changed later. Reported and tested by: pgj Diagnosed and reviewed by: kan (previous version) Approved by: re (bz)
2011-08-24 20:05:13 +00:00
#include "rtld_printf.h"
Move defines common between rtld and libsysdecode into the header, instead of copying inline into sources. Sponsored by: The FreeBSD Foundation MFC after: 1 week
2016-08-12 19:31:41 +00:00
#include "rtld_utrace.h"
Add support for preinit, init and fini arrays. Some ABIs, in particular on ARM, do require working init arrays. Traditional FreeBSD crt1 calls _init and _fini of the binary, instead of allowing runtime linker to arrange the calls. This was probably done to have the same crt code serve both statically and dynamically linked binaries. Since ABI mandates that first is called preinit array functions, then init, and then init array functions, the init have to be called from rtld now. To provide binary compatibility to old FreeBSD crt1, which calls _init itself, rtld only calls intializers and finalizers for main binary if binary has a note indicating that new crt was used for linking. Add parsing of ELF notes to rtld, and cache p_osrel value since we parsed it anyway. The patch is inspired by init_array support for DragonflyBSD, written by John Marino. Reviewed by: kan Tested by: andrew (arm, previous version), flo (sparc64, previous version) MFC after: 3 weeks
2012-03-11 20:03:09 +00:00
#include "notes.h"
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
/* Types. */
typedef void (*func_ptr_type)();
Implement dlinfo() function. Introdice RTLD_SELF special handle and properly process it within dlsym() and dlinfo() functions. The intention is to improve our compatibility with Solaris and to make a Java port easier. Partially submitted by: phantom
2003-02-13 17:47:44 +00:00
typedef void * (*path_enum_proc) (const char *path, size_t len, void *arg);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
/*
* Function declarations.
*/
After relocating the main program, but before calling any of the _init() functions, initialize the global variables "__progname" and "environ". This makes it possible for the _init() functions to call things like getenv() and err().
1999-04-21 04:06:57 +00:00
static const char *basename(const char *);
Use the newly provided aux vectors to get pagesize and osreldate information. Use local version of getpagesize(), rtld_getpagesize() in private allocator. Override the __getosreldate() previously fetched from libc_pic.a with local version that uses aux value if present. Note that __getosreldate() is used by rtld indirectly, by mmap(2) libc wrapper. To be able to utilize aux, split digest_dynamic() for use by init_rtld() into two parts, where the first one does not call malloc(), and the second part uses it. init_rtld() is able to initialize global variables before digest_dynamic2() calls. In particular, pagesize and osreldate are set up from the aux values. Now, rtld avoids (two) sysctl calls in startup. Tested by: marius (sparc64) MFC after: 1 month
2010-08-17 09:05:39 +00:00
static void digest_dynamic1(Obj_Entry *, int, const Elf_Dyn **,
Import the DragonFly BSD commit 4f0bc915b65fcf5a23214f6d221d65c80be68ad4 by John Marino <draco@marino.st>, with the following (edited) commit message Date: Sat, 24 Mar 2012 06:40:50 +0100 Subject: [PATCH 1/1] rtld: Implement DT_RUNPATH and -z nodefaultlib DT_RUNPATH is incorrectly being considered as an alias of DT_RPATH. The purpose of DT_RUNPATH is to have two different types of rpath: one that can be overridden by the environment variable LD_LIBRARY_PATH and one that can't. With the currently implementation, LD_LIBRARY_PATH will always trump any embedded rpath or runpath tags. Current path search order by rtld: ================================== LD_LIBRARY_PATH DT_RPATH / DT_RUNPATH (always the same) ldconfig hints file (default: /var/run/ld-elf.so.hints) /usr/lib New path search order by rtld: ============================== DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file /usr/lib The new path search matches how the linux runtime loader works. The other major added feature is support for linker flag "-z nodefaultlib". When this flag is passed to the linker, rtld will skip all references to the standard library search path ("/usr/lib" in this case but it could handle more color delimited paths) except in DT_RPATH and DT_RUNPATH. New path search order by rtld with -z nodefaultlib flag set: ============================================================ DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file (skips all references to /usr/lib) FreeBSD notes: - we fixed some bugs which were submitted to DragonFly and merged there as commit 1ff8a2bd3eb6e5587174c6a983303ea3a79e0002; - we added LD_LIBRARY_PATH_RPATH environment variable to switch to the previous behaviour of considering DT_RPATH a synonym for DT_RUNPATH; - the FreeBSD default search path is /lib:/usr/lib and not /usr/lib. Reviewed by: kan MFC after: 1 month MFC note: flip the ld_library_path_rpath default value for stable/9
2012-07-15 10:53:48 +00:00
const Elf_Dyn **, const Elf_Dyn **);
static void digest_dynamic2(Obj_Entry *, const Elf_Dyn *, const Elf_Dyn *,
const Elf_Dyn *);
Minor changes to make this work on sparc64. Approved by: jdp Tested on: alpha, i386, sparc64
2002-04-02 02:19:02 +00:00
static void digest_dynamic(Obj_Entry *, int);
Change many asserts into normal errors. They were all for conditions caused by invalid shared objects rather than by internal errors. Enable format string mismatch checking for _rtld_error().
1999-07-18 00:02:19 +00:00
static Obj_Entry *digest_phdr(const Elf_Phdr *, int, caddr_t, const char *);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
static Obj_Entry *dlcheck(void *);
Avoid bind lock recursion. When dlclose(3) unloads an object with filtees, it recursively calls dlclose(3) on each filtee in free_needed_filtees(). Introduce dlclose_locked() helper, called from free_needed_filtees() instead of dlclose(), and pass the bind lockstate down to avoid recursing. Reported and tested by: jhibbits Sponsored by: The FreeBSD Foundation MFC after: 1 week
2017-03-09 21:05:47 +00:00
static int dlclose_locked(void *, RtldLockState *);
Implement fdlopen(3), an rtld interface to load shared object by file descriptor. Requested and tested by: des (previous version) Reviewed by: des, kan (previous version) MFC after: 2 weeks
2012-01-07 10:33:01 +00:00
static Obj_Entry *dlopen_object(const char *name, int fd, Obj_Entry *refobj,
Propagate the current state of rtld_bind_lock to dlopen_object() calls through the filter loading call chain. This fixes attempts to write-lock the already locked rtld_bind_lock when filter loading is initiated by relocation of dlopening dso. Reported and tested by: Taku YAMAMOTO <taku tackymt homeip net> MFC after: 1 week
2012-04-12 10:32:22 +00:00
int lo_flags, int mode, RtldLockState *lockstate);
Implement rtld part of the support for -z nodlopen (see ld(1)). Reviewed by: kan MFC after: 3 weeks
2009-11-26 13:57:20 +00:00
static Obj_Entry *do_load_object(int, const char *, char *, struct stat *, int);
Implement dlinfo() function. Introdice RTLD_SELF special handle and properly process it within dlsym() and dlinfo() functions. The intention is to improve our compatibility with Solaris and to make a Java port easier. Partially submitted by: phantom
2003-02-13 17:47:44 +00:00
static int do_search_info(const Obj_Entry *obj, int, struct dl_serinfo *);
Add support for dlsym(RTLD_DEFAULT, ...).
2000-09-19 04:27:16 +00:00
static bool donelist_check(DoneList *, const Obj_Entry *);
Fix a bug in which a program called dlclose from a destructor and got an assert failure in the dynamic linker.
2001-01-05 04:36:17 +00:00
static void errmsg_restore(char *);
static char *errmsg_save(void);
Implement dlinfo() function. Introdice RTLD_SELF special handle and properly process it within dlsym() and dlinfo() functions. The intention is to improve our compatibility with Solaris and to make a Java port easier. Partially submitted by: phantom
2003-02-13 17:47:44 +00:00
static void *fill_search_info(const char *, size_t, void *);
Add the LD_LIBRARY_PATH_FDS environmental variable. This variable allows the loading of shared libraries via directory descriptors rather than via library paths. If LD_LIBRARY_PATH_FDS=3:4:12, the directories represented by file descriptors 3, 4 and 12 will searched for shared libraries before the normal path-based mechanisms are used. This allows us to execute unprivileged binaries from within a Capsicum sandbox even if they require shared libraries. Approved by: rwatson (mentor) Reviewed by: kib MFC after: 3 weeks Sponsored by: DARPA/AFRL
2014-06-20 17:08:32 +00:00
static char *find_library(const char *, const Obj_Entry *, int *);
Import the DragonFly BSD commit 4f0bc915b65fcf5a23214f6d221d65c80be68ad4 by John Marino <draco@marino.st>, with the following (edited) commit message Date: Sat, 24 Mar 2012 06:40:50 +0100 Subject: [PATCH 1/1] rtld: Implement DT_RUNPATH and -z nodefaultlib DT_RUNPATH is incorrectly being considered as an alias of DT_RPATH. The purpose of DT_RUNPATH is to have two different types of rpath: one that can be overridden by the environment variable LD_LIBRARY_PATH and one that can't. With the currently implementation, LD_LIBRARY_PATH will always trump any embedded rpath or runpath tags. Current path search order by rtld: ================================== LD_LIBRARY_PATH DT_RPATH / DT_RUNPATH (always the same) ldconfig hints file (default: /var/run/ld-elf.so.hints) /usr/lib New path search order by rtld: ============================== DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file /usr/lib The new path search matches how the linux runtime loader works. The other major added feature is support for linker flag "-z nodefaultlib". When this flag is passed to the linker, rtld will skip all references to the standard library search path ("/usr/lib" in this case but it could handle more color delimited paths) except in DT_RPATH and DT_RUNPATH. New path search order by rtld with -z nodefaultlib flag set: ============================================================ DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file (skips all references to /usr/lib) FreeBSD notes: - we fixed some bugs which were submitted to DragonFly and merged there as commit 1ff8a2bd3eb6e5587174c6a983303ea3a79e0002; - we added LD_LIBRARY_PATH_RPATH environment variable to switch to the previous behaviour of considering DT_RPATH a synonym for DT_RUNPATH; - the FreeBSD default search path is /lib:/usr/lib and not /usr/lib. Reviewed by: kan MFC after: 1 month MFC note: flip the ld_library_path_rpath default value for stable/9
2012-07-15 10:53:48 +00:00
static const char *gethints(bool);
rtld: Fix a race between dl_iterate_phdr() and dlclose(). Add a transient reference count to ensure that the phdr argument to the callback remains valid while the bind lock is dropped. Reviewed by: kib MFC after: 2 weeks Sponsored by: Dell EMC Isilon
2016-12-22 17:37:39 +00:00
static void hold_object(Obj_Entry *);
static void unhold_object(Obj_Entry *);
Revamp the symbol lookup algorithm to cope better with objects loaded separately by dlopen that have global symbols with identical names. Viewing each dlopened object as a DAG which is linked by its DT_NEEDED entries in the dynamic table, the search order is as follows: * If the referencing object was linked with -Bsymbolic, search it internally. * Search all dlopened DAGs containing the referencing object. * Search all objects loaded at program start up. * Search all objects which were dlopened() using the RTLD_GLOBAL flag (which is now supported too). The search terminates as soon as a strong definition is found. Lacking that, the first weak definition is used. These rules match those of Solaris, as best I could determine them from its vague manual pages and the results of experiments I performed. PR: misc/12438
1999-08-30 01:48:19 +00:00
static void init_dag(Obj_Entry *);
rtld: Fix a couple of bugs around the unloading of ELF filters. - Pass the correct object to unload_filtees(). - Use a marker to restart iteration after unload_filtees() has returned. It calls dlclose() and may recursively remove entries from the global object list, so TAILQ_FOREACH_SAFE is not sufficient. Reviewed by: kib MFC after: 2 weeks Sponsored by: Dell EMC Isilon
2016-12-22 17:44:27 +00:00
static void init_marker(Obj_Entry *);
Before calling mmap() on a shared library's text and data sections, rtld first calls mmap() with the arguments PROT_NONE and MAP_ANON to reserve a single, contiguous range of virtual addresses for the entire shared library. Later, rtld calls mmap() with the the shared library's file descriptor and the argument MAP_FIXED to place the text and data sections within the reserved range. The rationale for mapping shared libraries in this way is explained in the commit message for Revision 190885. However, this approach does have an unintended, negative consequence. Since the first call to mmap() specifies MAP_ANON and not the shared library's file descriptor, the kernel has no idea what alignment the vm object backing the file prefers. As a result, the reserved range's alignment is unlikely to be the same as the vm object's, and so mapping with superpages becomes impossible. To address this problem, this revision adds the argument MAP_ALIGNED_SUPER to the first call to mmap() if the text section is larger than the smallest superpage size. To determine if the text section is larger than the smallest superpage size, rtld must always fetch the page size information. As a result, the private code for fetching the base page size in rtld's builtin malloc is redundant. Eliminate it. Requested by: kib Tested by: zbb (on arm) Reviewed by: kib (an earlier version) Discussed with: jhb
2014-04-11 16:55:25 +00:00
static void init_pagesizes(Elf_Auxinfo **aux_info);
Use the newly provided aux vectors to get pagesize and osreldate information. Use local version of getpagesize(), rtld_getpagesize() in private allocator. Override the __getosreldate() previously fetched from libc_pic.a with local version that uses aux value if present. Note that __getosreldate() is used by rtld indirectly, by mmap(2) libc wrapper. To be able to utilize aux, split digest_dynamic() for use by init_rtld() into two parts, where the first one does not call malloc(), and the second part uses it. init_rtld() is able to initialize global variables before digest_dynamic2() calls. In particular, pagesize and osreldate are set up from the aux values. Now, rtld avoids (two) sysctl calls in startup. Tested by: marius (sparc64) MFC after: 1 month
2010-08-17 09:05:39 +00:00
static void init_rtld(caddr_t, Elf_Auxinfo **);
Clean up white spaces and fix style(9).
2006-09-19 16:48:08 +00:00
static void initlist_add_neededs(Needed_Entry *, Objlist *);
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
static void initlist_add_objects(Obj_Entry *, Obj_Entry *, Objlist *);
Style fixes. If it seems like a lot of lines of changes, it's because I moved some functions. Mr. Tidy likes them to be in alphabetical order.
1998-09-02 02:00:20 +00:00
static void linkmap_add(Obj_Entry *);
static void linkmap_delete(Obj_Entry *);
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
static void load_filtees(Obj_Entry *, int flags, RtldLockState *);
Avoid bind lock recursion. When dlclose(3) unloads an object with filtees, it recursively calls dlclose(3) on each filtee in free_needed_filtees(). Introduce dlclose_locked() helper, called from free_needed_filtees() instead of dlclose(), and pass the bind lockstate down to avoid recursing. Reported and tested by: jhibbits Sponsored by: The FreeBSD Foundation MFC after: 1 week
2017-03-09 21:05:47 +00:00
static void unload_filtees(Obj_Entry *, RtldLockState *);
Implement rtld part of the support for -z nodlopen (see ld(1)). Reviewed by: kan MFC after: 3 weeks
2009-11-26 13:57:20 +00:00
static int load_needed_objects(Obj_Entry *, int);
Make LD_PRELOAD work for ELF.
1998-09-22 02:09:56 +00:00
static int load_preload_objects(void);
Implement fdlopen(3), an rtld interface to load shared object by file descriptor. Requested and tested by: des (previous version) Reviewed by: des, kan (previous version) MFC after: 2 weeks
2012-01-07 10:33:01 +00:00
static Obj_Entry *load_object(const char *, int fd, const Obj_Entry *, int);
get_program_var_addr() must prefer the strong symbol to the weak one. Search global objects, together with main object and dependencies, for the requested symbol. Move the common code from symlook_default() into new helper symlook_global(), and use it both in symlook_global() and get_program_var_addr(). Supply lock state to get_program_var_addr(). Reviewed by: kan Tested by: Mykola Dzham <i levsha me>
2011-01-10 16:09:35 +00:00
static void map_stacks_exec(RtldLockState *);
For the main binary, postpone enforcing relro read-only protection until copy relocations are done. Newer binutils and lld seems to output copy into relro-protected range. Reported by: Rafael Espц╜ndola via emaste Sponsored by: The FreeBSD Foundation MFC after: 1 week
2017-01-12 15:54:03 +00:00
static int obj_enforce_relro(Obj_Entry *);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
static Obj_Entry *obj_from_addr(const void *);
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
static void objlist_call_fini(Objlist *, Obj_Entry *, RtldLockState *);
static void objlist_call_init(Objlist *, RtldLockState *);
Revamp the code that calls shared libraries' init and fini functions. Formerly the init functions were called in the opposite of the order in which libraries were loaded, and libraries were loaded according to a breadth-first traversal of the dependency graph. That ordering came from SVR4.0, and it was easy to implement but not always sensible. Now we do a depth-first walk over the dependency graph and call the init functions in an order such that each shared object's needed objects are initialized before the shared object itself. At the same time we build a list of finalization (fini) functions in the opposite order, to guarantee correct C++ destructor ordering whenever possible. (It may not be possible if dlopen and dlclose are used in strange ways, but we come as close as one can come.) The need for this renovation has become apparent as more programs have started using multithreading. The multithreaded C library libc_r requires initialization, whereas the standard libc does not. Since virtually every other object depends on the C library, it is important that it get initialized first.
2000-07-26 04:24:40 +00:00
static void objlist_clear(Objlist *);
Revamp the symbol lookup algorithm to cope better with objects loaded separately by dlopen that have global symbols with identical names. Viewing each dlopened object as a DAG which is linked by its DT_NEEDED entries in the dynamic table, the search order is as follows: * If the referencing object was linked with -Bsymbolic, search it internally. * Search all dlopened DAGs containing the referencing object. * Search all objects loaded at program start up. * Search all objects which were dlopened() using the RTLD_GLOBAL flag (which is now supported too). The search terminates as soon as a strong definition is found. Lacking that, the first weak definition is used. These rules match those of Solaris, as best I could determine them from its vague manual pages and the results of experiments I performed. PR: misc/12438
1999-08-30 01:48:19 +00:00
static Objlist_Entry *objlist_find(Objlist *, const Obj_Entry *);
Revamp the code that calls shared libraries' init and fini functions. Formerly the init functions were called in the opposite of the order in which libraries were loaded, and libraries were loaded according to a breadth-first traversal of the dependency graph. That ordering came from SVR4.0, and it was easy to implement but not always sensible. Now we do a depth-first walk over the dependency graph and call the init functions in an order such that each shared object's needed objects are initialized before the shared object itself. At the same time we build a list of finalization (fini) functions in the opposite order, to guarantee correct C++ destructor ordering whenever possible. (It may not be possible if dlopen and dlclose are used in strange ways, but we come as close as one can come.) The need for this renovation has become apparent as more programs have started using multithreading. The multithreaded C library libc_r requires initialization, whereas the standard libc does not. Since virtually every other object depends on the C library, it is important that it get initialized first.
2000-07-26 04:24:40 +00:00
static void objlist_init(Objlist *);
static void objlist_push_head(Objlist *, Obj_Entry *);
static void objlist_push_tail(Objlist *, Obj_Entry *);
Implement support for the interpose dso flag. Requested by: bf Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 1 week Approved by: re (glebius)
2013-10-07 08:19:30 +00:00
static void objlist_put_after(Objlist *, Obj_Entry *, Obj_Entry *);
Revamp the symbol lookup algorithm to cope better with objects loaded separately by dlopen that have global symbols with identical names. Viewing each dlopened object as a DAG which is linked by its DT_NEEDED entries in the dynamic table, the search order is as follows: * If the referencing object was linked with -Bsymbolic, search it internally. * Search all dlopened DAGs containing the referencing object. * Search all objects loaded at program start up. * Search all objects which were dlopened() using the RTLD_GLOBAL flag (which is now supported too). The search terminates as soon as a strong definition is found. Lacking that, the first weak definition is used. These rules match those of Solaris, as best I could determine them from its vague manual pages and the results of experiments I performed. PR: misc/12438
1999-08-30 01:48:19 +00:00
static void objlist_remove(Objlist *, Obj_Entry *);
For ld.so direct execution mode, implement -p option: search for the binary in $PATH. Sponsored by: The FreeBSD Foundation MFC after: 2 weeks Differential revision: https://reviews.freebsd.org/D10790
2017-05-23 10:00:52 +00:00
static int open_binary_fd(const char *argv0, bool search_in_path);
Allow rtld direct-exec to take a file descriptor. When executing rtld directly, allow a file descriptor to be explicitly specified rather than opened from the given path. This, together with the LD_LIBRARY_PATH_FDS environment variable, allows dynamically-linked applications to be executed from within capability mode. Also add some rudimentary argument parsing (without pulling in getopt or the like) to accept this file descriptor, a help (-h) option and a basic usage string. Reviewed by: kib Sponsored by: NSERC, RDC Differential Revision: https://reviews.freebsd.org/D10751
2017-05-17 22:51:28 +00:00
static int parse_args(char* argv[], int argc, bool *use_pathp, int *fdp);
Rename rtld's parse_libdir to parse_integer. This is a more accurate name, as the integer doesn't have to be a library directory descriptor. It is also a prerequisite for more argument parsing coming in the near future (e.g., parsing explicit binary descriptors). Reviewed by: kib MFC after: 2 weeks Sponsored by: NSERC
2017-05-16 13:27:44 +00:00
static int parse_integer(const char *);
Implement dlinfo() function. Introdice RTLD_SELF special handle and properly process it within dlsym() and dlinfo() functions. The intention is to improve our compatibility with Solaris and to make a Java port easier. Partially submitted by: phantom
2003-02-13 17:47:44 +00:00
static void *path_enumerate(const char *, path_enum_proc, void *);
Allow rtld direct-exec to take a file descriptor. When executing rtld directly, allow a file descriptor to be explicitly specified rather than opened from the given path. This, together with the LD_LIBRARY_PATH_FDS environment variable, allows dynamically-linked applications to be executed from within capability mode. Also add some rudimentary argument parsing (without pulling in getopt or the like) to accept this file descriptor, a help (-h) option and a basic usage string. Reviewed by: kib Sponsored by: NSERC, RDC Differential Revision: https://reviews.freebsd.org/D10751
2017-05-17 22:51:28 +00:00
static void print_usage(const char *argv0);
rtld: Fix a race between dl_iterate_phdr() and dlclose(). Add a transient reference count to ensure that the phdr argument to the callback remains valid while the bind lock is dropped. Reviewed by: kib MFC after: 2 weeks Sponsored by: Dell EMC Isilon
2016-12-22 17:37:39 +00:00
static void release_object(Obj_Entry *);
Ensure that for the object which is a dependency for some filtee, relocations are performed before the object's initializer is called. When dlopen()ing an object, relocate the whole DAG rooted in the object instead of only relocating the object itself and list of newly loaded dependencies. Reversed sequence currently can occur if the same object is a dependency for both filtee and filter, since filtees are loaded typically during the relocation processing, when some filter dependencies might be already loaded but not relocated yet. Reported and tested by: swills Reviewed by: kan MFC after: 1 week
2012-06-27 20:24:25 +00:00
static int relocate_object_dag(Obj_Entry *root, bool bind_now,
Obj_Entry *rtldobj, int flags, RtldLockState *lockstate);
static int relocate_object(Obj_Entry *obj, bool bind_now, Obj_Entry *rtldobj,
int flags, RtldLockState *lockstate);
Fix several problems with our ELF filters implementation. Do not relocate twice an object which happens to be needed by loaded binary (or dso) and some filtee opened due to symbol resolution when relocating need objects. Record the state of the relocation processing in Obj_Entry and short-circuit relocate_objects() if current object already processed. Do not call constructors for filtees loaded during the early relocation processing before image is initialized enough to run user-provided code. Filtees are loaded using dlopen_object(), which normally performs relocation and initialization. If filtee is lazy-loaded during the relocation of dso needed by the main object, dlopen_object() runs too earlier, when most runtime services are not yet ready. Postpone the constructors call to the time when main binary and depended libraries constructors are run, passing the new flag RTLD_LO_EARLY to dlopen_object(). Symbol lookups callers inform symlook_* functions about early stage of initialization with SYMLOOK_EARLY. Pass flags through all functions participating in object relocation. Use the opportunity and fix flags argument to find_symdef() in arch-specific reloc.c to use proper name SYMLOOK_IN_PLT instead of true, which happen to have the same numeric value. Reported and tested by: theraven Reviewed by: kan MFC after: 2 weeks
2012-03-20 13:20:49 +00:00
static int relocate_objects(Obj_Entry *, bool, Obj_Entry *, int,
RtldLockState *);
_rtld_bind() read-locks the bind lock, and possible plt resolution from the dispatcher would also acquire bind lock in read mode, which is the supported operation. plt is explicitely designed to allow safe multithreaded updates, so the shared lock do not cause problems. The error in r228435 is that it allows read lock acquisition after the write lock for the bind block. If we dlopened the shared object that contains IRELATIVE or jump slot which target is STT_GNU_IFUNC, then possible recursive plt resolve from the dispatcher would cause it. Postpone the resolution for irelative/ifunc right before initializers are called, and drop bind lock around calls to dispatcher. Use initlist to iterate over the objects instead of the ->next, due to drop of the bind lock in iteration. For i386/reloc.c:reloc_iresolve(), fix calculation of the dispatch function address for dso, by taking into account possible non-zero relocbase. MFC after: 3 weeks
2011-12-14 16:47:53 +00:00
static int resolve_objects_ifunc(Obj_Entry *first, bool bind_now,
Fix several problems with our ELF filters implementation. Do not relocate twice an object which happens to be needed by loaded binary (or dso) and some filtee opened due to symbol resolution when relocating need objects. Record the state of the relocation processing in Obj_Entry and short-circuit relocate_objects() if current object already processed. Do not call constructors for filtees loaded during the early relocation processing before image is initialized enough to run user-provided code. Filtees are loaded using dlopen_object(), which normally performs relocation and initialization. If filtee is lazy-loaded during the relocation of dso needed by the main object, dlopen_object() runs too earlier, when most runtime services are not yet ready. Postpone the constructors call to the time when main binary and depended libraries constructors are run, passing the new flag RTLD_LO_EARLY to dlopen_object(). Symbol lookups callers inform symlook_* functions about early stage of initialization with SYMLOOK_EARLY. Pass flags through all functions participating in object relocation. Use the opportunity and fix flags argument to find_symdef() in arch-specific reloc.c to use proper name SYMLOOK_IN_PLT instead of true, which happen to have the same numeric value. Reported and tested by: theraven Reviewed by: kan MFC after: 2 weeks
2012-03-20 13:20:49 +00:00
int flags, RtldLockState *lockstate);
Implement dlinfo() function. Introdice RTLD_SELF special handle and properly process it within dlsym() and dlinfo() functions. The intention is to improve our compatibility with Solaris and to make a Java port easier. Partially submitted by: phantom
2003-02-13 17:47:44 +00:00
static int rtld_dirname(const char *, char *);
Implement the dynamic string token substitution in the rpath and soneeded pathes. The $ORIGIN, $OSNAME, $OSREL and $PLATFORM tokens are supported. Enabling the substitution requires DF_ORIGIN flag in DT_FLAGS or DF_1_ORIGIN if DF_FLAGS_1, that may be set with -z origin gnu ld flag. Translation is unconditionally disabled for setuid/setgid processes. The $ORIGIN translation relies on the AT_EXECPATH auxinfo supplied by kernel. Requested by: maho Tested by: maho, pho Reviewed by: kan
2009-03-18 13:40:37 +00:00
static int rtld_dirname_abs(const char *, char *);
Implement fdlopen(3), an rtld interface to load shared object by file descriptor. Requested and tested by: des (previous version) Reviewed by: des, kan (previous version) MFC after: 2 weeks
2012-01-07 10:33:01 +00:00
static void *rtld_dlopen(const char *name, int fd, int mode);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
static void rtld_exit(void);
static char *search_library_path(const char *, const char *);
Add the LD_LIBRARY_PATH_FDS environmental variable. This variable allows the loading of shared libraries via directory descriptors rather than via library paths. If LD_LIBRARY_PATH_FDS=3:4:12, the directories represented by file descriptors 3, 4 and 12 will searched for shared libraries before the normal path-based mechanisms are used. This allows us to execute unprivileged binaries from within a Capsicum sandbox even if they require shared libraries. Approved by: rwatson (mentor) Reviewed by: kib MFC after: 3 weeks Sponsored by: DARPA/AFRL
2014-06-20 17:08:32 +00:00
static char *search_library_pathfds(const char *, const char *, int *);
get_program_var_addr() must prefer the strong symbol to the weak one. Search global objects, together with main object and dependencies, for the requested symbol. Move the common code from symlook_default() into new helper symlook_global(), and use it both in symlook_global() and get_program_var_addr(). Supply lock state to get_program_var_addr(). Reviewed by: kan Tested by: Mykola Dzham <i levsha me>
2011-01-10 16:09:35 +00:00
static const void **get_program_var_addr(const char *, RtldLockState *);
After relocating the main program, but before calling any of the _init() functions, initialize the global variables "__progname" and "environ". This makes it possible for the _init() functions to call things like getenv() and err().
1999-04-21 04:06:57 +00:00
static void set_program_var(const char *, const void *);
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
static int symlook_default(SymLook *, const Obj_Entry *refobj);
get_program_var_addr() must prefer the strong symbol to the weak one. Search global objects, together with main object and dependencies, for the requested symbol. Move the common code from symlook_default() into new helper symlook_global(), and use it both in symlook_global() and get_program_var_addr(). Supply lock state to get_program_var_addr(). Reviewed by: kan Tested by: Mykola Dzham <i levsha me>
2011-01-10 16:09:35 +00:00
static int symlook_global(SymLook *, DoneList *);
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
static void symlook_init_from_req(SymLook *, const SymLook *);
static int symlook_list(SymLook *, const Objlist *, DoneList *);
static int symlook_needed(SymLook *, const Needed_Entry *, DoneList *);
Add GNU hash support for rtld. Based on dragonflybsd support for GNU hash by John Marino <draco marino st> Reviewed by: kan Tested by: bapt MFC after: 2 weeks
2012-04-30 13:31:10 +00:00
static int symlook_obj1_sysv(SymLook *, const Obj_Entry *);
static int symlook_obj1_gnu(SymLook *, const Obj_Entry *);
Clean up white spaces and fix style(9).
2006-09-19 16:48:08 +00:00
static void trace_loaded_objects(Obj_Entry *);
Do not remove object from the lists at the unref_dag() stage. Introduce a new unlink_object() function and call it in unload_object() instead. Removing the object in unref_dag() is too early, rtld calls _fini() function after that and shared objects might fail resolve their own symbols.
2003-02-17 20:58:27 +00:00
static void unlink_object(Obj_Entry *);
Avoid bind lock recursion. When dlclose(3) unloads an object with filtees, it recursively calls dlclose(3) on each filtee in free_needed_filtees(). Introduce dlclose_locked() helper, called from free_needed_filtees() instead of dlclose(), and pass the bind lockstate down to avoid recursing. Reported and tested by: jhibbits Sponsored by: The FreeBSD Foundation MFC after: 1 week
2017-03-09 21:05:47 +00:00
static void unload_object(Obj_Entry *, RtldLockState *lockstate);
Revamp the symbol lookup algorithm to cope better with objects loaded separately by dlopen that have global symbols with identical names. Viewing each dlopened object as a DAG which is linked by its DT_NEEDED entries in the dynamic table, the search order is as follows: * If the referencing object was linked with -Bsymbolic, search it internally. * Search all dlopened DAGs containing the referencing object. * Search all objects loaded at program start up. * Search all objects which were dlopened() using the RTLD_GLOBAL flag (which is now supported too). The search terminates as soon as a strong definition is found. Lacking that, the first weak definition is used. These rules match those of Solaris, as best I could determine them from its vague manual pages and the results of experiments I performed. PR: misc/12438
1999-08-30 01:48:19 +00:00
static void unref_dag(Obj_Entry *);
Rethink the way we count module references. Simply following DT_NEEDED links is not flexible enough for cases where dynamically loaded modules form a dependency cycle. This should fix an infinite recursion problem encountered by Yahoo. Approved by: re (jhb)
2003-05-08 01:31:36 +00:00
static void ref_dag(Obj_Entry *);
Change interpretation of the DF_ORIGIN and DF_1_ORIGIN flags. According to standard, the presence of the flags only means that the object path must be resolved at the time object loading, instead of my reading that the flag is required to enable token substitution at all. The consequence is that -z origin linker flag is no longer required for the token substitution in the run/rpath or the needed library soname. It is only recommended if token substition is needed at dlopen(3) time, since namecache might drop the required entries at the time of resolution. Found, reviewed and tested by: emaste Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-04-27 18:41:31 +00:00
static char *origin_subst_one(Obj_Entry *, char *, const char *,
const char *, bool);
static char *origin_subst(Obj_Entry *, char *);
static bool obj_resolve_origin(Obj_Entry *obj);
Add support for preinit, init and fini arrays. Some ABIs, in particular on ARM, do require working init arrays. Traditional FreeBSD crt1 calls _init and _fini of the binary, instead of allowing runtime linker to arrange the calls. This was probably done to have the same crt code serve both statically and dynamically linked binaries. Since ABI mandates that first is called preinit array functions, then init, and then init array functions, the init have to be called from rtld now. To provide binary compatibility to old FreeBSD crt1, which calls _init itself, rtld only calls intializers and finalizers for main binary if binary has a note indicating that new crt was used for linking. Add parsing of ELF notes to rtld, and cache p_osrel value since we parsed it anyway. The patch is inspired by init_array support for DragonflyBSD, written by John Marino. Reviewed by: kan Tested by: andrew (arm, previous version), flo (sparc64, previous version) MFC after: 3 weeks
2012-03-11 20:03:09 +00:00
static void preinit_main(void);
Implement ELF symbol versioning using GNU semantics. This code aims to be compatible with symbol versioning support as implemented by GNU libc and documented by http://people.redhat.com/~drepper/symbol-versioning and LSB 3.0. Implement dlvsym() function to allow lookups for a specific version of a given symbol.
2005-12-18 19:43:33 +00:00
static int rtld_verify_versions(const Objlist *);
static int rtld_verify_object_versions(Obj_Entry *);
static void object_add_name(Obj_Entry *, const char *);
static int object_match_name(const Obj_Entry *, const char *);
Add various utrace's for use with ktrace to the ELF runtime linker. To activate the traces, set the LD_UTRACE (or LD_32_UTRACE) environment variable. This also includes code in kdump(8) to parse the traces. Reviewed by: kan, jdp MFC after: 2 weeks
2007-01-09 17:50:05 +00:00
static void ld_utrace_log(int, void *, void *, size_t, int, const char *);
Introduce implementation-private rtld interface _rtld_addr_phdr, which fills struct dl_phdr_info for the shared object that contains the specified address, if any. Idea and reviewed by: kan MFC after: 3 weeks
2010-08-23 15:27:03 +00:00
static void rtld_fill_dl_phdr_info(const Obj_Entry *obj,
struct dl_phdr_info *phdr_info);
Add GNU hash support for rtld. Based on dragonflybsd support for GNU hash by John Marino <draco marino st> Reviewed by: kan Tested by: bapt MFC after: 2 weeks
2012-04-30 13:31:10 +00:00
static uint32_t gnu_hash(const char *);
Split the symlook_obj1 into a loop iterating over the ELF object symbol hash elements, and a helper matched_symbol() which match the given hash entry and request, performing needed type and version checks. Based on dragonflybsd support for GNU hash by John Marino <draco marino st> Reviewed by: kan Tested by: bapt MFC after: 2 weeks
2012-04-30 13:29:21 +00:00
static bool matched_symbol(SymLook *, const Obj_Entry *, Sym_Match_Result *,
const unsigned long);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
Change compiler setting to make default visibility of the symbols for rtld on x86 to be hidden. This is a micro-optimization, which allows intrinsic references inside rtld to be handled without indirection through PLT. The visibility of rtld symbols for other objects in the symbol namespace is controlled by a version script. Reviewed by: kan, jilles Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2015-03-29 18:53:21 +00:00
void r_debug_state(struct r_debug *, struct link_map *) __noinline __exported;
void _r_debug_postinit(struct link_map *) __noinline __exported;
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
Do not erronously export 'openat' symbol from rtld. The symbol leaked after r276630 since lib/libc/sys/openat.c defines versions for openat using .symver (version script cannot assign two versions to one symbol), and rtld uses openat. Instead, directly use __sys_openat(). Reported and tested by: antoine Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-01-04 00:33:41 +00:00
int __sys_openat(int, const char *, int, ...);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
/*
* Data declarations.
*/
static char *error_message; /* Message for dlerror(), or NULL */
Change compiler setting to make default visibility of the symbols for rtld on x86 to be hidden. This is a micro-optimization, which allows intrinsic references inside rtld to be handled without indirection through PLT. The visibility of rtld symbols for other objects in the symbol namespace is controlled by a version script. Reviewed by: kan, jilles Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2015-03-29 18:53:21 +00:00
struct r_debug r_debug __exported; /* for GDB; */
Use the environment variable LD_LIBMAP_DISABLE to disable libmap.conf(5) functionality.
2003-05-31 14:45:11 +00:00
static bool libmap_disable; /* Disable libmap */
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
static bool ld_loadfltr; /* Immediate filters processing */
Description from Dan: Another handy libmap patch. Lets you do stuff like this: LD_LIBMAP="libpthread.so.1=libthr.so.1" mythreadedapp If you already have a program-specific override in libmap.conf, note that you must use a program-specific override in LD_LIBMAP: LD_LIBMAP="[mythreadedapp],libpthread.so.1=libthr.so.1" mythreadedapp PR: bin/74471 Submitted by: Dan Nelson <dnelson AT allantgroup.com> MFC after: 2 weeks
2005-02-04 02:46:41 +00:00
static char *libmap_override; /* Maps to use in addition to libmap.conf */
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
static bool trust; /* False for setuid and setgid programs */
If "dangerous" environment variables (LD_PRELOAD, LD_LIBMAP, LD_LIBMAP_DISABLE, LD_LIBRARY_PATH) are used, then make sure the libraries being loaded aren't on a noexec-mounted filesystem. This is a compromise position: I'm assuming that nobody will be silly enough to set the noexec mount flag on part of the default library path, in order to avoid adding extra overhead into the common case (where those environment variables aren't used). Discussed with: csjp, secteam MFC after: 1 week
2005-03-24 10:12:29 +00:00
static bool dangerous_ld_env; /* True if environment variables have been
used to affect the libraries loaded */
Implement LD_BIND_NOT knob for rtld. From the manpage: When set to a nonempty string, prevents modifications of the PLT slots when doing bindings. As result, each call of the PLT-resolved function is resolved. In combination with debug output, this provides complete account of all bind actions at runtime. Same feature exists on Linux and Solaris. Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2017-03-15 21:11:57 +00:00
bool ld_bind_not; /* Disable PLT update */
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
static char *ld_bind_now; /* Environment variable for immediate binding */
static char *ld_debug; /* Environment variable for debugging */
static char *ld_library_path; /* Environment variable for search path */
Add the LD_LIBRARY_PATH_FDS environmental variable. This variable allows the loading of shared libraries via directory descriptors rather than via library paths. If LD_LIBRARY_PATH_FDS=3:4:12, the directories represented by file descriptors 3, 4 and 12 will searched for shared libraries before the normal path-based mechanisms are used. This allows us to execute unprivileged binaries from within a Capsicum sandbox even if they require shared libraries. Approved by: rwatson (mentor) Reviewed by: kib MFC after: 3 weeks Sponsored by: DARPA/AFRL
2014-06-20 17:08:32 +00:00
static char *ld_library_dirs; /* Environment variable for library descriptors */
Make LD_PRELOAD work for ELF.
1998-09-22 02:09:56 +00:00
static char *ld_preload; /* Environment variable for libraries to
load first */
Support for a new environment variable, LD_ELF_HINTS_PATH for overriding the rtld hints file. This environment variable would be unset if the process is considered as tainted with setuid/setgid. This feature gives a convenient way of using a custom set of shared library that is not located in the default location and switch back. Feature requested by: iXsystems Original patch by: John Hixson MFC after: 2 weeks
2009-03-23 16:49:00 +00:00
static char *ld_elf_hints_path; /* Environment variable for alternative hints path */
Add support for ldd.
1998-05-01 08:39:27 +00:00
static char *ld_tracing; /* Called from ldd to print libs */
Add various utrace's for use with ktrace to the ELF runtime linker. To activate the traces, set the LD_UTRACE (or LD_32_UTRACE) environment variable. This also includes code in kdump(8) to parse the traces. Reviewed by: kan, jdp MFC after: 2 weeks
2007-01-09 17:50:05 +00:00
static char *ld_utrace; /* Use utrace() to log events. */
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
static struct obj_entry_q obj_list; /* Queue of all loaded objects */
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
static Obj_Entry *obj_main; /* The main program shared object */
static Obj_Entry obj_rtld; /* The dynamic linker shared object */
Solve the dynamic linker's problems with multithreaded programs once and for all (I hope). Packages such as wine, JDK, and linuxthreads should no longer have any problems with re-entering the dynamic linker. This commit replaces the locking used in the dynamic linker with a new spinlock-based reader/writer lock implementation. Brian Fundakowski Feldman <green> argued for this from the very beginning, but it took me a long time to come around to his point of view. Spinlocks are the only kinds of locks that work with all thread packages. But on uniprocessor systems they can be inefficient, because while a contender for the lock is spinning the holder of the lock cannot make any progress toward releasing it. To alleviate this disadvantage I have borrowed a trick from Sleepycat's Berkeley DB implementation. When spinning for a lock, the requester does a nanosleep() call for 1 usec. each time around the loop. This will generally yield the CPU to other threads, allowing the lock holder to finish its business and release the lock. I chose 1 usec. as the minimum sleep which would with reasonable certainty not be rounded down to 0. The formerly machine-independent file "lockdflt.c" has been moved into the architecture-specific subdirectories by repository copy. It now contains the machine-dependent spinlocking code. For the spinlocks I used the very nifty "simple, non-scalable reader-preference lock" which I found at <http://www.cs.rochester.edu/u/scott/synchronization/pseudocode/rw.html> on all CPUs except the 80386 (the specific CPU model, not the architecture). The 80386 CPU doesn't support the necessary "cmpxchg" instruction, so on that CPU a simple exclusive test-and-set lock is used instead. 80386 CPUs are detected at initialization time by trying to execute "cmpxchg" and catching the resulting SIGILL signal. To reduce contention for the locks, I have revamped a couple of key data structures, permitting all common operations to be done under non-exclusive (reader) locking. The only operations that require exclusive locking now are the rare intrusive operations such as dlopen() and dlclose(). The dllockinit() interface is now deprecated. It still exists, but only as a do-nothing stub. I plan to remove it as soon as is reasonably possible. (From the very beginning it was clearly labeled as experimental and subject to change.) As far as I know, only the linuxthreads port uses dllockinit(). This interface turned out to have several problems. As one example, when the dynamic linker called a client-supplied locking function, that function sometimes needed lazy binding, causing re-entry into the dynamic linker and a big looping mess. And in any case, it turned out to be too burdensome to require threads packages to register themselves with the dynamic linker.
2000-07-08 04:10:38 +00:00
static unsigned int obj_count; /* Number of objects in obj_list */
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
static unsigned int obj_loads; /* Number of loads of objects (gen count) */
Revamp the symbol lookup algorithm to cope better with objects loaded separately by dlopen that have global symbols with identical names. Viewing each dlopened object as a DAG which is linked by its DT_NEEDED entries in the dynamic table, the search order is as follows: * If the referencing object was linked with -Bsymbolic, search it internally. * Search all dlopened DAGs containing the referencing object. * Search all objects loaded at program start up. * Search all objects which were dlopened() using the RTLD_GLOBAL flag (which is now supported too). The search terminates as soon as a strong definition is found. Lacking that, the first weak definition is used. These rules match those of Solaris, as best I could determine them from its vague manual pages and the results of experiments I performed. PR: misc/12438
1999-08-30 01:48:19 +00:00
static Objlist list_global = /* Objects dlopened with RTLD_GLOBAL */
STAILQ_HEAD_INITIALIZER(list_global);
static Objlist list_main = /* Objects loaded at program startup */
STAILQ_HEAD_INITIALIZER(list_main);
Revamp the code that calls shared libraries' init and fini functions. Formerly the init functions were called in the opposite of the order in which libraries were loaded, and libraries were loaded according to a breadth-first traversal of the dependency graph. That ordering came from SVR4.0, and it was easy to implement but not always sensible. Now we do a depth-first walk over the dependency graph and call the init functions in an order such that each shared object's needed objects are initialized before the shared object itself. At the same time we build a list of finalization (fini) functions in the opposite order, to guarantee correct C++ destructor ordering whenever possible. (It may not be possible if dlopen and dlclose are used in strange ways, but we come as close as one can come.) The need for this renovation has become apparent as more programs have started using multithreading. The multithreaded C library libc_r requires initialization, whereas the standard libc does not. Since virtually every other object depends on the C library, it is important that it get initialized first.
2000-07-26 04:24:40 +00:00
static Objlist list_fini = /* Objects needing fini() calls */
STAILQ_HEAD_INITIALIZER(list_fini);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
Check for undefined weak symbols during PLT binding on powerpc64, and do not attempt to copy NULL function descriptors. This fixes LD_BIND_NOW on powerpc64 after r211706. Reviewed by: kib
2010-09-12 17:04:51 +00:00
Elf_Sym sym_zero; /* For resolving undefined weak refs. */
Resolve undefined weak references to a value of 0. This solves the "__deregister_frame_info" problem that was seen when combining a program linked using the old gcc with shared libraries that were built using egcs.
1999-04-05 02:36:40 +00:00
Pass two pointer parameters to the r_debug_state() hook function, thus allowing a debugger or other trace tool to easily grab the addresses of the needed structures off the stack. This change is transparent to gdb, which locates the link_map list and transfers it to debugger memory for comparison purposes. A sample program will be committed showing how this can be used. Reviewed by: John Polstra <jdp@FreeBSD.org>
2000-08-26 05:13:29 +00:00
#define GDB_STATE(s,m) r_debug.r_state = s; r_debug_state(&r_debug,m);
Add GDB support. The method and some of the code came from NetBSD's elf runtime linker.
1998-04-30 07:48:02 +00:00
Add alpha support. Submitted by: John Birrell <jb@cimlogic.com.au> (with extra hacks by me) Obtained from: Probably NetBSD
1998-09-04 19:03:57 +00:00
extern Elf_Dyn _DYNAMIC;
Eliminate all machine-dependent code from the main source body and the Makefile, and move it down into the architecture-specific subdirectories. Eliminate an asm() statement for the i386. Make the dynamic linker work if it is built as an executable instead of as a shared library. See i386/Makefile.inc to find out how to do it. Note, this change is not enabled and it might never be enabled. But it might be useful in the future. Building the dynamic linker as an executable should make it start up faster, because it won't have any relocations. But in practice I suspect the difference is negligible.
1999-04-09 00:28:43 +00:00
#pragma weak _DYNAMIC
Resolve undefined weak references to a value of 0. This solves the "__deregister_frame_info" problem that was seen when combining a program linked using the old gcc with shared libraries that were built using egcs.
1999-04-05 02:36:40 +00:00
Change compiler setting to make default visibility of the symbols for rtld on x86 to be hidden. This is a micro-optimization, which allows intrinsic references inside rtld to be handled without indirection through PLT. The visibility of rtld symbols for other objects in the symbol namespace is controlled by a version script. Reviewed by: kan, jilles Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2015-03-29 18:53:21 +00:00
int dlclose(void *) __exported;
char *dlerror(void) __exported;
void *dlopen(const char *, int) __exported;
void *fdlopen(int, int) __exported;
void *dlsym(void *, const char *) __exported;
dlfunc_t dlfunc(void *, const char *) __exported;
void *dlvsym(void *, const char *, const char *) __exported;
int dladdr(const void *, Dl_info *) __exported;
void dllockinit(void *, void *(*)(void *), void (*)(void *), void (*)(void *),
void (*)(void *), void (*)(void *), void (*)(void *)) __exported;
int dlinfo(void *, int , void *) __exported;
int dl_iterate_phdr(__dl_iterate_hdr_callback, void *) __exported;
int _rtld_addr_phdr(const void *, struct dl_phdr_info *) __exported;
int _rtld_get_stack_prot(void) __exported;
int _rtld_is_dlopened(void *) __exported;
void _rtld_error(const char *, ...) __exported;
Before calling mmap() on a shared library's text and data sections, rtld first calls mmap() with the arguments PROT_NONE and MAP_ANON to reserve a single, contiguous range of virtual addresses for the entire shared library. Later, rtld calls mmap() with the the shared library's file descriptor and the argument MAP_FIXED to place the text and data sections within the reserved range. The rationale for mapping shared libraries in this way is explained in the commit message for Revision 190885. However, this approach does have an unintended, negative consequence. Since the first call to mmap() specifies MAP_ANON and not the shared library's file descriptor, the kernel has no idea what alignment the vm object backing the file prefers. As a result, the reserved range's alignment is unlikely to be the same as the vm object's, and so mapping with superpages becomes impossible. To address this problem, this revision adds the argument MAP_ALIGNED_SUPER to the first call to mmap() if the text section is larger than the smallest superpage size. To determine if the text section is larger than the smallest superpage size, rtld must always fetch the page size information. As a result, the private code for fetching the base page size in rtld's builtin malloc is redundant. Eliminate it. Requested by: kib Tested by: zbb (on arm) Reviewed by: kib (an earlier version) Discussed with: jhb
2014-04-11 16:55:25 +00:00
int npagesizes, osreldate;
size_t *pagesizes;
Use the newly provided aux vectors to get pagesize and osreldate information. Use local version of getpagesize(), rtld_getpagesize() in private allocator. Override the __getosreldate() previously fetched from libc_pic.a with local version that uses aux value if present. Note that __getosreldate() is used by rtld indirectly, by mmap(2) libc wrapper. To be able to utilize aux, split digest_dynamic() for use by init_rtld() into two parts, where the first one does not call malloc(), and the second part uses it. init_rtld() is able to initialize global variables before digest_dynamic2() calls. In particular, pagesize and osreldate are set up from the aux values. Now, rtld avoids (two) sysctl calls in startup. Tested by: marius (sparc64) MFC after: 1 month
2010-08-17 09:05:39 +00:00
Rtld on diet part 1: Provide rtld-private implementations of __stack_chk_guard, __stack_chk_fail() and __chk_fail() symbols, to be used by functions linked from libc_pic.a. This avoids use of libc stack_protector.c, which pulls in syslog(3) and stdio as dependency. Also, do initialize rtld-private copy __stack_chk_guard, previously libc-provided one was not initialized, since we do not call rtld object _init() methods. Reviewed by: kan MFC after: 3 weeks
2012-03-12 12:15:47 +00:00
long __stack_chk_guard[8] = {0, 0, 0, 0, 0, 0, 0, 0};
When loading dso without PT_GNU_STACK phdr, only call __pthread_map_stacks_exec() on architectures that allow executable stacks. Reported and tested by: marcel (ia64)
2011-01-25 21:12:31 +00:00
static int stack_prot = PROT_READ | PROT_WRITE | RTLD_DEFAULT_STACK_EXEC;
In rtld, read the initial stack access mode from AT_STACKPROT as set by kernel, and parse PT_GNU_STACK phdr from linked and loaded dsos. If the loaded dso requires executable stack, as specified by PF_X bit of p_flags of PT_GNU_STACK phdr, but current stack protection does not permit execution, the __pthread_map_stacks_exec symbol is looked up and called. It should be implemented in libc or threading library and change the protection mode of all thread stacks to be executable. Provide a private interface _rtld_get_stack_prot() to export the stack access mode as calculated by rtld. Reviewed by: kan
2011-01-08 17:11:49 +00:00
static int max_stack_flags;
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
/*
* Global declarations normally provided by crt1. The dynamic linker is
Revamp the mechanism for enumerating and calling shared objects' init and fini functions. Now the code is very careful to hold no locks when calling these functions. Thus the dynamic linker cannot be re-entered with a lock already held. Remove the tolerance for recursive locking that I added in revision 1.2 of dllockinit.c. Recursive locking shouldn't happen any more. Mozilla and JDK users: I'd appreciate confirmation that things still work right (or at least the same) with these changes.
2000-01-09 21:13:48 +00:00
* not built with crt1, so we have to provide them ourselves.
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
*/
char *__progname;
char **environ;
Add support for preinit, init and fini arrays. Some ABIs, in particular on ARM, do require working init arrays. Traditional FreeBSD crt1 calls _init and _fini of the binary, instead of allowing runtime linker to arrange the calls. This was probably done to have the same crt code serve both statically and dynamically linked binaries. Since ABI mandates that first is called preinit array functions, then init, and then init array functions, the init have to be called from rtld now. To provide binary compatibility to old FreeBSD crt1, which calls _init itself, rtld only calls intializers and finalizers for main binary if binary has a note indicating that new crt was used for linking. Add parsing of ELF notes to rtld, and cache p_osrel value since we parsed it anyway. The patch is inspired by init_array support for DragonflyBSD, written by John Marino. Reviewed by: kan Tested by: andrew (arm, previous version), flo (sparc64, previous version) MFC after: 3 weeks
2012-03-11 20:03:09 +00:00
/*
* Used to pass argc, argv to init functions.
*/
int main_argc;
char **main_argv;
Add support for Thread Local Storage.
2004-08-03 08:51:00 +00:00
/*
* Globals to control TLS allocation.
*/
size_t tls_last_offset; /* Static TLS offset of last module */
size_t tls_last_size; /* Static TLS size of last module */
size_t tls_static_space; /* Static TLS space allocated */
For variant II static TLS, properly align tls segments. Pre-calculate the max required alignment for the static tls segments, and honor it when carving the pieces for next module, from the static space. Use aligned allocator to get properly-aligned dynamic blocks. Reported by: dt71@gmx.com Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 1 week
2013-12-06 21:39:45 +00:00
size_t tls_static_max_align;
Add support for Thread Local Storage.
2004-08-03 08:51:00 +00:00
int tls_dtv_generation = 1; /* Used to detect when dtv size changes */
int tls_max_index = 1; /* Largest module index allocated */
Import the DragonFly BSD commit 4f0bc915b65fcf5a23214f6d221d65c80be68ad4 by John Marino <draco@marino.st>, with the following (edited) commit message Date: Sat, 24 Mar 2012 06:40:50 +0100 Subject: [PATCH 1/1] rtld: Implement DT_RUNPATH and -z nodefaultlib DT_RUNPATH is incorrectly being considered as an alias of DT_RPATH. The purpose of DT_RUNPATH is to have two different types of rpath: one that can be overridden by the environment variable LD_LIBRARY_PATH and one that can't. With the currently implementation, LD_LIBRARY_PATH will always trump any embedded rpath or runpath tags. Current path search order by rtld: ================================== LD_LIBRARY_PATH DT_RPATH / DT_RUNPATH (always the same) ldconfig hints file (default: /var/run/ld-elf.so.hints) /usr/lib New path search order by rtld: ============================== DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file /usr/lib The new path search matches how the linux runtime loader works. The other major added feature is support for linker flag "-z nodefaultlib". When this flag is passed to the linker, rtld will skip all references to the standard library search path ("/usr/lib" in this case but it could handle more color delimited paths) except in DT_RPATH and DT_RUNPATH. New path search order by rtld with -z nodefaultlib flag set: ============================================================ DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file (skips all references to /usr/lib) FreeBSD notes: - we fixed some bugs which were submitted to DragonFly and merged there as commit 1ff8a2bd3eb6e5587174c6a983303ea3a79e0002; - we added LD_LIBRARY_PATH_RPATH environment variable to switch to the previous behaviour of considering DT_RPATH a synonym for DT_RUNPATH; - the FreeBSD default search path is /lib:/usr/lib and not /usr/lib. Reviewed by: kan MFC after: 1 month MFC note: flip the ld_library_path_rpath default value for stable/9
2012-07-15 10:53:48 +00:00
bool ld_library_path_rpath = false;
Rather than using the #define for path names, indirect through a char * variable that could change for different executable types detected.
2015-10-31 04:39:55 +00:00
/*
* Globals for path names, and such
*/
Use variable names that aren't as prone to dyslexic confusion. Suggested by: jmallet@
2015-12-18 05:45:49 +00:00
char *ld_elf_hints_default = _PATH_ELF_HINTS;
Rather than using the #define for path names, indirect through a char * variable that could change for different executable types detected.
2015-10-31 04:39:55 +00:00
char *ld_path_libmap_conf = _PATH_LIBMAP_CONF;
char *ld_path_rtld = _PATH_RTLD;
char *ld_standard_library_path = STANDARD_LIBRARY_PATH;
char *ld_env_prefix = LD_;
Solve the dynamic linker's problems with multithreaded programs once and for all (I hope). Packages such as wine, JDK, and linuxthreads should no longer have any problems with re-entering the dynamic linker. This commit replaces the locking used in the dynamic linker with a new spinlock-based reader/writer lock implementation. Brian Fundakowski Feldman <green> argued for this from the very beginning, but it took me a long time to come around to his point of view. Spinlocks are the only kinds of locks that work with all thread packages. But on uniprocessor systems they can be inefficient, because while a contender for the lock is spinning the holder of the lock cannot make any progress toward releasing it. To alleviate this disadvantage I have borrowed a trick from Sleepycat's Berkeley DB implementation. When spinning for a lock, the requester does a nanosleep() call for 1 usec. each time around the loop. This will generally yield the CPU to other threads, allowing the lock holder to finish its business and release the lock. I chose 1 usec. as the minimum sleep which would with reasonable certainty not be rounded down to 0. The formerly machine-independent file "lockdflt.c" has been moved into the architecture-specific subdirectories by repository copy. It now contains the machine-dependent spinlocking code. For the spinlocks I used the very nifty "simple, non-scalable reader-preference lock" which I found at <http://www.cs.rochester.edu/u/scott/synchronization/pseudocode/rw.html> on all CPUs except the 80386 (the specific CPU model, not the architecture). The 80386 CPU doesn't support the necessary "cmpxchg" instruction, so on that CPU a simple exclusive test-and-set lock is used instead. 80386 CPUs are detected at initialization time by trying to execute "cmpxchg" and catching the resulting SIGILL signal. To reduce contention for the locks, I have revamped a couple of key data structures, permitting all common operations to be done under non-exclusive (reader) locking. The only operations that require exclusive locking now are the rare intrusive operations such as dlopen() and dlclose(). The dllockinit() interface is now deprecated. It still exists, but only as a do-nothing stub. I plan to remove it as soon as is reasonably possible. (From the very beginning it was clearly labeled as experimental and subject to change.) As far as I know, only the linuxthreads port uses dllockinit(). This interface turned out to have several problems. As one example, when the dynamic linker called a client-supplied locking function, that function sometimes needed lazy binding, causing re-entry into the dynamic linker and a big looping mess. And in any case, it turned out to be too burdensome to require threads packages to register themselves with the dynamic linker.
2000-07-08 04:10:38 +00:00
/*
* Fill in a DoneList with an allocation large enough to hold all of
* the currently-loaded objects. Keep this as a macro since it calls
* alloca and we want that to occur within the scope of the caller.
*/
#define donelist_init(dlp) \
((dlp)->objs = alloca(obj_count * sizeof (dlp)->objs[0]), \
assert((dlp)->objs != NULL), \
(dlp)->num_alloc = obj_count, \
(dlp)->num_used = 0)
Add various utrace's for use with ktrace to the ELF runtime linker. To activate the traces, set the LD_UTRACE (or LD_32_UTRACE) environment variable. This also includes code in kdump(8) to parse the traces. Reviewed by: kan, jdp MFC after: 2 weeks
2007-01-09 17:50:05 +00:00
#define LD_UTRACE(e, h, mb, ms, r, n) do { \
if (ld_utrace != NULL) \
ld_utrace_log(e, h, mb, ms, r, n); \
} while (0)
static void
ld_utrace_log(int event, void *handle, void *mapbase, size_t mapsize,
int refcnt, const char *name)
{
struct utrace_rtld ut;
Move defines common between rtld and libsysdecode into the header, instead of copying inline into sources. Sponsored by: The FreeBSD Foundation MFC after: 1 week
2016-08-12 19:31:41 +00:00
static const char rtld_utrace_sig[RTLD_UTRACE_SIG_SZ] = RTLD_UTRACE_SIG;
Add various utrace's for use with ktrace to the ELF runtime linker. To activate the traces, set the LD_UTRACE (or LD_32_UTRACE) environment variable. This also includes code in kdump(8) to parse the traces. Reviewed by: kan, jdp MFC after: 2 weeks
2007-01-09 17:50:05 +00:00
Move defines common between rtld and libsysdecode into the header, instead of copying inline into sources. Sponsored by: The FreeBSD Foundation MFC after: 1 week
2016-08-12 19:31:41 +00:00
memcpy(ut.sig, rtld_utrace_sig, sizeof(ut.sig));
Add various utrace's for use with ktrace to the ELF runtime linker. To activate the traces, set the LD_UTRACE (or LD_32_UTRACE) environment variable. This also includes code in kdump(8) to parse the traces. Reviewed by: kan, jdp MFC after: 2 weeks
2007-01-09 17:50:05 +00:00
ut.event = event;
ut.handle = handle;
ut.mapbase = mapbase;
ut.mapsize = mapsize;
ut.refcnt = refcnt;
bzero(ut.name, sizeof(ut.name));
if (name)
strlcpy(ut.name, name, sizeof(ut.name));
utrace(&ut, sizeof(ut));
}
If md_exec_hook is defined, provide a way to create the strings for the environment variables we look up at runtime. Otherwise, there's no way they will change, optimize it at compile time. Differential Review: https://reviews.freebsd.org/D2718
2016-01-03 04:32:13 +00:00
#ifdef RTLD_VARIANT_ENV_NAMES
/*
* construct the env variable based on the type of binary that's
* running.
*/
static inline const char *
_LD(const char *var)
{
static char buffer[128];
strlcpy(buffer, ld_env_prefix, sizeof(buffer));
strlcat(buffer, var, sizeof(buffer));
return (buffer);
}
#else
#define _LD(x) LD_ x
#endif
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
/*
* Main entry point for dynamic linking. The first argument is the
* stack pointer. The stack is expected to be laid out as described
* in the SVR4 ABI specification, Intel 386 Processor Supplement.
* Specifically, the stack pointer points to a word containing
* ARGC. Following that in the stack is a null-terminated sequence
* of pointers to argument strings. Then comes a null-terminated
* sequence of pointers to environment strings. Finally, there is a
* sequence of "auxiliary vector" entries.
*
* The second argument points to a place to store the dynamic linker's
Add alpha support. Submitted by: John Birrell <jb@cimlogic.com.au> (with extra hacks by me) Obtained from: Probably NetBSD
1998-09-04 19:03:57 +00:00
* exit procedure pointer and the third to a place to store the main
* program's object.
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
*
* The return value is the main program's entry point.
*/
func_ptr_type
Add alpha support. Submitted by: John Birrell <jb@cimlogic.com.au> (with extra hacks by me) Obtained from: Probably NetBSD
1998-09-04 19:03:57 +00:00
_rtld(Elf_Addr *sp, func_ptr_type *exit_proc, Obj_Entry **objp)
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
{
Make ld-elf.so.1 directly executable. Check if passed phdr is actually phdr of the interpreter itself, and decide that this is the case of direct execution. In this case, the binary to activate is specified in the argv[1]. After opening it, shift down on-stack structure with argv, env and aux vectors to emulate execution of the binary and not of the interpreter. Reviewed by: emaste Sponsored by: The FreeBSD Foundation MFC after: 2 weeks Differential revision: https://reviews.freebsd.org/D10701
2017-05-15 18:48:58 +00:00
Elf_Auxinfo *aux, *auxp, *auxpf, *aux_info[AT_COUNT];
Add support for Thread Local Storage.
2004-08-03 08:51:00 +00:00
Objlist_Entry *entry;
In _rtld(), reorder local declarations to compact the block and partially sort them by style(9). Move locals declarations from nested blocks into the block at function start. Discussed with: emaste MFC after: 1 week
2017-05-13 18:59:27 +00:00
Obj_Entry *last_interposer, *obj, *preload_tail;
const Elf_Phdr *phdr;
Revamp the code that calls shared libraries' init and fini functions. Formerly the init functions were called in the opposite of the order in which libraries were loaded, and libraries were loaded according to a breadth-first traversal of the dependency graph. That ordering came from SVR4.0, and it was easy to implement but not always sensible. Now we do a depth-first walk over the dependency graph and call the init functions in an order such that each shared object's needed objects are initialized before the shared object itself. At the same time we build a list of finalization (fini) functions in the opposite order, to guarantee correct C++ destructor ordering whenever possible. (It may not be possible if dlopen and dlclose are used in strange ways, but we come as close as one can come.) The need for this renovation has become apparent as more programs have started using multithreading. The multithreaded C library libc_r requires initialization, whereas the standard libc does not. Since virtually every other object depends on the C library, it is important that it get initialized first.
2000-07-26 04:24:40 +00:00
Objlist initlist;
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
RtldLockState lockstate;
Pretend that there is some security when executing in direct mode. Do not allow direct exec if we the process is suid. Try to follow Unix permission checks for DACs, ignore ACLs. Reviewed by: emaste Sponsored by: The FreeBSD Foundation MFC after: 2 weeks Differential revision: https://reviews.freebsd.org/D10750
2017-05-16 19:53:38 +00:00
struct stat st;
Make ld-elf.so.1 directly executable. Check if passed phdr is actually phdr of the interpreter itself, and decide that this is the case of direct execution. In this case, the binary to activate is specified in the argv[1]. After opening it, shift down on-stack structure with argv, env and aux vectors to emulate execution of the binary and not of the interpreter. Reviewed by: emaste Sponsored by: The FreeBSD Foundation MFC after: 2 weeks Differential revision: https://reviews.freebsd.org/D10701
2017-05-15 18:48:58 +00:00
Elf_Addr *argcp;
char **argv, *argv0, **env, **envp, *kexecpath, *library_path_rpath;
In _rtld(), reorder local declarations to compact the block and partially sort them by style(9). Move locals declarations from nested blocks into the block at function start. Discussed with: emaste MFC after: 1 week
2017-05-13 18:59:27 +00:00
caddr_t imgentry;
char buf[MAXPATHLEN];
Allow rtld direct-exec to take a file descriptor. When executing rtld directly, allow a file descriptor to be explicitly specified rather than opened from the given path. This, together with the LD_LIBRARY_PATH_FDS environment variable, allows dynamically-linked applications to be executed from within capability mode. Also add some rudimentary argument parsing (without pulling in getopt or the like) to accept this file descriptor, a help (-h) option and a basic usage string. Reviewed by: kib Sponsored by: NSERC, RDC Differential Revision: https://reviews.freebsd.org/D10751
2017-05-17 22:51:28 +00:00
int argc, fd, i, mib[2], phnum, rtld_argc;
Rtld on diet part 1: Provide rtld-private implementations of __stack_chk_guard, __stack_chk_fail() and __chk_fail() symbols, to be used by functions linked from libc_pic.a. This avoids use of libc stack_protector.c, which pulls in syslog(3) and stdio as dependency. Also, do initialize rtld-private copy __stack_chk_guard, previously libc-provided one was not initialized, since we do not call rtld object _init() methods. Reviewed by: kan MFC after: 3 weeks
2012-03-12 12:15:47 +00:00
size_t len;
Allow rtld direct-exec to take a file descriptor. When executing rtld directly, allow a file descriptor to be explicitly specified rather than opened from the given path. This, together with the LD_LIBRARY_PATH_FDS environment variable, allows dynamically-linked applications to be executed from within capability mode. Also add some rudimentary argument parsing (without pulling in getopt or the like) to accept this file descriptor, a help (-h) option and a basic usage string. Reviewed by: kib Sponsored by: NSERC, RDC Differential Revision: https://reviews.freebsd.org/D10751
2017-05-17 22:51:28 +00:00
bool dir_enable, explicit_fd, search_in_path;
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
/*
* On entry, the dynamic linker itself has not been relocated yet.
* Be very careful not to reference any global data until after
* init_rtld has returned. It is OK to reference file-scope statics
* and string constants, and to call static and global functions.
*/
/* Find the auxiliary vector on the stack. */
Make ld-elf.so.1 directly executable. Check if passed phdr is actually phdr of the interpreter itself, and decide that this is the case of direct execution. In this case, the binary to activate is specified in the argv[1]. After opening it, shift down on-stack structure with argv, env and aux vectors to emulate execution of the binary and not of the interpreter. Reviewed by: emaste Sponsored by: The FreeBSD Foundation MFC after: 2 weeks Differential revision: https://reviews.freebsd.org/D10701
2017-05-15 18:48:58 +00:00
argcp = sp;
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
argc = *sp++;
argv = (char **) sp;
sp += argc + 1; /* Skip over arguments and NULL terminator */
env = (char **) sp;
while (*sp++ != 0) /* Skip over environment, and NULL terminator */
;
Add alpha support. Submitted by: John Birrell <jb@cimlogic.com.au> (with extra hacks by me) Obtained from: Probably NetBSD
1998-09-04 19:03:57 +00:00
aux = (Elf_Auxinfo *) sp;
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
/* Digest the auxiliary vector. */
for (i = 0; i < AT_COUNT; i++)
aux_info[i] = NULL;
for (auxp = aux; auxp->a_type != AT_NULL; auxp++) {
if (auxp->a_type < AT_COUNT)
aux_info[auxp->a_type] = auxp;
}
/* Initialize and relocate ourselves. */
assert(aux_info[AT_BASE] != NULL);
Use the newly provided aux vectors to get pagesize and osreldate information. Use local version of getpagesize(), rtld_getpagesize() in private allocator. Override the __getosreldate() previously fetched from libc_pic.a with local version that uses aux value if present. Note that __getosreldate() is used by rtld indirectly, by mmap(2) libc wrapper. To be able to utilize aux, split digest_dynamic() for use by init_rtld() into two parts, where the first one does not call malloc(), and the second part uses it. init_rtld() is able to initialize global variables before digest_dynamic2() calls. In particular, pagesize and osreldate are set up from the aux values. Now, rtld avoids (two) sysctl calls in startup. Tested by: marius (sparc64) MFC after: 1 month
2010-08-17 09:05:39 +00:00
init_rtld((caddr_t) aux_info[AT_BASE]->a_un.a_ptr, aux_info);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
__progname = obj_rtld.path;
Change many asserts into normal errors. They were all for conditions caused by invalid shared objects rather than by internal errors. Enable format string mismatch checking for _rtld_error().
1999-07-18 00:02:19 +00:00
argv0 = argv[0] != NULL ? argv[0] : "(null)";
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
environ = env;
Add support for preinit, init and fini arrays. Some ABIs, in particular on ARM, do require working init arrays. Traditional FreeBSD crt1 calls _init and _fini of the binary, instead of allowing runtime linker to arrange the calls. This was probably done to have the same crt code serve both statically and dynamically linked binaries. Since ABI mandates that first is called preinit array functions, then init, and then init array functions, the init have to be called from rtld now. To provide binary compatibility to old FreeBSD crt1, which calls _init itself, rtld only calls intializers and finalizers for main binary if binary has a note indicating that new crt was used for linking. Add parsing of ELF notes to rtld, and cache p_osrel value since we parsed it anyway. The patch is inspired by init_array support for DragonflyBSD, written by John Marino. Reviewed by: kan Tested by: andrew (arm, previous version), flo (sparc64, previous version) MFC after: 3 weeks
2012-03-11 20:03:09 +00:00
main_argc = argc;
main_argv = argv;
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
Properly handle absent AT_CANARY aux entry. Submitted by: Andrey Zonov <andrey zonov org> MFC after: 3 days
2012-04-05 18:47:54 +00:00
if (aux_info[AT_CANARY] != NULL &&
aux_info[AT_CANARY]->a_un.a_ptr != NULL) {
Rtld on diet part 1: Provide rtld-private implementations of __stack_chk_guard, __stack_chk_fail() and __chk_fail() symbols, to be used by functions linked from libc_pic.a. This avoids use of libc stack_protector.c, which pulls in syslog(3) and stdio as dependency. Also, do initialize rtld-private copy __stack_chk_guard, previously libc-provided one was not initialized, since we do not call rtld object _init() methods. Reviewed by: kan MFC after: 3 weeks
2012-03-12 12:15:47 +00:00
i = aux_info[AT_CANARYLEN]->a_un.a_val;
if (i > sizeof(__stack_chk_guard))
i = sizeof(__stack_chk_guard);
memcpy(__stack_chk_guard, aux_info[AT_CANARY]->a_un.a_ptr, i);
} else {
mib[0] = CTL_KERN;
mib[1] = KERN_ARND;
len = sizeof(__stack_chk_guard);
if (sysctl(mib, 2, __stack_chk_guard, &len, NULL, 0) == -1 ||
len != sizeof(__stack_chk_guard)) {
/* If sysctl was unsuccessful, use the "terminator canary". */
((unsigned char *)(void *)__stack_chk_guard)[0] = 0;
((unsigned char *)(void *)__stack_chk_guard)[1] = 0;
((unsigned char *)(void *)__stack_chk_guard)[2] = '\n';
((unsigned char *)(void *)__stack_chk_guard)[3] = 255;
}
}
- use issetugid() - be paranoid about honoring LD_LIBMAP_DISABLE. Suggested by: rwatson
2003-05-31 15:24:29 +00:00
trust = !issetugid();
Use the environment variable LD_LIBMAP_DISABLE to disable libmap.conf(5) functionality.
2003-05-31 14:45:11 +00:00
Restore ABI variants now that ldconfig groks -soft. In addition, as a transition mechanism, if we don't have /usr/libsoft, assume that soft float ABI binaries are the default, so treat them as default binaries. When we've fully transitioned, it will make no sense to do this stat, and it will be removed.
2016-01-18 21:40:20 +00:00
md_abi_variant_hook(aux_info);
Create a generalized exec hook that different architectures can hook into if they need to, but default to no action. Differential Review: https://reviews.freebsd.org/D2718
2016-01-03 04:32:02 +00:00
Make ld-elf.so.1 directly executable. Check if passed phdr is actually phdr of the interpreter itself, and decide that this is the case of direct execution. In this case, the binary to activate is specified in the argv[1]. After opening it, shift down on-stack structure with argv, env and aux vectors to emulate execution of the binary and not of the interpreter. Reviewed by: emaste Sponsored by: The FreeBSD Foundation MFC after: 2 weeks Differential revision: https://reviews.freebsd.org/D10701
2017-05-15 18:48:58 +00:00
fd = -1;
if (aux_info[AT_EXECFD] != NULL) {
fd = aux_info[AT_EXECFD]->a_un.a_val;
} else {
assert(aux_info[AT_PHDR] != NULL);
phdr = (const Elf_Phdr *)aux_info[AT_PHDR]->a_un.a_ptr;
if (phdr == obj_rtld.phdr) {
Pretend that there is some security when executing in direct mode. Do not allow direct exec if we the process is suid. Try to follow Unix permission checks for DACs, ignore ACLs. Reviewed by: emaste Sponsored by: The FreeBSD Foundation MFC after: 2 weeks Differential revision: https://reviews.freebsd.org/D10750
2017-05-16 19:53:38 +00:00
if (!trust) {
rtld_printf("Tainted process refusing to run binary %s\n",
argv0);
rtld_die();
}
Make ld-elf.so.1 directly executable. Check if passed phdr is actually phdr of the interpreter itself, and decide that this is the case of direct execution. In this case, the binary to activate is specified in the argv[1]. After opening it, shift down on-stack structure with argv, env and aux vectors to emulate execution of the binary and not of the interpreter. Reviewed by: emaste Sponsored by: The FreeBSD Foundation MFC after: 2 weeks Differential revision: https://reviews.freebsd.org/D10701
2017-05-15 18:48:58 +00:00
dbg("opening main program in direct exec mode");
if (argc >= 2) {
Allow rtld direct-exec to take a file descriptor. When executing rtld directly, allow a file descriptor to be explicitly specified rather than opened from the given path. This, together with the LD_LIBRARY_PATH_FDS environment variable, allows dynamically-linked applications to be executed from within capability mode. Also add some rudimentary argument parsing (without pulling in getopt or the like) to accept this file descriptor, a help (-h) option and a basic usage string. Reviewed by: kib Sponsored by: NSERC, RDC Differential Revision: https://reviews.freebsd.org/D10751
2017-05-17 22:51:28 +00:00
rtld_argc = parse_args(argv, argc, &search_in_path, &fd);
argv0 = argv[rtld_argc];
explicit_fd = (fd != -1);
if (!explicit_fd)
For ld.so direct execution mode, implement -p option: search for the binary in $PATH. Sponsored by: The FreeBSD Foundation MFC after: 2 weeks Differential revision: https://reviews.freebsd.org/D10790
2017-05-23 10:00:52 +00:00
fd = open_binary_fd(argv0, search_in_path);
Pretend that there is some security when executing in direct mode. Do not allow direct exec if we the process is suid. Try to follow Unix permission checks for DACs, ignore ACLs. Reviewed by: emaste Sponsored by: The FreeBSD Foundation MFC after: 2 weeks Differential revision: https://reviews.freebsd.org/D10750
2017-05-16 19:53:38 +00:00
if (fstat(fd, &st) == -1) {
Allow rtld direct-exec to take a file descriptor. When executing rtld directly, allow a file descriptor to be explicitly specified rather than opened from the given path. This, together with the LD_LIBRARY_PATH_FDS environment variable, allows dynamically-linked applications to be executed from within capability mode. Also add some rudimentary argument parsing (without pulling in getopt or the like) to accept this file descriptor, a help (-h) option and a basic usage string. Reviewed by: kib Sponsored by: NSERC, RDC Differential Revision: https://reviews.freebsd.org/D10751
2017-05-17 22:51:28 +00:00
_rtld_error("failed to fstat FD %d (%s): %s", fd,
explicit_fd ? "user-provided descriptor" : argv0,
Pretend that there is some security when executing in direct mode. Do not allow direct exec if we the process is suid. Try to follow Unix permission checks for DACs, ignore ACLs. Reviewed by: emaste Sponsored by: The FreeBSD Foundation MFC after: 2 weeks Differential revision: https://reviews.freebsd.org/D10750
2017-05-16 19:53:38 +00:00
rtld_strerror(errno));
rtld_die();
}
/*
* Rough emulation of the permission checks done by
* execve(2), only Unix DACs are checked, ACLs are
* ignored. Preserve the semantic of disabling owner
* to execute if owner x bit is cleared, even if
* others x bit is enabled.
* mmap(2) does not allow to mmap with PROT_EXEC if
* binary' file comes from noexec mount. We cannot
* set VV_TEXT on the binary.
*/
dir_enable = false;
if (st.st_uid == geteuid()) {
if ((st.st_mode & S_IXUSR) != 0)
dir_enable = true;
} else if (st.st_gid == getegid()) {
if ((st.st_mode & S_IXGRP) != 0)
dir_enable = true;
} else if ((st.st_mode & S_IXOTH) != 0) {
dir_enable = true;
}
if (!dir_enable) {
rtld_printf("No execute permission for binary %s\n",
argv0);
rtld_die();
}
Make ld-elf.so.1 directly executable. Check if passed phdr is actually phdr of the interpreter itself, and decide that this is the case of direct execution. In this case, the binary to activate is specified in the argv[1]. After opening it, shift down on-stack structure with argv, env and aux vectors to emulate execution of the binary and not of the interpreter. Reviewed by: emaste Sponsored by: The FreeBSD Foundation MFC after: 2 weeks Differential revision: https://reviews.freebsd.org/D10701
2017-05-15 18:48:58 +00:00
/*
* For direct exec mode, argv[0] is the interpreter
Allow rtld direct-exec to take a file descriptor. When executing rtld directly, allow a file descriptor to be explicitly specified rather than opened from the given path. This, together with the LD_LIBRARY_PATH_FDS environment variable, allows dynamically-linked applications to be executed from within capability mode. Also add some rudimentary argument parsing (without pulling in getopt or the like) to accept this file descriptor, a help (-h) option and a basic usage string. Reviewed by: kib Sponsored by: NSERC, RDC Differential Revision: https://reviews.freebsd.org/D10751
2017-05-17 22:51:28 +00:00
* name, we must remove it and shift arguments left
* before invoking binary main. Since stack layout
Make ld-elf.so.1 directly executable. Check if passed phdr is actually phdr of the interpreter itself, and decide that this is the case of direct execution. In this case, the binary to activate is specified in the argv[1]. After opening it, shift down on-stack structure with argv, env and aux vectors to emulate execution of the binary and not of the interpreter. Reviewed by: emaste Sponsored by: The FreeBSD Foundation MFC after: 2 weeks Differential revision: https://reviews.freebsd.org/D10701
2017-05-15 18:48:58 +00:00
* places environment pointers and aux vectors right
* after the terminating NULL, we must shift
* environment and aux as well.
*/
Allow rtld direct-exec to take a file descriptor. When executing rtld directly, allow a file descriptor to be explicitly specified rather than opened from the given path. This, together with the LD_LIBRARY_PATH_FDS environment variable, allows dynamically-linked applications to be executed from within capability mode. Also add some rudimentary argument parsing (without pulling in getopt or the like) to accept this file descriptor, a help (-h) option and a basic usage string. Reviewed by: kib Sponsored by: NSERC, RDC Differential Revision: https://reviews.freebsd.org/D10751
2017-05-17 22:51:28 +00:00
main_argc = argc - rtld_argc;
for (i = 0; i <= main_argc; i++)
argv[i] = argv[i + rtld_argc];
*argcp -= rtld_argc;
environ = env = envp = argv + main_argc + 1;
Make ld-elf.so.1 directly executable. Check if passed phdr is actually phdr of the interpreter itself, and decide that this is the case of direct execution. In this case, the binary to activate is specified in the argv[1]. After opening it, shift down on-stack structure with argv, env and aux vectors to emulate execution of the binary and not of the interpreter. Reviewed by: emaste Sponsored by: The FreeBSD Foundation MFC after: 2 weeks Differential revision: https://reviews.freebsd.org/D10701
2017-05-15 18:48:58 +00:00
do {
Allow rtld direct-exec to take a file descriptor. When executing rtld directly, allow a file descriptor to be explicitly specified rather than opened from the given path. This, together with the LD_LIBRARY_PATH_FDS environment variable, allows dynamically-linked applications to be executed from within capability mode. Also add some rudimentary argument parsing (without pulling in getopt or the like) to accept this file descriptor, a help (-h) option and a basic usage string. Reviewed by: kib Sponsored by: NSERC, RDC Differential Revision: https://reviews.freebsd.org/D10751
2017-05-17 22:51:28 +00:00
*envp = *(envp + rtld_argc);
Make ld-elf.so.1 directly executable. Check if passed phdr is actually phdr of the interpreter itself, and decide that this is the case of direct execution. In this case, the binary to activate is specified in the argv[1]. After opening it, shift down on-stack structure with argv, env and aux vectors to emulate execution of the binary and not of the interpreter. Reviewed by: emaste Sponsored by: The FreeBSD Foundation MFC after: 2 weeks Differential revision: https://reviews.freebsd.org/D10701
2017-05-15 18:48:58 +00:00
envp++;
} while (*envp != NULL);
aux = auxp = (Elf_Auxinfo *)envp;
Allow rtld direct-exec to take a file descriptor. When executing rtld directly, allow a file descriptor to be explicitly specified rather than opened from the given path. This, together with the LD_LIBRARY_PATH_FDS environment variable, allows dynamically-linked applications to be executed from within capability mode. Also add some rudimentary argument parsing (without pulling in getopt or the like) to accept this file descriptor, a help (-h) option and a basic usage string. Reviewed by: kib Sponsored by: NSERC, RDC Differential Revision: https://reviews.freebsd.org/D10751
2017-05-17 22:51:28 +00:00
auxpf = (Elf_Auxinfo *)(envp + rtld_argc);
Make ld-elf.so.1 directly executable. Check if passed phdr is actually phdr of the interpreter itself, and decide that this is the case of direct execution. In this case, the binary to activate is specified in the argv[1]. After opening it, shift down on-stack structure with argv, env and aux vectors to emulate execution of the binary and not of the interpreter. Reviewed by: emaste Sponsored by: The FreeBSD Foundation MFC after: 2 weeks Differential revision: https://reviews.freebsd.org/D10701
2017-05-15 18:48:58 +00:00
for (;; auxp++, auxpf++) {
*auxp = *auxpf;
if (auxp->a_type == AT_NULL)
break;
}
} else {
rtld_printf("no binary\n");
rtld_die();
}
}
}
Use a macro to create the names for the library path names. This will allow later substitution at run time instead of compile time of the environment variable name prefix. Differential Review: https://reviews.freebsd.org/D2718
2015-12-27 23:04:12 +00:00
ld_bind_now = getenv(_LD("BIND_NOW"));
Implement LD_BIND_NOT knob for rtld. From the manpage: When set to a nonempty string, prevents modifications of the PLT slots when doing bindings. As result, each call of the PLT-resolved function is resolved. In combination with debug output, this provides complete account of all bind actions at runtime. Same feature exists on Linux and Solaris. Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2017-03-15 21:11:57 +00:00
In the event a process is tainted (setuid/setgid binaries), un-set any potentially dangerous environment variables all together. It should be noted that the run-time linker will not honnor these environment variables if the process is tainted currently. However, once a child of the tainted process calls setuid(2), it's status as being tainted (as defined by issetugid(2)) will be removed. This could be problematic because subsequent activations of the run-time linker could honnor these dangerous variables. This is more of an anti foot-shot mechanism, there is nothing I am aware of in base that does this, however there may be third party utilities which do, and there is no real negative impact of clearing these environment variables. Discussed on: secteam Reviewed by: cperciva PR: kern/109836 MFC after: 2 weeks
2007-05-17 18:00:27 +00:00
/*
* If the process is tainted, then we un-set the dangerous environment
* variables. The process will be marked as tainted until setuid(2)
* is called. If any child process calls setuid(2) we do not want any
* future processes to honor the potentially un-safe variables.
*/
if (!trust) {
Use a macro to create the names for the library path names. This will allow later substitution at run time instead of compile time of the environment variable name prefix. Differential Review: https://reviews.freebsd.org/D2718
2015-12-27 23:04:12 +00:00
if (unsetenv(_LD("PRELOAD")) || unsetenv(_LD("LIBMAP")) ||
unsetenv(_LD("LIBRARY_PATH")) || unsetenv(_LD("LIBRARY_PATH_FDS")) ||
Disable LD_BIND_NOT for setugid processes. Requested by: jilles Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2017-03-15 23:47:19 +00:00
unsetenv(_LD("LIBMAP_DISABLE")) || unsetenv(_LD("BIND_NOT")) ||
Use a macro to create the names for the library path names. This will allow later substitution at run time instead of compile time of the environment variable name prefix. Differential Review: https://reviews.freebsd.org/D2718
2015-12-27 23:04:12 +00:00
unsetenv(_LD("DEBUG")) || unsetenv(_LD("ELF_HINTS_PATH")) ||
unsetenv(_LD("LOADFLTR")) || unsetenv(_LD("LIBRARY_PATH_RPATH"))) {
Fix local root vulnerability. Security: Advisory will be coming soon. X-MFC-After: 30 seconds
2009-12-01 02:57:06 +00:00
_rtld_error("environment corrupt; aborting");
Make die available as rtld_die for use by MD relocation code Reviewed by: kib Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D2213
2015-04-02 21:35:36 +00:00
rtld_die();
Fix local root vulnerability. Security: Advisory will be coming soon. X-MFC-After: 30 seconds
2009-12-01 02:57:06 +00:00
}
In the event a process is tainted (setuid/setgid binaries), un-set any potentially dangerous environment variables all together. It should be noted that the run-time linker will not honnor these environment variables if the process is tainted currently. However, once a child of the tainted process calls setuid(2), it's status as being tainted (as defined by issetugid(2)) will be removed. This could be problematic because subsequent activations of the run-time linker could honnor these dangerous variables. This is more of an anti foot-shot mechanism, there is nothing I am aware of in base that does this, however there may be third party utilities which do, and there is no real negative impact of clearing these environment variables. Discussed on: secteam Reviewed by: cperciva PR: kern/109836 MFC after: 2 weeks
2007-05-17 18:00:27 +00:00
}
Use a macro to create the names for the library path names. This will allow later substitution at run time instead of compile time of the environment variable name prefix. Differential Review: https://reviews.freebsd.org/D2718
2015-12-27 23:04:12 +00:00
ld_debug = getenv(_LD("DEBUG"));
Disable LD_BIND_NOT for setugid processes. Requested by: jilles Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2017-03-15 23:47:19 +00:00
if (ld_bind_now == NULL)
ld_bind_not = getenv(_LD("BIND_NOT")) != NULL;
Use a macro to create the names for the library path names. This will allow later substitution at run time instead of compile time of the environment variable name prefix. Differential Review: https://reviews.freebsd.org/D2718
2015-12-27 23:04:12 +00:00
libmap_disable = getenv(_LD("LIBMAP_DISABLE")) != NULL;
libmap_override = getenv(_LD("LIBMAP"));
ld_library_path = getenv(_LD("LIBRARY_PATH"));
ld_library_dirs = getenv(_LD("LIBRARY_PATH_FDS"));
ld_preload = getenv(_LD("PRELOAD"));
ld_elf_hints_path = getenv(_LD("ELF_HINTS_PATH"));
ld_loadfltr = getenv(_LD("LOADFLTR")) != NULL;
library_path_rpath = getenv(_LD("LIBRARY_PATH_RPATH"));
Import the DragonFly BSD commit 4f0bc915b65fcf5a23214f6d221d65c80be68ad4 by John Marino <draco@marino.st>, with the following (edited) commit message Date: Sat, 24 Mar 2012 06:40:50 +0100 Subject: [PATCH 1/1] rtld: Implement DT_RUNPATH and -z nodefaultlib DT_RUNPATH is incorrectly being considered as an alias of DT_RPATH. The purpose of DT_RUNPATH is to have two different types of rpath: one that can be overridden by the environment variable LD_LIBRARY_PATH and one that can't. With the currently implementation, LD_LIBRARY_PATH will always trump any embedded rpath or runpath tags. Current path search order by rtld: ================================== LD_LIBRARY_PATH DT_RPATH / DT_RUNPATH (always the same) ldconfig hints file (default: /var/run/ld-elf.so.hints) /usr/lib New path search order by rtld: ============================== DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file /usr/lib The new path search matches how the linux runtime loader works. The other major added feature is support for linker flag "-z nodefaultlib". When this flag is passed to the linker, rtld will skip all references to the standard library search path ("/usr/lib" in this case but it could handle more color delimited paths) except in DT_RPATH and DT_RUNPATH. New path search order by rtld with -z nodefaultlib flag set: ============================================================ DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file (skips all references to /usr/lib) FreeBSD notes: - we fixed some bugs which were submitted to DragonFly and merged there as commit 1ff8a2bd3eb6e5587174c6a983303ea3a79e0002; - we added LD_LIBRARY_PATH_RPATH environment variable to switch to the previous behaviour of considering DT_RPATH a synonym for DT_RUNPATH; - the FreeBSD default search path is /lib:/usr/lib and not /usr/lib. Reviewed by: kan MFC after: 1 month MFC note: flip the ld_library_path_rpath default value for stable/9
2012-07-15 10:53:48 +00:00
if (library_path_rpath != NULL) {
if (library_path_rpath[0] == 'y' ||
library_path_rpath[0] == 'Y' ||
library_path_rpath[0] == '1')
ld_library_path_rpath = true;
else
ld_library_path_rpath = false;
}
In the event a process is tainted (setuid/setgid binaries), un-set any potentially dangerous environment variables all together. It should be noted that the run-time linker will not honnor these environment variables if the process is tainted currently. However, once a child of the tainted process calls setuid(2), it's status as being tainted (as defined by issetugid(2)) will be removed. This could be problematic because subsequent activations of the run-time linker could honnor these dangerous variables. This is more of an anti foot-shot mechanism, there is nothing I am aware of in base that does this, however there may be third party utilities which do, and there is no real negative impact of clearing these environment variables. Discussed on: secteam Reviewed by: cperciva PR: kern/109836 MFC after: 2 weeks
2007-05-17 18:00:27 +00:00
dangerous_ld_env = libmap_disable || (libmap_override != NULL) ||
Support for a new environment variable, LD_ELF_HINTS_PATH for overriding the rtld hints file. This environment variable would be unset if the process is considered as tainted with setuid/setgid. This feature gives a convenient way of using a custom set of shared library that is not located in the default location and switch back. Feature requested by: iXsystems Original patch by: John Hixson MFC after: 2 weeks
2009-03-23 16:49:00 +00:00
(ld_library_path != NULL) || (ld_preload != NULL) ||
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
(ld_elf_hints_path != NULL) || ld_loadfltr;
Use a macro to create the names for the library path names. This will allow later substitution at run time instead of compile time of the environment variable name prefix. Differential Review: https://reviews.freebsd.org/D2718
2015-12-27 23:04:12 +00:00
ld_tracing = getenv(_LD("TRACE_LOADED_OBJECTS"));
ld_utrace = getenv(_LD("UTRACE"));
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
Support for a new environment variable, LD_ELF_HINTS_PATH for overriding the rtld hints file. This environment variable would be unset if the process is considered as tainted with setuid/setgid. This feature gives a convenient way of using a custom set of shared library that is not located in the default location and switch back. Feature requested by: iXsystems Original patch by: John Hixson MFC after: 2 weeks
2009-03-23 16:49:00 +00:00
if ((ld_elf_hints_path == NULL) || strlen(ld_elf_hints_path) == 0)
Use variable names that aren't as prone to dyslexic confusion. Suggested by: jmallet@
2015-12-18 05:45:49 +00:00
ld_elf_hints_path = ld_elf_hints_default;
Support for a new environment variable, LD_ELF_HINTS_PATH for overriding the rtld hints file. This environment variable would be unset if the process is considered as tainted with setuid/setgid. This feature gives a convenient way of using a custom set of shared library that is not located in the default location and switch back. Feature requested by: iXsystems Original patch by: John Hixson MFC after: 2 weeks
2009-03-23 16:49:00 +00:00
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
if (ld_debug != NULL && *ld_debug != '\0')
debug = 1;
dbg("%s is initialized, base address = %p", __progname,
(caddr_t) aux_info[AT_BASE]->a_un.a_ptr);
Eliminate all machine-dependent code from the main source body and the Makefile, and move it down into the architecture-specific subdirectories. Eliminate an asm() statement for the i386. Make the dynamic linker work if it is built as an executable instead of as a shared library. See i386/Makefile.inc to find out how to do it. Note, this change is not enabled and it might never be enabled. But it might be useful in the future. Building the dynamic linker as an executable should make it start up faster, because it won't have any relocations. But in practice I suspect the difference is negligible.
1999-04-09 00:28:43 +00:00
dbg("RTLD dynamic = %p", obj_rtld.dynamic);
dbg("RTLD pltgot = %p", obj_rtld.pltgot);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
dbg("initializing thread locks");
lockdflt_init();
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
/*
* Load the main program, or process its program header if it is
* already loaded.
*/
Make ld-elf.so.1 directly executable. Check if passed phdr is actually phdr of the interpreter itself, and decide that this is the case of direct execution. In this case, the binary to activate is specified in the argv[1]. After opening it, shift down on-stack structure with argv, env and aux vectors to emulate execution of the binary and not of the interpreter. Reviewed by: emaste Sponsored by: The FreeBSD Foundation MFC after: 2 weeks Differential revision: https://reviews.freebsd.org/D10701
2017-05-15 18:48:58 +00:00
if (fd != -1) { /* Load the main program. */
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
dbg("loading main program");
When checking to see if a shared object is already loaded, look for a device/inode match if no pathname match is found.
1999-08-30 01:50:41 +00:00
obj_main = map_object(fd, argv0, NULL);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
close(fd);
if (obj_main == NULL)
Make die available as rtld_die for use by MD relocation code Reviewed by: kib Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D2213
2015-04-02 21:35:36 +00:00
rtld_die();
In rtld, read the initial stack access mode from AT_STACKPROT as set by kernel, and parse PT_GNU_STACK phdr from linked and loaded dsos. If the loaded dso requires executable stack, as specified by PF_X bit of p_flags of PT_GNU_STACK phdr, but current stack protection does not permit execution, the __pthread_map_stacks_exec symbol is looked up and called. It should be implemented in libc or threading library and change the protection mode of all thread stacks to be executable. Provide a private interface _rtld_get_stack_prot() to export the stack access mode as calculated by rtld. Reviewed by: kan
2011-01-08 17:11:49 +00:00
max_stack_flags = obj->stack_flags;
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
} else { /* Main program already loaded. */
dbg("processing main program's program header");
assert(aux_info[AT_PHDR] != NULL);
Add alpha support. Submitted by: John Birrell <jb@cimlogic.com.au> (with extra hacks by me) Obtained from: Probably NetBSD
1998-09-04 19:03:57 +00:00
phdr = (const Elf_Phdr *) aux_info[AT_PHDR]->a_un.a_ptr;
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
assert(aux_info[AT_PHNUM] != NULL);
phnum = aux_info[AT_PHNUM]->a_un.a_val;
assert(aux_info[AT_PHENT] != NULL);
Add alpha support. Submitted by: John Birrell <jb@cimlogic.com.au> (with extra hacks by me) Obtained from: Probably NetBSD
1998-09-04 19:03:57 +00:00
assert(aux_info[AT_PHENT]->a_un.a_val == sizeof(Elf_Phdr));
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
assert(aux_info[AT_ENTRY] != NULL);
In _rtld(), reorder local declarations to compact the block and partially sort them by style(9). Move locals declarations from nested blocks into the block at function start. Discussed with: emaste MFC after: 1 week
2017-05-13 18:59:27 +00:00
imgentry = (caddr_t) aux_info[AT_ENTRY]->a_un.a_ptr;
if ((obj_main = digest_phdr(phdr, phnum, imgentry, argv0)) == NULL)
Make die available as rtld_die for use by MD relocation code Reviewed by: kib Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D2213
2015-04-02 21:35:36 +00:00
rtld_die();
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
}
Make ld-elf.so.1 directly executable. Check if passed phdr is actually phdr of the interpreter itself, and decide that this is the case of direct execution. In this case, the binary to activate is specified in the argv[1]. After opening it, shift down on-stack structure with argv, env and aux vectors to emulate execution of the binary and not of the interpreter. Reviewed by: emaste Sponsored by: The FreeBSD Foundation MFC after: 2 weeks Differential revision: https://reviews.freebsd.org/D10701
2017-05-15 18:48:58 +00:00
if (aux_info[AT_EXECPATH] != NULL && fd == -1) {
Implement the dynamic string token substitution in the rpath and soneeded pathes. The $ORIGIN, $OSNAME, $OSREL and $PLATFORM tokens are supported. Enabling the substitution requires DF_ORIGIN flag in DT_FLAGS or DF_1_ORIGIN if DF_FLAGS_1, that may be set with -z origin gnu ld flag. Translation is unconditionally disabled for setuid/setgid processes. The $ORIGIN translation relies on the AT_EXECPATH auxinfo supplied by kernel. Requested by: maho Tested by: maho, pho Reviewed by: kan
2009-03-18 13:40:37 +00:00
kexecpath = aux_info[AT_EXECPATH]->a_un.a_ptr;
dbg("AT_EXECPATH %p %s", kexecpath, kexecpath);
if (kexecpath[0] == '/')
obj_main->path = kexecpath;
else if (getcwd(buf, sizeof(buf)) == NULL ||
strlcat(buf, "/", sizeof(buf)) >= sizeof(buf) ||
strlcat(buf, kexecpath, sizeof(buf)) >= sizeof(buf))
obj_main->path = xstrdup(argv0);
else
obj_main->path = xstrdup(buf);
} else {
Make ld-elf.so.1 directly executable. Check if passed phdr is actually phdr of the interpreter itself, and decide that this is the case of direct execution. In this case, the binary to activate is specified in the argv[1]. After opening it, shift down on-stack structure with argv, env and aux vectors to emulate execution of the binary and not of the interpreter. Reviewed by: emaste Sponsored by: The FreeBSD Foundation MFC after: 2 weeks Differential revision: https://reviews.freebsd.org/D10701
2017-05-15 18:48:58 +00:00
dbg("No AT_EXECPATH or direct exec");
Implement the dynamic string token substitution in the rpath and soneeded pathes. The $ORIGIN, $OSNAME, $OSREL and $PLATFORM tokens are supported. Enabling the substitution requires DF_ORIGIN flag in DT_FLAGS or DF_1_ORIGIN if DF_FLAGS_1, that may be set with -z origin gnu ld flag. Translation is unconditionally disabled for setuid/setgid processes. The $ORIGIN translation relies on the AT_EXECPATH auxinfo supplied by kernel. Requested by: maho Tested by: maho, pho Reviewed by: kan
2009-03-18 13:40:37 +00:00
obj_main->path = xstrdup(argv0);
}
dbg("obj_main path %s", obj_main->path);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
obj_main->mainprog = true;
Get the actual pathname of the dynamic linker from the executable's PT_INTERP program header entry, to ensure that gdb always finds the right dynamic linker. Use obj->relocbase to simplify a few calculations where appropriate.
1999-08-30 01:54:13 +00:00
In rtld, read the initial stack access mode from AT_STACKPROT as set by kernel, and parse PT_GNU_STACK phdr from linked and loaded dsos. If the loaded dso requires executable stack, as specified by PF_X bit of p_flags of PT_GNU_STACK phdr, but current stack protection does not permit execution, the __pthread_map_stacks_exec symbol is looked up and called. It should be implemented in libc or threading library and change the protection mode of all thread stacks to be executable. Provide a private interface _rtld_get_stack_prot() to export the stack access mode as calculated by rtld. Reviewed by: kan
2011-01-08 17:11:49 +00:00
if (aux_info[AT_STACKPROT] != NULL &&
aux_info[AT_STACKPROT]->a_un.a_val != 0)
stack_prot = aux_info[AT_STACKPROT]->a_un.a_val;
The runtime linker needs to include a path to itself in the link map it exports to the debugger. It currently has two choices: it can use a compiled-in path (/libexec/ld-elf.so.1) or it can use the path stored in the interpreter path in the binary being executed. The runtime linker currently prefers the second. However, this is usually wrong for compat32 binaries since the binary specifies the path of rtld on a 32-bit system (/libexec/ld-elf.so.1) instead of the actual path (/libexec/ld-elf32.so.1). For now, always assume the compiled in path (/libexec/ld-elf32.so.1) as the rtld path and ignore the path in the binary for the 32-bit runtime linker. Differential Revision: https://reviews.freebsd.org/D1236 Reviewed by: kib
2014-12-02 16:22:08 +00:00
#ifndef COMPAT_32BIT
Get the actual pathname of the dynamic linker from the executable's PT_INTERP program header entry, to ensure that gdb always finds the right dynamic linker. Use obj->relocbase to simplify a few calculations where appropriate.
1999-08-30 01:54:13 +00:00
/*
* Get the actual dynamic linker pathname from the executable if
* possible. (It should always be possible.) That ensures that
* gdb will find the right dynamic linker even if a non-standard
* one is being used.
*/
if (obj_main->interp != NULL &&
strcmp(obj_main->interp, obj_rtld.path) != 0) {
free(obj_rtld.path);
obj_rtld.path = xstrdup(obj_main->interp);
If we change obj_rtld.path after initialising __progname, make sure we change __progname to point at the new storage otherwise it ends up pointing at freed memory which leads to confusing garbled error messages.
2004-03-29 18:37:37 +00:00
__progname = obj_rtld.path;
Get the actual pathname of the dynamic linker from the executable's PT_INTERP program header entry, to ensure that gdb always finds the right dynamic linker. Use obj->relocbase to simplify a few calculations where appropriate.
1999-08-30 01:54:13 +00:00
}
The runtime linker needs to include a path to itself in the link map it exports to the debugger. It currently has two choices: it can use a compiled-in path (/libexec/ld-elf.so.1) or it can use the path stored in the interpreter path in the binary being executed. The runtime linker currently prefers the second. However, this is usually wrong for compat32 binaries since the binary specifies the path of rtld on a 32-bit system (/libexec/ld-elf.so.1) instead of the actual path (/libexec/ld-elf32.so.1). For now, always assume the compiled in path (/libexec/ld-elf32.so.1) as the rtld path and ignore the path in the binary for the 32-bit runtime linker. Differential Revision: https://reviews.freebsd.org/D1236 Reviewed by: kib
2014-12-02 16:22:08 +00:00
#endif
Get the actual pathname of the dynamic linker from the executable's PT_INTERP program header entry, to ensure that gdb always finds the right dynamic linker. Use obj->relocbase to simplify a few calculations where appropriate.
1999-08-30 01:54:13 +00:00
Minor changes to make this work on sparc64. Approved by: jdp Tested on: alpha, i386, sparc64
2002-04-02 02:19:02 +00:00
digest_dynamic(obj_main, 0);
Add GNU hash support for rtld. Based on dragonflybsd support for GNU hash by John Marino <draco marino st> Reviewed by: kan Tested by: bapt MFC after: 2 weeks
2012-04-30 13:31:10 +00:00
dbg("%s valid_hash_sysv %d valid_hash_gnu %d dynsymcount %d",
obj_main->path, obj_main->valid_hash_sysv, obj_main->valid_hash_gnu,
obj_main->dynsymcount);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
Add GDB support. The method and some of the code came from NetBSD's elf runtime linker.
1998-04-30 07:48:02 +00:00
linkmap_add(obj_main);
linkmap_add(&obj_rtld);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
/* Link the main program into the list of objects. */
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
TAILQ_INSERT_HEAD(&obj_list, obj_main, next);
Solve the dynamic linker's problems with multithreaded programs once and for all (I hope). Packages such as wine, JDK, and linuxthreads should no longer have any problems with re-entering the dynamic linker. This commit replaces the locking used in the dynamic linker with a new spinlock-based reader/writer lock implementation. Brian Fundakowski Feldman <green> argued for this from the very beginning, but it took me a long time to come around to his point of view. Spinlocks are the only kinds of locks that work with all thread packages. But on uniprocessor systems they can be inefficient, because while a contender for the lock is spinning the holder of the lock cannot make any progress toward releasing it. To alleviate this disadvantage I have borrowed a trick from Sleepycat's Berkeley DB implementation. When spinning for a lock, the requester does a nanosleep() call for 1 usec. each time around the loop. This will generally yield the CPU to other threads, allowing the lock holder to finish its business and release the lock. I chose 1 usec. as the minimum sleep which would with reasonable certainty not be rounded down to 0. The formerly machine-independent file "lockdflt.c" has been moved into the architecture-specific subdirectories by repository copy. It now contains the machine-dependent spinlocking code. For the spinlocks I used the very nifty "simple, non-scalable reader-preference lock" which I found at <http://www.cs.rochester.edu/u/scott/synchronization/pseudocode/rw.html> on all CPUs except the 80386 (the specific CPU model, not the architecture). The 80386 CPU doesn't support the necessary "cmpxchg" instruction, so on that CPU a simple exclusive test-and-set lock is used instead. 80386 CPUs are detected at initialization time by trying to execute "cmpxchg" and catching the resulting SIGILL signal. To reduce contention for the locks, I have revamped a couple of key data structures, permitting all common operations to be done under non-exclusive (reader) locking. The only operations that require exclusive locking now are the rare intrusive operations such as dlopen() and dlclose(). The dllockinit() interface is now deprecated. It still exists, but only as a do-nothing stub. I plan to remove it as soon as is reasonably possible. (From the very beginning it was clearly labeled as experimental and subject to change.) As far as I know, only the linuxthreads port uses dllockinit(). This interface turned out to have several problems. As one example, when the dynamic linker called a client-supplied locking function, that function sometimes needed lazy binding, causing re-entry into the dynamic linker and a big looping mess. And in any case, it turned out to be too burdensome to require threads packages to register themselves with the dynamic linker.
2000-07-08 04:10:38 +00:00
obj_count++;
Implement dl_iterate_phdr function. Convert boolean flags in internal Obj_Entry structure into bitfields. Properly check for loaded segment alignment in map_object.
2007-04-03 18:31:20 +00:00
obj_loads++;
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
Resolve undefined weak references to a value of 0. This solves the "__deregister_frame_info" problem that was seen when combining a program linked using the old gcc with shared libraries that were built using egcs.
1999-04-05 02:36:40 +00:00
/* Initialize a fake symbol for resolving undefined weak references. */
sym_zero.st_info = ELF_ST_INFO(STB_GLOBAL, STT_NOTYPE);
Fix handling of weak references to undefined symbols on ia64: o Set st_shndx for sym_zero to SHN_UNDEF instead of SHN_ABS. This gives us something to reliably test against. o For weak references to undefined sysmbols (as indicated by having st_shndx equals SHN_UNDEF) in the context of OPDs, the address of the OPD is to be zero, not the address of the function it contains. o For weak references to undefined symbols in all other cases (only DIR64LSB at this time), the actual relocated value is to be zero, not the value prior to relocating. Roughly speaking, weak references to undefined symbols are no-ops. Tested on: i386, ia64
2002-04-27 05:32:51 +00:00
sym_zero.st_shndx = SHN_UNDEF;
Calculate relocation base for the main object, and apply the relocation adjustment for all virtual addresses encoded into the ELF structures of it. PIE binary could and should be loaded at non-zero mapbase. For sym_zero pseudosymbol used as a return value from find_symdef() for undefined weak symbols, st_value also should be adjusted, since _rtld_bind corrects symbol values by relocbase. Discussed with: bz Reviewed by: kan Tested by: bz (i386, amd64), bsam (linux) MFC after: some time
2009-10-10 15:27:10 +00:00
sym_zero.st_value = -(uintptr_t)obj_main->relocbase;
Resolve undefined weak references to a value of 0. This solves the "__deregister_frame_info" problem that was seen when combining a program linked using the old gcc with shared libraries that were built using egcs.
1999-04-05 02:36:40 +00:00
Use the environment variable LD_LIBMAP_DISABLE to disable libmap.conf(5) functionality.
2003-05-31 14:45:11 +00:00
if (!libmap_disable)
Description from Dan: Another handy libmap patch. Lets you do stuff like this: LD_LIBMAP="libpthread.so.1=libthr.so.1" mythreadedapp If you already have a program-specific override in libmap.conf, note that you must use a program-specific override in LD_LIBMAP: LD_LIBMAP="[mythreadedapp],libpthread.so.1=libthr.so.1" mythreadedapp PR: bin/74471 Submitted by: Dan Nelson <dnelson AT allantgroup.com> MFC after: 2 weeks
2005-02-04 02:46:41 +00:00
libmap_disable = (bool)lm_init(libmap_override);
Dynamic object dependency mapping: libmap. This is an optional feature, disabled by default. This will be useful to people testing the various POSIX threading libraries under -CURRENT but can easily serve other needs.
2003-04-07 16:21:26 +00:00
Make LD_PRELOAD work for ELF.
1998-09-22 02:09:56 +00:00
dbg("loading LD_PRELOAD libraries");
if (load_preload_objects() == -1)
Make die available as rtld_die for use by MD relocation code Reviewed by: kib Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D2213
2015-04-02 21:35:36 +00:00
rtld_die();
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
preload_tail = globallist_curr(TAILQ_LAST(&obj_list, obj_entry_q));
Make LD_PRELOAD work for ELF.
1998-09-22 02:09:56 +00:00
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
dbg("loading needed objects");
Implement rtld part of the support for -z nodlopen (see ld(1)). Reviewed by: kan MFC after: 3 weeks
2009-11-26 13:57:20 +00:00
if (load_needed_objects(obj_main, 0) == -1)
Make die available as rtld_die for use by MD relocation code Reviewed by: kib Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D2213
2015-04-02 21:35:36 +00:00
rtld_die();
Revamp the symbol lookup algorithm to cope better with objects loaded separately by dlopen that have global symbols with identical names. Viewing each dlopened object as a DAG which is linked by its DT_NEEDED entries in the dynamic table, the search order is as follows: * If the referencing object was linked with -Bsymbolic, search it internally. * Search all dlopened DAGs containing the referencing object. * Search all objects loaded at program start up. * Search all objects which were dlopened() using the RTLD_GLOBAL flag (which is now supported too). The search terminates as soon as a strong definition is found. Lacking that, the first weak definition is used. These rules match those of Solaris, as best I could determine them from its vague manual pages and the results of experiments I performed. PR: misc/12438
1999-08-30 01:48:19 +00:00
Revamp the code that calls shared libraries' init and fini functions. Formerly the init functions were called in the opposite of the order in which libraries were loaded, and libraries were loaded according to a breadth-first traversal of the dependency graph. That ordering came from SVR4.0, and it was easy to implement but not always sensible. Now we do a depth-first walk over the dependency graph and call the init functions in an order such that each shared object's needed objects are initialized before the shared object itself. At the same time we build a list of finalization (fini) functions in the opposite order, to guarantee correct C++ destructor ordering whenever possible. (It may not be possible if dlopen and dlclose are used in strange ways, but we come as close as one can come.) The need for this renovation has become apparent as more programs have started using multithreading. The multithreaded C library libc_r requires initialization, whereas the standard libc does not. Since virtually every other object depends on the C library, it is important that it get initialized first.
2000-07-26 04:24:40 +00:00
/* Make a list of all objects loaded at startup. */
Implement support for the interpose dso flag. Requested by: bf Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 1 week Approved by: re (glebius)
2013-10-07 08:19:30 +00:00
last_interposer = obj_main;
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
TAILQ_FOREACH(obj, &obj_list, next) {
if (obj->marker)
continue;
Implement support for the interpose dso flag. Requested by: bf Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 1 week Approved by: re (glebius)
2013-10-07 08:19:30 +00:00
if (obj->z_interpose && obj != obj_main) {
objlist_put_after(&list_main, last_interposer, obj);
last_interposer = obj;
} else {
objlist_push_tail(&list_main, obj);
}
Rethink the way we count module references. Simply following DT_NEEDED links is not flexible enough for cases where dynamically loaded modules form a dependency cycle. This should fix an infinite recursion problem encountered by Yahoo. Approved by: re (jhb)
2003-05-08 01:31:36 +00:00
obj->refcount++;
}
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
Implement ELF symbol versioning using GNU semantics. This code aims to be compatible with symbol versioning support as implemented by GNU libc and documented by http://people.redhat.com/~drepper/symbol-versioning and LSB 3.0. Implement dlvsym() function to allow lookups for a specific version of a given symbol.
2005-12-18 19:43:33 +00:00
dbg("checking for required versions");
if (rtld_verify_versions(&list_main) == -1 && !ld_tracing)
Make die available as rtld_die for use by MD relocation code Reviewed by: kib Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D2213
2015-04-02 21:35:36 +00:00
rtld_die();
Implement ELF symbol versioning using GNU semantics. This code aims to be compatible with symbol versioning support as implemented by GNU libc and documented by http://people.redhat.com/~drepper/symbol-versioning and LSB 3.0. Implement dlvsym() function to allow lookups for a specific version of a given symbol.
2005-12-18 19:43:33 +00:00
Add support for ldd.
1998-05-01 08:39:27 +00:00
if (ld_tracing) { /* We're done */
trace_loaded_objects(obj_main);
exit(0);
}
Use a macro to create the names for the library path names. This will allow later substitution at run time instead of compile time of the environment variable name prefix. Differential Review: https://reviews.freebsd.org/D2718
2015-12-27 23:04:12 +00:00
if (getenv(_LD("DUMP_REL_PRE")) != NULL) {
Provide a mechanism for dumping relocation information. Setting the LD_DUMP_REL_PRE or LD_DUMP_REL_POST environment variables cause rtld-elf to output a table of all relocations. This is useful for debugging.
2003-06-19 03:55:38 +00:00
dump_relocations(obj_main);
exit (0);
}
Setting up TLS block for the main thread must be done after the relocations are processed, since tls initialization section might be itself subject for relocations. Only set up of the block is postponed, the tls block offsets are allocated before relocation processing, since TLS-related relocations may need offsets ready. Reported by: ale PR: threads/161344 Reviewed by: kan MFC after: 1 week
2011-10-08 12:39:47 +00:00
/*
* Processing tls relocations requires having the tls offsets
* initialized. Prepare offsets before starting initial
* relocation processing.
*/
dbg("initializing initial thread local storage offsets");
Do the TLS offset allocations before relocations, as otherwise there can be overlap in the TLS offsets, if the relocations are done in a certain order. Approved by: dfr, grehan (mentor)
2004-11-02 09:42:21 +00:00
STAILQ_FOREACH(entry, &list_main, link) {
/*
* Allocate all the initial objects out of the static TLS
* block even if they didn't ask for it.
*/
allocate_tls_offset(entry->obj);
}
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
if (relocate_objects(obj_main,
Fix several problems with our ELF filters implementation. Do not relocate twice an object which happens to be needed by loaded binary (or dso) and some filtee opened due to symbol resolution when relocating need objects. Record the state of the relocation processing in Obj_Entry and short-circuit relocate_objects() if current object already processed. Do not call constructors for filtees loaded during the early relocation processing before image is initialized enough to run user-provided code. Filtees are loaded using dlopen_object(), which normally performs relocation and initialization. If filtee is lazy-loaded during the relocation of dso needed by the main object, dlopen_object() runs too earlier, when most runtime services are not yet ready. Postpone the constructors call to the time when main binary and depended libraries constructors are run, passing the new flag RTLD_LO_EARLY to dlopen_object(). Symbol lookups callers inform symlook_* functions about early stage of initialization with SYMLOOK_EARLY. Pass flags through all functions participating in object relocation. Use the opportunity and fix flags argument to find_symdef() in arch-specific reloc.c to use proper name SYMLOOK_IN_PLT instead of true, which happen to have the same numeric value. Reported and tested by: theraven Reviewed by: kan MFC after: 2 weeks
2012-03-20 13:20:49 +00:00
ld_bind_now != NULL && *ld_bind_now != '\0',
&obj_rtld, SYMLOOK_EARLY, NULL) == -1)
Make die available as rtld_die for use by MD relocation code Reviewed by: kib Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D2213
2015-04-02 21:35:36 +00:00
rtld_die();
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
dbg("doing copy relocations");
if (do_copy_relocations(obj_main) == -1)
Make die available as rtld_die for use by MD relocation code Reviewed by: kib Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D2213
2015-04-02 21:35:36 +00:00
rtld_die();
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
For the main binary, postpone enforcing relro read-only protection until copy relocations are done. Newer binutils and lld seems to output copy into relro-protected range. Reported by: Rafael Espц╜ndola via emaste Sponsored by: The FreeBSD Foundation MFC after: 1 week
2017-01-12 15:54:03 +00:00
dbg("enforcing main obj relro");
if (obj_enforce_relro(obj_main) == -1)
rtld_die();
Use a macro to create the names for the library path names. This will allow later substitution at run time instead of compile time of the environment variable name prefix. Differential Review: https://reviews.freebsd.org/D2718
2015-12-27 23:04:12 +00:00
if (getenv(_LD("DUMP_REL_POST")) != NULL) {
Provide a mechanism for dumping relocation information. Setting the LD_DUMP_REL_PRE or LD_DUMP_REL_POST environment variables cause rtld-elf to output a table of all relocations. This is useful for debugging.
2003-06-19 03:55:38 +00:00
dump_relocations(obj_main);
exit (0);
}
Setting up TLS block for the main thread must be done after the relocations are processed, since tls initialization section might be itself subject for relocations. Only set up of the block is postponed, the tls block offsets are allocated before relocation processing, since TLS-related relocations may need offsets ready. Reported by: ale PR: threads/161344 Reviewed by: kan MFC after: 1 week
2011-10-08 12:39:47 +00:00
/*
* Setup TLS for main thread. This must be done after the
* relocations are processed, since tls initialization section
* might be the subject for relocations.
*/
dbg("initializing initial thread local storage");
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
allocate_initial_tls(globallist_curr(TAILQ_FIRST(&obj_list)));
Setting up TLS block for the main thread must be done after the relocations are processed, since tls initialization section might be itself subject for relocations. Only set up of the block is postponed, the tls block offsets are allocated before relocation processing, since TLS-related relocations may need offsets ready. Reported by: ale PR: threads/161344 Reviewed by: kan MFC after: 1 week
2011-10-08 12:39:47 +00:00
After relocating the main program, but before calling any of the _init() functions, initialize the global variables "__progname" and "environ". This makes it possible for the _init() functions to call things like getenv() and err().
1999-04-21 04:06:57 +00:00
dbg("initializing key program variables");
set_program_var("__progname", argv[0] != NULL ? basename(argv[0]) : "");
set_program_var("environ", env);
Provide the starting image with the pointer to ELF aux vector. It is written into the __elf_aux_vector variable, if the symbol is present. Idea from: kan Tested by: marius (sparc64) MFC after: 1 month
2010-08-17 09:08:28 +00:00
set_program_var("__elf_aux_vector", aux);
After relocating the main program, but before calling any of the _init() functions, initialize the global variables "__progname" and "environ". This makes it possible for the _init() functions to call things like getenv() and err().
1999-04-21 04:06:57 +00:00
Revamp the code that calls shared libraries' init and fini functions. Formerly the init functions were called in the opposite of the order in which libraries were loaded, and libraries were loaded according to a breadth-first traversal of the dependency graph. That ordering came from SVR4.0, and it was easy to implement but not always sensible. Now we do a depth-first walk over the dependency graph and call the init functions in an order such that each shared object's needed objects are initialized before the shared object itself. At the same time we build a list of finalization (fini) functions in the opposite order, to guarantee correct C++ destructor ordering whenever possible. (It may not be possible if dlopen and dlclose are used in strange ways, but we come as close as one can come.) The need for this renovation has become apparent as more programs have started using multithreading. The multithreaded C library libc_r requires initialization, whereas the standard libc does not. Since virtually every other object depends on the C library, it is important that it get initialized first.
2000-07-26 04:24:40 +00:00
/* Make a list of init functions to call. */
objlist_init(&initlist);
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
initlist_add_objects(globallist_curr(TAILQ_FIRST(&obj_list)),
preload_tail, &initlist);
Revamp the code that calls shared libraries' init and fini functions. Formerly the init functions were called in the opposite of the order in which libraries were loaded, and libraries were loaded according to a breadth-first traversal of the dependency graph. That ordering came from SVR4.0, and it was easy to implement but not always sensible. Now we do a depth-first walk over the dependency graph and call the init functions in an order such that each shared object's needed objects are initialized before the shared object itself. At the same time we build a list of finalization (fini) functions in the opposite order, to guarantee correct C++ destructor ordering whenever possible. (It may not be possible if dlopen and dlclose are used in strange ways, but we come as close as one can come.) The need for this renovation has become apparent as more programs have started using multithreading. The multithreaded C library libc_r requires initialization, whereas the standard libc does not. Since virtually every other object depends on the C library, it is important that it get initialized first.
2000-07-26 04:24:40 +00:00
Pass two pointer parameters to the r_debug_state() hook function, thus allowing a debugger or other trace tool to easily grab the addresses of the needed structures off the stack. This change is transparent to gdb, which locates the link_map list and transfers it to debugger memory for comparison purposes. A sample program will be committed showing how this can be used. Reviewed by: John Polstra <jdp@FreeBSD.org>
2000-08-26 05:13:29 +00:00
r_debug_state(NULL, &obj_main->linkmap); /* say hello to gdb! */
Shake hands with GDB a little bit earlier so that it is possible to debug the init functions. Submitted by: dfr
1999-07-03 23:54:02 +00:00
get_program_var_addr() must prefer the strong symbol to the weak one. Search global objects, together with main object and dependencies, for the requested symbol. Move the common code from symlook_default() into new helper symlook_global(), and use it both in symlook_global() and get_program_var_addr(). Supply lock state to get_program_var_addr(). Reviewed by: kan Tested by: Mykola Dzham <i levsha me>
2011-01-10 16:09:35 +00:00
map_stacks_exec(NULL);
Pass CPUID[1] %edx (cpu_feature), %ecx (cpu_feature2) and CPUID[7].%ebx (cpu_stdext_feature), %ecx (cpu_stdext_feature2) to the ifunc resolvers on x86. It is much more clean to use CPUID instruction in usermode to retrieve this information than to pass AT_HWCAP aux vector from kernel, on x86. Still, the change does allow for use of AT_HWCAP on arches where it is needed, by passing aux array to ifunc_init() initializer which should prepare arguments for ifunc resolvers. Current signature for resolvers on x86 is func_t iresolve(uint32_t cpu_feature, uint32_t cpu_feature2, uint32_t cpu_stdext_feature, uint32_t cpu_stdext_feature2); where arguments have identical meaning as the kernel variables of the same name. The ABIs allow to use resolvers with the void or shortened list of arguments. Reviewed by: jhb Sponsored by: The FreeBSD Foundation MFC after: 1 week Differential revision: https://reviews.freebsd.org/D8448
2016-11-15 09:43:26 +00:00
ifunc_init(aux);
In rtld, read the initial stack access mode from AT_STACKPROT as set by kernel, and parse PT_GNU_STACK phdr from linked and loaded dsos. If the loaded dso requires executable stack, as specified by PF_X bit of p_flags of PT_GNU_STACK phdr, but current stack protection does not permit execution, the __pthread_map_stacks_exec symbol is looked up and called. It should be implemented in libc or threading library and change the protection mode of all thread stacks to be executable. Provide a private interface _rtld_get_stack_prot() to export the stack access mode as calculated by rtld. Reviewed by: kan
2011-01-08 17:11:49 +00:00
Postpone the resolution of IRELATIVE relocations and IFUNC-targeted relocations until tls is initialized and stacks permissions correctly set. This allows the ifunc to call malloc(3) and some other heavy services. Add debug banner. MFC after: 3 days
2012-01-04 17:17:11 +00:00
dbg("resolving ifuncs");
if (resolve_objects_ifunc(obj_main,
Fix several problems with our ELF filters implementation. Do not relocate twice an object which happens to be needed by loaded binary (or dso) and some filtee opened due to symbol resolution when relocating need objects. Record the state of the relocation processing in Obj_Entry and short-circuit relocate_objects() if current object already processed. Do not call constructors for filtees loaded during the early relocation processing before image is initialized enough to run user-provided code. Filtees are loaded using dlopen_object(), which normally performs relocation and initialization. If filtee is lazy-loaded during the relocation of dso needed by the main object, dlopen_object() runs too earlier, when most runtime services are not yet ready. Postpone the constructors call to the time when main binary and depended libraries constructors are run, passing the new flag RTLD_LO_EARLY to dlopen_object(). Symbol lookups callers inform symlook_* functions about early stage of initialization with SYMLOOK_EARLY. Pass flags through all functions participating in object relocation. Use the opportunity and fix flags argument to find_symdef() in arch-specific reloc.c to use proper name SYMLOOK_IN_PLT instead of true, which happen to have the same numeric value. Reported and tested by: theraven Reviewed by: kan MFC after: 2 weeks
2012-03-20 13:20:49 +00:00
ld_bind_now != NULL && *ld_bind_now != '\0', SYMLOOK_EARLY,
NULL) == -1)
Make die available as rtld_die for use by MD relocation code Reviewed by: kib Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D2213
2015-04-02 21:35:36 +00:00
rtld_die();
Postpone the resolution of IRELATIVE relocations and IFUNC-targeted relocations until tls is initialized and stacks permissions correctly set. This allows the ifunc to call malloc(3) and some other heavy services. Add debug banner. MFC after: 3 days
2012-01-04 17:17:11 +00:00
Add support for preinit, init and fini arrays. Some ABIs, in particular on ARM, do require working init arrays. Traditional FreeBSD crt1 calls _init and _fini of the binary, instead of allowing runtime linker to arrange the calls. This was probably done to have the same crt code serve both statically and dynamically linked binaries. Since ABI mandates that first is called preinit array functions, then init, and then init array functions, the init have to be called from rtld now. To provide binary compatibility to old FreeBSD crt1, which calls _init itself, rtld only calls intializers and finalizers for main binary if binary has a note indicating that new crt was used for linking. Add parsing of ELF notes to rtld, and cache p_osrel value since we parsed it anyway. The patch is inspired by init_array support for DragonflyBSD, written by John Marino. Reviewed by: kan Tested by: andrew (arm, previous version), flo (sparc64, previous version) MFC after: 3 weeks
2012-03-11 20:03:09 +00:00
if (!obj_main->crt_no_init) {
/*
* Make sure we don't call the main program's init and fini
* functions for binaries linked with old crt1 which calls
* _init itself.
*/
obj_main->init = obj_main->fini = (Elf_Addr)NULL;
obj_main->preinit_array = obj_main->init_array =
obj_main->fini_array = (Elf_Addr)NULL;
}
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
wlock_acquire(rtld_bind_lock, &lockstate);
Add support for preinit, init and fini arrays. Some ABIs, in particular on ARM, do require working init arrays. Traditional FreeBSD crt1 calls _init and _fini of the binary, instead of allowing runtime linker to arrange the calls. This was probably done to have the same crt code serve both statically and dynamically linked binaries. Since ABI mandates that first is called preinit array functions, then init, and then init array functions, the init have to be called from rtld now. To provide binary compatibility to old FreeBSD crt1, which calls _init itself, rtld only calls intializers and finalizers for main binary if binary has a note indicating that new crt was used for linking. Add parsing of ELF notes to rtld, and cache p_osrel value since we parsed it anyway. The patch is inspired by init_array support for DragonflyBSD, written by John Marino. Reviewed by: kan Tested by: andrew (arm, previous version), flo (sparc64, previous version) MFC after: 3 weeks
2012-03-11 20:03:09 +00:00
if (obj_main->crt_no_init)
preinit_main();
Make sure internal rtld malloc routines are not called from unlocked contexts as rtld's malloc is not thread safe and is only supposed to be called with exclusive bind lock already held. The originating PR submitted a patch on top of different pre-requisite workaroud for unsafe dlopen calls, and the patch was midief slighlty to apply to stock sources for the purpose of this commit. Running rtld malloc from unlocked contexts is a bug on its own. PR: 126950 Submited by: Oleg Dolgov
2008-09-03 01:05:32 +00:00
objlist_call_init(&initlist, &lockstate);
Add a postinit debugger hook to rtld. This will be used by dtrace(1) to halt the victim process before its entry point is called, at which point probes and DOF data are registered with the kernel. The r_debug_state hook cannot be used for this purpose, as it is called before the program's init routines are invoked and in particular before DOF data is registered (via drti.o). Reviewed by: kib MFC after: 2 weeks
2014-05-06 18:07:58 +00:00
_r_debug_postinit(&obj_main->linkmap);
Revamp the code that calls shared libraries' init and fini functions. Formerly the init functions were called in the opposite of the order in which libraries were loaded, and libraries were loaded according to a breadth-first traversal of the dependency graph. That ordering came from SVR4.0, and it was easy to implement but not always sensible. Now we do a depth-first walk over the dependency graph and call the init functions in an order such that each shared object's needed objects are initialized before the shared object itself. At the same time we build a list of finalization (fini) functions in the opposite order, to guarantee correct C++ destructor ordering whenever possible. (It may not be possible if dlopen and dlclose are used in strange ways, but we come as close as one can come.) The need for this renovation has become apparent as more programs have started using multithreading. The multithreaded C library libc_r requires initialization, whereas the standard libc does not. Since virtually every other object depends on the C library, it is important that it get initialized first.
2000-07-26 04:24:40 +00:00
objlist_clear(&initlist);
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
dbg("loading filtees");
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
TAILQ_FOREACH(obj, &obj_list, next) {
if (obj->marker)
continue;
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
if (ld_loadfltr || obj->z_loadfltr)
load_filtees(obj, 0, &lockstate);
}
lock_release(rtld_bind_lock, &lockstate);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
dbg("transferring control to program entry point = %p", obj_main->entry);
/* Return the exit procedure and the program entry point. */
Add alpha support. Submitted by: John Birrell <jb@cimlogic.com.au> (with extra hacks by me) Obtained from: Probably NetBSD
1998-09-04 19:03:57 +00:00
*exit_proc = rtld_exit;
*objp = obj_main;
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
return (func_ptr_type) obj_main->entry;
}
Add support for STT_GNU_IFUNC and R_MACHINE_IRELATIVE GNU extensions to rtld on 386 and amd64. This adds runtime bits neccessary for the use of the dispatch functions from the dynamically-linked executables and shared libraries. To allow use of external references from the dispatch function, resolution of the R_MACHINE_IRESOLVE relocations in PLT is postponed until GOT entries for PLT are prepared, and normal resolution of the GOT entries is finished. Similar to how it is done by GNU, IRELATIVE relocations are resolved in advance, instead of normal lazy handling for PLT. Move the init_pltgot() call before the relocations for the object are processed. MFC after: 3 weeks
2011-12-12 11:03:14 +00:00
void *
rtld_resolve_ifunc(const Obj_Entry *obj, const Elf_Sym *def)
{
void *ptr;
Elf_Addr target;
ptr = (void *)make_function_pointer(def, obj);
Pass CPUID[1] %edx (cpu_feature), %ecx (cpu_feature2) and CPUID[7].%ebx (cpu_stdext_feature), %ecx (cpu_stdext_feature2) to the ifunc resolvers on x86. It is much more clean to use CPUID instruction in usermode to retrieve this information than to pass AT_HWCAP aux vector from kernel, on x86. Still, the change does allow for use of AT_HWCAP on arches where it is needed, by passing aux array to ifunc_init() initializer which should prepare arguments for ifunc resolvers. Current signature for resolvers on x86 is func_t iresolve(uint32_t cpu_feature, uint32_t cpu_feature2, uint32_t cpu_stdext_feature, uint32_t cpu_stdext_feature2); where arguments have identical meaning as the kernel variables of the same name. The ABIs allow to use resolvers with the void or shortened list of arguments. Reviewed by: jhb Sponsored by: The FreeBSD Foundation MFC after: 1 week Differential revision: https://reviews.freebsd.org/D8448
2016-11-15 09:43:26 +00:00
target = call_ifunc_resolver(ptr);
Add support for STT_GNU_IFUNC and R_MACHINE_IRELATIVE GNU extensions to rtld on 386 and amd64. This adds runtime bits neccessary for the use of the dispatch functions from the dynamically-linked executables and shared libraries. To allow use of external references from the dispatch function, resolution of the R_MACHINE_IRESOLVE relocations in PLT is postponed until GOT entries for PLT are prepared, and normal resolution of the GOT entries is finished. Similar to how it is done by GNU, IRELATIVE relocations are resolved in advance, instead of normal lazy handling for PLT. Move the init_pltgot() call before the relocations for the object are processed. MFC after: 3 weeks
2011-12-12 11:03:14 +00:00
return ((void *)target);
}
Fix _mips_rtld_bind() to handle ELF filters. MIPS does not use the common _rtld_bind() to handle runtime binding. Instead, it uses a private _mips_rtld_bind(). Update _mips_rtld_bind() to include the changes made to _rtld_bind() in r216695 and r218476 to support upgrading the read-locked rtld_bind_lock to a write lock when an object with a filter is encountered. While here, add a 'where' variable to track the location of the fixup in the GOT to make the code flow more closely match _rtld_bind(). Reviewed by: kib Obtained from: CheriBSD Sponsored by: DARPA / AFRL Differential Revision: https://reviews.freebsd.org/D8625
2016-11-23 20:21:53 +00:00
/*
* NB: MIPS uses a private version of this function (_mips_rtld_bind).
* Changes to this function should be applied there as well.
*/
Fix a serious performance bug for large programs on the Alpha, discovered by Hidetoshi Shimokawa. Large programs need multiple GOTs. The lazy binding stub in the PLT can be reached from any of these GOTs, but the dynamic linker only has enough information to fix up the first GOT entry. Thus calls through the other GOTs went through the time-consuming lazy binding process on every call. This fix rewrites the PLT entries themselves to bypass the lazy binding. Tested by Hidetoshi Shimokawa and Steve Price. Reviewed by: Doug Rabson <dfr@freebsd.org>
1999-06-25 02:53:59 +00:00
Elf_Addr
Make our ELF64 type definitions match standards. In particular this means: o Remove Elf64_Quarter, o Redefine Elf64_Half to be 16-bit, o Redefine Elf64_Word to be 32-bit, o Add Elf64_Xword and Elf64_Sxword for 64-bit entities, o Use Elf_Size in MI code to abstract the difference between Elf32_Word and Elf64_Word. o Add Elf_Ssize as the signed counterpart of Elf_Size. MFC after: 2 weeks
2005-12-18 04:52:37 +00:00
_rtld_bind(Obj_Entry *obj, Elf_Size reloff)
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
{
Add alpha support. Submitted by: John Birrell <jb@cimlogic.com.au> (with extra hacks by me) Obtained from: Probably NetBSD
1998-09-04 19:03:57 +00:00
const Elf_Rel *rel;
const Elf_Sym *def;
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
const Obj_Entry *defobj;
Add alpha support. Submitted by: John Birrell <jb@cimlogic.com.au> (with extra hacks by me) Obtained from: Probably NetBSD
1998-09-04 19:03:57 +00:00
Elf_Addr *where;
Fix a serious performance bug for large programs on the Alpha, discovered by Hidetoshi Shimokawa. Large programs need multiple GOTs. The lazy binding stub in the PLT can be reached from any of these GOTs, but the dynamic linker only has enough information to fix up the first GOT entry. Thus calls through the other GOTs went through the time-consuming lazy binding process on every call. This fix rewrites the PLT entries themselves to bypass the lazy binding. Tested by Hidetoshi Shimokawa and Steve Price. Reviewed by: Doug Rabson <dfr@freebsd.org>
1999-06-25 02:53:59 +00:00
Elf_Addr target;
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
RtldLockState lockstate;
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
rlock_acquire(rtld_bind_lock, &lockstate);
Use sigsetjmp/siglongjmp with disabled signal mask access for lock upgrade in rtld. There is no need to care about the mask, which causes a lot of unneeded sigprocmask(2) calls during each symbol lookup.
2011-02-09 09:20:27 +00:00
if (sigsetjmp(lockstate.env, 0) != 0)
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
lock_upgrade(rtld_bind_lock, &lockstate);
Add alpha support. Submitted by: John Birrell <jb@cimlogic.com.au> (with extra hacks by me) Obtained from: Probably NetBSD
1998-09-04 19:03:57 +00:00
if (obj->pltrel)
rel = (const Elf_Rel *) ((caddr_t) obj->pltrel + reloff);
else
rel = (const Elf_Rel *) ((caddr_t) obj->pltrela + reloff);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
Add alpha support. Submitted by: John Birrell <jb@cimlogic.com.au> (with extra hacks by me) Obtained from: Probably NetBSD
1998-09-04 19:03:57 +00:00
where = (Elf_Addr *) (obj->relocbase + rel->r_offset);
Pass the correct flag to find_symdef() from _rtld_bind(). When symbol versioning was added to rtld, the boolean 'in_plt' argument to find_symdef() was converted to a bitmask of flags. The first flag added was 'SYMLOOK_IN_PLT' which replaced the 'in_plt' bool. This happened to still work by accident as SYMLOOK_IN_PLT had the value of 1 which is the same as 'true', so there should be no functional change. Tested on: amd64 Reviewed by: kan MFC after: 2 weeks Sponsored by: DARPA / AFRL
2016-11-08 22:41:11 +00:00
def = find_symdef(ELF_R_SYM(rel->r_info), obj, &defobj, SYMLOOK_IN_PLT,
NULL, &lockstate);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
if (def == NULL)
Make die available as rtld_die for use by MD relocation code Reviewed by: kib Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D2213
2015-04-02 21:35:36 +00:00
rtld_die();
Add support for STT_GNU_IFUNC and R_MACHINE_IRELATIVE GNU extensions to rtld on 386 and amd64. This adds runtime bits neccessary for the use of the dispatch functions from the dynamically-linked executables and shared libraries. To allow use of external references from the dispatch function, resolution of the R_MACHINE_IRESOLVE relocations in PLT is postponed until GOT entries for PLT are prepared, and normal resolution of the GOT entries is finished. Similar to how it is done by GNU, IRELATIVE relocations are resolved in advance, instead of normal lazy handling for PLT. Move the init_pltgot() call before the relocations for the object are processed. MFC after: 3 weeks
2011-12-12 11:03:14 +00:00
if (ELF_ST_TYPE(def->st_info) == STT_GNU_IFUNC)
target = (Elf_Addr)rtld_resolve_ifunc(defobj, def);
else
target = (Elf_Addr)(defobj->relocbase + def->st_value);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
dbg("\"%s\" in \"%s\" ==> %p in \"%s\"",
defobj->strtab + def->st_name, basename(obj->path),
Fix a serious performance bug for large programs on the Alpha, discovered by Hidetoshi Shimokawa. Large programs need multiple GOTs. The lazy binding stub in the PLT can be reached from any of these GOTs, but the dynamic linker only has enough information to fix up the first GOT entry. Thus calls through the other GOTs went through the time-consuming lazy binding process on every call. This fix rewrites the PLT entries themselves to bypass the lazy binding. Tested by Hidetoshi Shimokawa and Steve Price. Reviewed by: Doug Rabson <dfr@freebsd.org>
1999-06-25 02:53:59 +00:00
(void *)target, basename(defobj->path));
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
Add ia64 support. Various adjustments were made to existing targets to cope with a few interface changes required by the ia64. In particular, function pointers on ia64 need special treatment in rtld.
2001-10-15 18:48:42 +00:00
/*
* Write the new contents for the jmpslot. Note that depending on
* architecture, the value which we need to return back to the
* lazy binding trampoline may or may not be the target
* address. The value returned from reloc_jmpslot() is the value
* that the trampoline needs.
*/
Fix the handling of high PLT entries (> 32764) on sparc64. This requires additional arguments to reloc_jmpslot(), which is why MI code and MD code of other platforms had to be changed. Reviewed by: jake Approved by: re
2002-11-18 22:08:50 +00:00
target = reloc_jmpslot(where, target, defobj, obj, rel);
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
lock_release(rtld_bind_lock, &lockstate);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
return target;
}
/*
* Error reporting function. Use it like printf. If formats the message
* into a buffer, and sets things up so that the next call to dlerror()
* will return the message.
*/
void
_rtld_error(const char *fmt, ...)
{
static char buf[512];
va_list ap;
va_start(ap, fmt);
Rtld links with the specially built pic static libc library to get some C runtime services, like printf(). Unfortunately, the multithread-safeness measures in the libc do not work in rtld environment. Rip the kernel printf() implementation and use it in the rtld instead of libc version. This printf does not require any shared global data and thus is mt-safe. Systematically use rtld_printf() and related functions, remove the calls to err(3). Note that stdio is still pulled from libc due to libmap implementaion using fopen(). This is safe but unoptimal, and can be changed later. Reported and tested by: pgj Diagnosed and reviewed by: kan (previous version) Approved by: re (bz)
2011-08-24 20:05:13 +00:00
rtld_vsnprintf(buf, sizeof buf, fmt, ap);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
error_message = buf;
va_end(ap);
Extend LD_UTRACE by also generating utrace(2) log events for runtime linker errors. Reviewed by: kib, jhb Approved by: jhb(mentor) MFC after: 1 week Differential Revision: D9347
2017-01-30 08:38:32 +00:00
LD_UTRACE(UTRACE_RTLD_ERROR, NULL, NULL, 0, 0, error_message);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
}
Fix a bug in which a program called dlclose from a destructor and got an assert failure in the dynamic linker.
2001-01-05 04:36:17 +00:00
/*
* Return a dynamically-allocated copy of the current error message, if any.
*/
static char *
errmsg_save(void)
{
return error_message == NULL ? NULL : xstrdup(error_message);
}
/*
* Restore the current error message from a copy which was previously saved
* by errmsg_save(). The copy is freed.
*/
static void
errmsg_restore(char *saved_msg)
{
if (saved_msg == NULL)
error_message = NULL;
else {
_rtld_error("%s", saved_msg);
free(saved_msg);
}
}
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
static const char *
basename(const char *name)
{
const char *p = strrchr(name, '/');
return p != NULL ? p + 1 : name;
}
Implement the dynamic string token substitution in the rpath and soneeded pathes. The $ORIGIN, $OSNAME, $OSREL and $PLATFORM tokens are supported. Enabling the substitution requires DF_ORIGIN flag in DT_FLAGS or DF_1_ORIGIN if DF_FLAGS_1, that may be set with -z origin gnu ld flag. Translation is unconditionally disabled for setuid/setgid processes. The $ORIGIN translation relies on the AT_EXECPATH auxinfo supplied by kernel. Requested by: maho Tested by: maho, pho Reviewed by: kan
2009-03-18 13:40:37 +00:00
static struct utsname uts;
The origin_subst_one() function limits the length of the string to PATH_MAX after the token substitution. This is wrong, because origin_subst_one() performs the substitution on the whole rpath and similar strings, which contain several pathes separated by colon. As result, long (but correct) rpath consisting of many path elements is rejected by the function. Correct the problem by rewriting the origin_subst_one() to perform two passes, first pass to calculate the number of substitutions to be performed, and second pass to generate the resulting string. Second pass allocates the memory for the result based on the count from the first pass, without enforcing a limit. Reported and tested by: pgj Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2013-04-15 20:06:56 +00:00
static char *
Change interpretation of the DF_ORIGIN and DF_1_ORIGIN flags. According to standard, the presence of the flags only means that the object path must be resolved at the time object loading, instead of my reading that the flag is required to enable token substitution at all. The consequence is that -z origin linker flag is no longer required for the token substitution in the run/rpath or the needed library soname. It is only recommended if token substition is needed at dlopen(3) time, since namecache might drop the required entries at the time of resolution. Found, reviewed and tested by: emaste Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-04-27 18:41:31 +00:00
origin_subst_one(Obj_Entry *obj, char *real, const char *kw,
const char *subst, bool may_free)
Implement the dynamic string token substitution in the rpath and soneeded pathes. The $ORIGIN, $OSNAME, $OSREL and $PLATFORM tokens are supported. Enabling the substitution requires DF_ORIGIN flag in DT_FLAGS or DF_1_ORIGIN if DF_FLAGS_1, that may be set with -z origin gnu ld flag. Translation is unconditionally disabled for setuid/setgid processes. The $ORIGIN translation relies on the AT_EXECPATH auxinfo supplied by kernel. Requested by: maho Tested by: maho, pho Reviewed by: kan
2009-03-18 13:40:37 +00:00
{
The origin_subst_one() function limits the length of the string to PATH_MAX after the token substitution. This is wrong, because origin_subst_one() performs the substitution on the whole rpath and similar strings, which contain several pathes separated by colon. As result, long (but correct) rpath consisting of many path elements is rejected by the function. Correct the problem by rewriting the origin_subst_one() to perform two passes, first pass to calculate the number of substitutions to be performed, and second pass to generate the resulting string. Second pass allocates the memory for the result based on the count from the first pass, without enforcing a limit. Reported and tested by: pgj Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2013-04-15 20:06:56 +00:00
char *p, *p1, *res, *resp;
int subst_len, kw_len, subst_count, old_len, new_len;
kw_len = strlen(kw);
/*
libexec: minor spelling fixes in comments. No functional change.
2016-05-01 19:39:23 +00:00
* First, count the number of the keyword occurrences, to
The origin_subst_one() function limits the length of the string to PATH_MAX after the token substitution. This is wrong, because origin_subst_one() performs the substitution on the whole rpath and similar strings, which contain several pathes separated by colon. As result, long (but correct) rpath consisting of many path elements is rejected by the function. Correct the problem by rewriting the origin_subst_one() to perform two passes, first pass to calculate the number of substitutions to be performed, and second pass to generate the resulting string. Second pass allocates the memory for the result based on the count from the first pass, without enforcing a limit. Reported and tested by: pgj Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2013-04-15 20:06:56 +00:00
* preallocate the final string.
*/
for (p = real, subst_count = 0;; p = p1 + kw_len, subst_count++) {
p1 = strstr(p, kw);
if (p1 == NULL)
break;
}
/*
* If the keyword is not found, just return.
Change interpretation of the DF_ORIGIN and DF_1_ORIGIN flags. According to standard, the presence of the flags only means that the object path must be resolved at the time object loading, instead of my reading that the flag is required to enable token substitution at all. The consequence is that -z origin linker flag is no longer required for the token substitution in the run/rpath or the needed library soname. It is only recommended if token substition is needed at dlopen(3) time, since namecache might drop the required entries at the time of resolution. Found, reviewed and tested by: emaste Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-04-27 18:41:31 +00:00
*
* Return non-substituted string if resolution failed. We
* cannot do anything more reasonable, the failure mode of the
* caller is unresolved library anyway.
The origin_subst_one() function limits the length of the string to PATH_MAX after the token substitution. This is wrong, because origin_subst_one() performs the substitution on the whole rpath and similar strings, which contain several pathes separated by colon. As result, long (but correct) rpath consisting of many path elements is rejected by the function. Correct the problem by rewriting the origin_subst_one() to perform two passes, first pass to calculate the number of substitutions to be performed, and second pass to generate the resulting string. Second pass allocates the memory for the result based on the count from the first pass, without enforcing a limit. Reported and tested by: pgj Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2013-04-15 20:06:56 +00:00
*/
Change interpretation of the DF_ORIGIN and DF_1_ORIGIN flags. According to standard, the presence of the flags only means that the object path must be resolved at the time object loading, instead of my reading that the flag is required to enable token substitution at all. The consequence is that -z origin linker flag is no longer required for the token substitution in the run/rpath or the needed library soname. It is only recommended if token substition is needed at dlopen(3) time, since namecache might drop the required entries at the time of resolution. Found, reviewed and tested by: emaste Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-04-27 18:41:31 +00:00
if (subst_count == 0 || (obj != NULL && !obj_resolve_origin(obj)))
The origin_subst_one() function limits the length of the string to PATH_MAX after the token substitution. This is wrong, because origin_subst_one() performs the substitution on the whole rpath and similar strings, which contain several pathes separated by colon. As result, long (but correct) rpath consisting of many path elements is rejected by the function. Correct the problem by rewriting the origin_subst_one() to perform two passes, first pass to calculate the number of substitutions to be performed, and second pass to generate the resulting string. Second pass allocates the memory for the result based on the count from the first pass, without enforcing a limit. Reported and tested by: pgj Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2013-04-15 20:06:56 +00:00
return (may_free ? real : xstrdup(real));
Change interpretation of the DF_ORIGIN and DF_1_ORIGIN flags. According to standard, the presence of the flags only means that the object path must be resolved at the time object loading, instead of my reading that the flag is required to enable token substitution at all. The consequence is that -z origin linker flag is no longer required for the token substitution in the run/rpath or the needed library soname. It is only recommended if token substition is needed at dlopen(3) time, since namecache might drop the required entries at the time of resolution. Found, reviewed and tested by: emaste Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-04-27 18:41:31 +00:00
if (obj != NULL)
subst = obj->origin_path;
The origin_subst_one() function limits the length of the string to PATH_MAX after the token substitution. This is wrong, because origin_subst_one() performs the substitution on the whole rpath and similar strings, which contain several pathes separated by colon. As result, long (but correct) rpath consisting of many path elements is rejected by the function. Correct the problem by rewriting the origin_subst_one() to perform two passes, first pass to calculate the number of substitutions to be performed, and second pass to generate the resulting string. Second pass allocates the memory for the result based on the count from the first pass, without enforcing a limit. Reported and tested by: pgj Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2013-04-15 20:06:56 +00:00
/*
* There is indeed something to substitute. Calculate the
* length of the resulting string, and allocate it.
*/
subst_len = strlen(subst);
old_len = strlen(real);
new_len = old_len + (subst_len - kw_len) * subst_count;
res = xmalloc(new_len + 1);
/*
* Now, execute the substitution loop.
*/
Properly terminate the result string for intermediate results, to allow the final strcpy() to start at the intended place. Reported and tested by: pgj Pointy hat to: kib MFC after: 3 days
2013-04-29 21:12:25 +00:00
for (p = real, resp = res, *resp = '\0';;) {
The origin_subst_one() function limits the length of the string to PATH_MAX after the token substitution. This is wrong, because origin_subst_one() performs the substitution on the whole rpath and similar strings, which contain several pathes separated by colon. As result, long (but correct) rpath consisting of many path elements is rejected by the function. Correct the problem by rewriting the origin_subst_one() to perform two passes, first pass to calculate the number of substitutions to be performed, and second pass to generate the resulting string. Second pass allocates the memory for the result based on the count from the first pass, without enforcing a limit. Reported and tested by: pgj Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2013-04-15 20:06:56 +00:00
p1 = strstr(p, kw);
if (p1 != NULL) {
/* Copy the prefix before keyword. */
memcpy(resp, p, p1 - p);
resp += p1 - p;
/* Keyword replacement. */
memcpy(resp, subst, subst_len);
resp += subst_len;
Properly terminate the result string for intermediate results, to allow the final strcpy() to start at the intended place. Reported and tested by: pgj Pointy hat to: kib MFC after: 3 days
2013-04-29 21:12:25 +00:00
*resp = '\0';
The origin_subst_one() function limits the length of the string to PATH_MAX after the token substitution. This is wrong, because origin_subst_one() performs the substitution on the whole rpath and similar strings, which contain several pathes separated by colon. As result, long (but correct) rpath consisting of many path elements is rejected by the function. Correct the problem by rewriting the origin_subst_one() to perform two passes, first pass to calculate the number of substitutions to be performed, and second pass to generate the resulting string. Second pass allocates the memory for the result based on the count from the first pass, without enforcing a limit. Reported and tested by: pgj Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2013-04-15 20:06:56 +00:00
p = p1 + kw_len;
} else
break;
}
/* Copy to the end of string and finish. */
strcat(resp, p);
if (may_free)
free(real);
return (res);
Implement the dynamic string token substitution in the rpath and soneeded pathes. The $ORIGIN, $OSNAME, $OSREL and $PLATFORM tokens are supported. Enabling the substitution requires DF_ORIGIN flag in DT_FLAGS or DF_1_ORIGIN if DF_FLAGS_1, that may be set with -z origin gnu ld flag. Translation is unconditionally disabled for setuid/setgid processes. The $ORIGIN translation relies on the AT_EXECPATH auxinfo supplied by kernel. Requested by: maho Tested by: maho, pho Reviewed by: kan
2009-03-18 13:40:37 +00:00
}
static char *
Change interpretation of the DF_ORIGIN and DF_1_ORIGIN flags. According to standard, the presence of the flags only means that the object path must be resolved at the time object loading, instead of my reading that the flag is required to enable token substitution at all. The consequence is that -z origin linker flag is no longer required for the token substitution in the run/rpath or the needed library soname. It is only recommended if token substition is needed at dlopen(3) time, since namecache might drop the required entries at the time of resolution. Found, reviewed and tested by: emaste Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-04-27 18:41:31 +00:00
origin_subst(Obj_Entry *obj, char *real)
Implement the dynamic string token substitution in the rpath and soneeded pathes. The $ORIGIN, $OSNAME, $OSREL and $PLATFORM tokens are supported. Enabling the substitution requires DF_ORIGIN flag in DT_FLAGS or DF_1_ORIGIN if DF_FLAGS_1, that may be set with -z origin gnu ld flag. Translation is unconditionally disabled for setuid/setgid processes. The $ORIGIN translation relies on the AT_EXECPATH auxinfo supplied by kernel. Requested by: maho Tested by: maho, pho Reviewed by: kan
2009-03-18 13:40:37 +00:00
{
The origin_subst_one() function limits the length of the string to PATH_MAX after the token substitution. This is wrong, because origin_subst_one() performs the substitution on the whole rpath and similar strings, which contain several pathes separated by colon. As result, long (but correct) rpath consisting of many path elements is rejected by the function. Correct the problem by rewriting the origin_subst_one() to perform two passes, first pass to calculate the number of substitutions to be performed, and second pass to generate the resulting string. Second pass allocates the memory for the result based on the count from the first pass, without enforcing a limit. Reported and tested by: pgj Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2013-04-15 20:06:56 +00:00
char *res1, *res2, *res3, *res4;
Implement the dynamic string token substitution in the rpath and soneeded pathes. The $ORIGIN, $OSNAME, $OSREL and $PLATFORM tokens are supported. Enabling the substitution requires DF_ORIGIN flag in DT_FLAGS or DF_1_ORIGIN if DF_FLAGS_1, that may be set with -z origin gnu ld flag. Translation is unconditionally disabled for setuid/setgid processes. The $ORIGIN translation relies on the AT_EXECPATH auxinfo supplied by kernel. Requested by: maho Tested by: maho, pho Reviewed by: kan
2009-03-18 13:40:37 +00:00
Change interpretation of the DF_ORIGIN and DF_1_ORIGIN flags. According to standard, the presence of the flags only means that the object path must be resolved at the time object loading, instead of my reading that the flag is required to enable token substitution at all. The consequence is that -z origin linker flag is no longer required for the token substitution in the run/rpath or the needed library soname. It is only recommended if token substition is needed at dlopen(3) time, since namecache might drop the required entries at the time of resolution. Found, reviewed and tested by: emaste Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-04-27 18:41:31 +00:00
if (obj == NULL || !trust)
return (xstrdup(real));
The origin_subst_one() function limits the length of the string to PATH_MAX after the token substitution. This is wrong, because origin_subst_one() performs the substitution on the whole rpath and similar strings, which contain several pathes separated by colon. As result, long (but correct) rpath consisting of many path elements is rejected by the function. Correct the problem by rewriting the origin_subst_one() to perform two passes, first pass to calculate the number of substitutions to be performed, and second pass to generate the resulting string. Second pass allocates the memory for the result based on the count from the first pass, without enforcing a limit. Reported and tested by: pgj Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2013-04-15 20:06:56 +00:00
if (uts.sysname[0] == '\0') {
if (uname(&uts) != 0) {
_rtld_error("utsname failed: %d", errno);
return (NULL);
}
Implement the dynamic string token substitution in the rpath and soneeded pathes. The $ORIGIN, $OSNAME, $OSREL and $PLATFORM tokens are supported. Enabling the substitution requires DF_ORIGIN flag in DT_FLAGS or DF_1_ORIGIN if DF_FLAGS_1, that may be set with -z origin gnu ld flag. Translation is unconditionally disabled for setuid/setgid processes. The $ORIGIN translation relies on the AT_EXECPATH auxinfo supplied by kernel. Requested by: maho Tested by: maho, pho Reviewed by: kan
2009-03-18 13:40:37 +00:00
}
Change interpretation of the DF_ORIGIN and DF_1_ORIGIN flags. According to standard, the presence of the flags only means that the object path must be resolved at the time object loading, instead of my reading that the flag is required to enable token substitution at all. The consequence is that -z origin linker flag is no longer required for the token substitution in the run/rpath or the needed library soname. It is only recommended if token substition is needed at dlopen(3) time, since namecache might drop the required entries at the time of resolution. Found, reviewed and tested by: emaste Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-04-27 18:41:31 +00:00
res1 = origin_subst_one(obj, real, "$ORIGIN", NULL, false);
res2 = origin_subst_one(NULL, res1, "$OSNAME", uts.sysname, true);
res3 = origin_subst_one(NULL, res2, "$OSREL", uts.release, true);
res4 = origin_subst_one(NULL, res3, "$PLATFORM", uts.machine, true);
The origin_subst_one() function limits the length of the string to PATH_MAX after the token substitution. This is wrong, because origin_subst_one() performs the substitution on the whole rpath and similar strings, which contain several pathes separated by colon. As result, long (but correct) rpath consisting of many path elements is rejected by the function. Correct the problem by rewriting the origin_subst_one() to perform two passes, first pass to calculate the number of substitutions to be performed, and second pass to generate the resulting string. Second pass allocates the memory for the result based on the count from the first pass, without enforcing a limit. Reported and tested by: pgj Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2013-04-15 20:06:56 +00:00
return (res4);
Implement the dynamic string token substitution in the rpath and soneeded pathes. The $ORIGIN, $OSNAME, $OSREL and $PLATFORM tokens are supported. Enabling the substitution requires DF_ORIGIN flag in DT_FLAGS or DF_1_ORIGIN if DF_FLAGS_1, that may be set with -z origin gnu ld flag. Translation is unconditionally disabled for setuid/setgid processes. The $ORIGIN translation relies on the AT_EXECPATH auxinfo supplied by kernel. Requested by: maho Tested by: maho, pho Reviewed by: kan
2009-03-18 13:40:37 +00:00
}
Make die available as rtld_die for use by MD relocation code Reviewed by: kib Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D2213
2015-04-02 21:35:36 +00:00
void
rtld_die(void)
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
{
const char *msg = dlerror();
if (msg == NULL)
msg = "Fatal error";
Rtld links with the specially built pic static libc library to get some C runtime services, like printf(). Unfortunately, the multithread-safeness measures in the libc do not work in rtld environment. Rip the kernel printf() implementation and use it in the rtld instead of libc version. This printf does not require any shared global data and thus is mt-safe. Systematically use rtld_printf() and related functions, remove the calls to err(3). Note that stdio is still pulled from libc due to libmap implementaion using fopen(). This is safe but unoptimal, and can be changed later. Reported and tested by: pgj Diagnosed and reviewed by: kan (previous version) Approved by: re (bz)
2011-08-24 20:05:13 +00:00
rtld_fdputstr(STDERR_FILENO, msg);
Add missed EOL when die() was converted to use rtld_fdputstr() instead of errx(). Reported by: amdmi3 PR: bin/165075 MFC after: 3 days
2012-02-13 11:15:29 +00:00
rtld_fdputchar(STDERR_FILENO, '\n');
Rtld links with the specially built pic static libc library to get some C runtime services, like printf(). Unfortunately, the multithread-safeness measures in the libc do not work in rtld environment. Rip the kernel printf() implementation and use it in the rtld instead of libc version. This printf does not require any shared global data and thus is mt-safe. Systematically use rtld_printf() and related functions, remove the calls to err(3). Note that stdio is still pulled from libc due to libmap implementaion using fopen(). This is safe but unoptimal, and can be changed later. Reported and tested by: pgj Diagnosed and reviewed by: kan (previous version) Approved by: re (bz)
2011-08-24 20:05:13 +00:00
_exit(1);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
}
/*
* Process a shared object's DYNAMIC section, and save the important
* information in its Obj_Entry structure.
*/
static void
Use the newly provided aux vectors to get pagesize and osreldate information. Use local version of getpagesize(), rtld_getpagesize() in private allocator. Override the __getosreldate() previously fetched from libc_pic.a with local version that uses aux value if present. Note that __getosreldate() is used by rtld indirectly, by mmap(2) libc wrapper. To be able to utilize aux, split digest_dynamic() for use by init_rtld() into two parts, where the first one does not call malloc(), and the second part uses it. init_rtld() is able to initialize global variables before digest_dynamic2() calls. In particular, pagesize and osreldate are set up from the aux values. Now, rtld avoids (two) sysctl calls in startup. Tested by: marius (sparc64) MFC after: 1 month
2010-08-17 09:05:39 +00:00
digest_dynamic1(Obj_Entry *obj, int early, const Elf_Dyn **dyn_rpath,
Import the DragonFly BSD commit 4f0bc915b65fcf5a23214f6d221d65c80be68ad4 by John Marino <draco@marino.st>, with the following (edited) commit message Date: Sat, 24 Mar 2012 06:40:50 +0100 Subject: [PATCH 1/1] rtld: Implement DT_RUNPATH and -z nodefaultlib DT_RUNPATH is incorrectly being considered as an alias of DT_RPATH. The purpose of DT_RUNPATH is to have two different types of rpath: one that can be overridden by the environment variable LD_LIBRARY_PATH and one that can't. With the currently implementation, LD_LIBRARY_PATH will always trump any embedded rpath or runpath tags. Current path search order by rtld: ================================== LD_LIBRARY_PATH DT_RPATH / DT_RUNPATH (always the same) ldconfig hints file (default: /var/run/ld-elf.so.hints) /usr/lib New path search order by rtld: ============================== DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file /usr/lib The new path search matches how the linux runtime loader works. The other major added feature is support for linker flag "-z nodefaultlib". When this flag is passed to the linker, rtld will skip all references to the standard library search path ("/usr/lib" in this case but it could handle more color delimited paths) except in DT_RPATH and DT_RUNPATH. New path search order by rtld with -z nodefaultlib flag set: ============================================================ DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file (skips all references to /usr/lib) FreeBSD notes: - we fixed some bugs which were submitted to DragonFly and merged there as commit 1ff8a2bd3eb6e5587174c6a983303ea3a79e0002; - we added LD_LIBRARY_PATH_RPATH environment variable to switch to the previous behaviour of considering DT_RPATH a synonym for DT_RUNPATH; - the FreeBSD default search path is /lib:/usr/lib and not /usr/lib. Reviewed by: kan MFC after: 1 month MFC note: flip the ld_library_path_rpath default value for stable/9
2012-07-15 10:53:48 +00:00
const Elf_Dyn **dyn_soname, const Elf_Dyn **dyn_runpath)
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
{
Add alpha support. Submitted by: John Birrell <jb@cimlogic.com.au> (with extra hacks by me) Obtained from: Probably NetBSD
1998-09-04 19:03:57 +00:00
const Elf_Dyn *dynp;
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
Needed_Entry **needed_tail = &obj->needed;
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
Needed_Entry **needed_filtees_tail = &obj->needed_filtees;
Needed_Entry **needed_aux_filtees_tail = &obj->needed_aux_filtees;
Add GNU hash support for rtld. Based on dragonflybsd support for GNU hash by John Marino <draco marino st> Reviewed by: kan Tested by: bapt MFC after: 2 weeks
2012-04-30 13:31:10 +00:00
const Elf_Hashelt *hashtab;
const Elf32_Word *hashval;
Elf32_Word bkt, nmaskwords;
int bloom_size32;
Add alpha support. Submitted by: John Birrell <jb@cimlogic.com.au> (with extra hacks by me) Obtained from: Probably NetBSD
1998-09-04 19:03:57 +00:00
int plttype = DT_REL;
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
Use the newly provided aux vectors to get pagesize and osreldate information. Use local version of getpagesize(), rtld_getpagesize() in private allocator. Override the __getosreldate() previously fetched from libc_pic.a with local version that uses aux value if present. Note that __getosreldate() is used by rtld indirectly, by mmap(2) libc wrapper. To be able to utilize aux, split digest_dynamic() for use by init_rtld() into two parts, where the first one does not call malloc(), and the second part uses it. init_rtld() is able to initialize global variables before digest_dynamic2() calls. In particular, pagesize and osreldate are set up from the aux values. Now, rtld avoids (two) sysctl calls in startup. Tested by: marius (sparc64) MFC after: 1 month
2010-08-17 09:05:39 +00:00
*dyn_rpath = NULL;
*dyn_soname = NULL;
Import the DragonFly BSD commit 4f0bc915b65fcf5a23214f6d221d65c80be68ad4 by John Marino <draco@marino.st>, with the following (edited) commit message Date: Sat, 24 Mar 2012 06:40:50 +0100 Subject: [PATCH 1/1] rtld: Implement DT_RUNPATH and -z nodefaultlib DT_RUNPATH is incorrectly being considered as an alias of DT_RPATH. The purpose of DT_RUNPATH is to have two different types of rpath: one that can be overridden by the environment variable LD_LIBRARY_PATH and one that can't. With the currently implementation, LD_LIBRARY_PATH will always trump any embedded rpath or runpath tags. Current path search order by rtld: ================================== LD_LIBRARY_PATH DT_RPATH / DT_RUNPATH (always the same) ldconfig hints file (default: /var/run/ld-elf.so.hints) /usr/lib New path search order by rtld: ============================== DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file /usr/lib The new path search matches how the linux runtime loader works. The other major added feature is support for linker flag "-z nodefaultlib". When this flag is passed to the linker, rtld will skip all references to the standard library search path ("/usr/lib" in this case but it could handle more color delimited paths) except in DT_RPATH and DT_RUNPATH. New path search order by rtld with -z nodefaultlib flag set: ============================================================ DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file (skips all references to /usr/lib) FreeBSD notes: - we fixed some bugs which were submitted to DragonFly and merged there as commit 1ff8a2bd3eb6e5587174c6a983303ea3a79e0002; - we added LD_LIBRARY_PATH_RPATH environment variable to switch to the previous behaviour of considering DT_RPATH a synonym for DT_RUNPATH; - the FreeBSD default search path is /lib:/usr/lib and not /usr/lib. Reviewed by: kan MFC after: 1 month MFC note: flip the ld_library_path_rpath default value for stable/9
2012-07-15 10:53:48 +00:00
*dyn_runpath = NULL;
Use the newly provided aux vectors to get pagesize and osreldate information. Use local version of getpagesize(), rtld_getpagesize() in private allocator. Override the __getosreldate() previously fetched from libc_pic.a with local version that uses aux value if present. Note that __getosreldate() is used by rtld indirectly, by mmap(2) libc wrapper. To be able to utilize aux, split digest_dynamic() for use by init_rtld() into two parts, where the first one does not call malloc(), and the second part uses it. init_rtld() is able to initialize global variables before digest_dynamic2() calls. In particular, pagesize and osreldate are set up from the aux values. Now, rtld avoids (two) sysctl calls in startup. Tested by: marius (sparc64) MFC after: 1 month
2010-08-17 09:05:39 +00:00
Initialise some uninitialised variables. Thanks to: valgrind
2003-12-31 15:10:41 +00:00
obj->bind_now = false;
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
for (dynp = obj->dynamic; dynp->d_tag != DT_NULL; dynp++) {
switch (dynp->d_tag) {
case DT_REL:
Add alpha support. Submitted by: John Birrell <jb@cimlogic.com.au> (with extra hacks by me) Obtained from: Probably NetBSD
1998-09-04 19:03:57 +00:00
obj->rel = (const Elf_Rel *) (obj->relocbase + dynp->d_un.d_ptr);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
break;
case DT_RELSZ:
obj->relsize = dynp->d_un.d_val;
break;
case DT_RELENT:
Add alpha support. Submitted by: John Birrell <jb@cimlogic.com.au> (with extra hacks by me) Obtained from: Probably NetBSD
1998-09-04 19:03:57 +00:00
assert(dynp->d_un.d_val == sizeof(Elf_Rel));
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
break;
case DT_JMPREL:
Add alpha support. Submitted by: John Birrell <jb@cimlogic.com.au> (with extra hacks by me) Obtained from: Probably NetBSD
1998-09-04 19:03:57 +00:00
obj->pltrel = (const Elf_Rel *)
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
(obj->relocbase + dynp->d_un.d_ptr);
break;
case DT_PLTRELSZ:
obj->pltrelsize = dynp->d_un.d_val;
break;
case DT_RELA:
Add alpha support. Submitted by: John Birrell <jb@cimlogic.com.au> (with extra hacks by me) Obtained from: Probably NetBSD
1998-09-04 19:03:57 +00:00
obj->rela = (const Elf_Rela *) (obj->relocbase + dynp->d_un.d_ptr);
break;
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
case DT_RELASZ:
Add alpha support. Submitted by: John Birrell <jb@cimlogic.com.au> (with extra hacks by me) Obtained from: Probably NetBSD
1998-09-04 19:03:57 +00:00
obj->relasize = dynp->d_un.d_val;
break;
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
case DT_RELAENT:
Add alpha support. Submitted by: John Birrell <jb@cimlogic.com.au> (with extra hacks by me) Obtained from: Probably NetBSD
1998-09-04 19:03:57 +00:00
assert(dynp->d_un.d_val == sizeof(Elf_Rela));
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
break;
case DT_PLTREL:
Add alpha support. Submitted by: John Birrell <jb@cimlogic.com.au> (with extra hacks by me) Obtained from: Probably NetBSD
1998-09-04 19:03:57 +00:00
plttype = dynp->d_un.d_val;
assert(dynp->d_un.d_val == DT_REL || plttype == DT_RELA);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
break;
case DT_SYMTAB:
Add alpha support. Submitted by: John Birrell <jb@cimlogic.com.au> (with extra hacks by me) Obtained from: Probably NetBSD
1998-09-04 19:03:57 +00:00
obj->symtab = (const Elf_Sym *)
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
(obj->relocbase + dynp->d_un.d_ptr);
break;
case DT_SYMENT:
Add alpha support. Submitted by: John Birrell <jb@cimlogic.com.au> (with extra hacks by me) Obtained from: Probably NetBSD
1998-09-04 19:03:57 +00:00
assert(dynp->d_un.d_val == sizeof(Elf_Sym));
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
break;
case DT_STRTAB:
obj->strtab = (const char *) (obj->relocbase + dynp->d_un.d_ptr);
break;
case DT_STRSZ:
obj->strsize = dynp->d_un.d_val;
break;
Implement ELF symbol versioning using GNU semantics. This code aims to be compatible with symbol versioning support as implemented by GNU libc and documented by http://people.redhat.com/~drepper/symbol-versioning and LSB 3.0. Implement dlvsym() function to allow lookups for a specific version of a given symbol.
2005-12-18 19:43:33 +00:00
case DT_VERNEED:
obj->verneed = (const Elf_Verneed *) (obj->relocbase +
dynp->d_un.d_val);
break;
case DT_VERNEEDNUM:
obj->verneednum = dynp->d_un.d_val;
break;
case DT_VERDEF:
obj->verdef = (const Elf_Verdef *) (obj->relocbase +
dynp->d_un.d_val);
break;
case DT_VERDEFNUM:
obj->verdefnum = dynp->d_un.d_val;
break;
case DT_VERSYM:
obj->versyms = (const Elf_Versym *)(obj->relocbase +
dynp->d_un.d_val);
break;
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
case DT_HASH:
{
Add GNU hash support for rtld. Based on dragonflybsd support for GNU hash by John Marino <draco marino st> Reviewed by: kan Tested by: bapt MFC after: 2 weeks
2012-04-30 13:31:10 +00:00
hashtab = (const Elf_Hashelt *)(obj->relocbase +
dynp->d_un.d_ptr);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
obj->nbuckets = hashtab[0];
obj->nchains = hashtab[1];
obj->buckets = hashtab + 2;
obj->chains = obj->buckets + obj->nbuckets;
Add GNU hash support for rtld. Based on dragonflybsd support for GNU hash by John Marino <draco marino st> Reviewed by: kan Tested by: bapt MFC after: 2 weeks
2012-04-30 13:31:10 +00:00
obj->valid_hash_sysv = obj->nbuckets > 0 && obj->nchains > 0 &&
obj->buckets != NULL;
}
break;
case DT_GNU_HASH:
{
hashtab = (const Elf_Hashelt *)(obj->relocbase +
dynp->d_un.d_ptr);
obj->nbuckets_gnu = hashtab[0];
obj->symndx_gnu = hashtab[1];
nmaskwords = hashtab[2];
bloom_size32 = (__ELF_WORD_SIZE / 32) * nmaskwords;
obj->maskwords_bm_gnu = nmaskwords - 1;
obj->shift2_gnu = hashtab[3];
obj->bloom_gnu = (Elf_Addr *) (hashtab + 4);
obj->buckets_gnu = hashtab + 4 + bloom_size32;
obj->chain_zero_gnu = obj->buckets_gnu + obj->nbuckets_gnu -
obj->symndx_gnu;
Use powerof2(). Remove single-use variable. Submitted by: Conrad Meyer Differential Revision: https://reviews.freebsd.org/D1724 MFC after: 1 week
2015-01-30 15:32:35 +00:00
/* Number of bitmask words is required to be power of 2 */
obj->valid_hash_gnu = powerof2(nmaskwords) &&
obj->nbuckets_gnu > 0 && obj->buckets_gnu != NULL;
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
}
break;
case DT_NEEDED:
Eliminate all machine-dependent code from the main source body and the Makefile, and move it down into the architecture-specific subdirectories. Eliminate an asm() statement for the i386. Make the dynamic linker work if it is built as an executable instead of as a shared library. See i386/Makefile.inc to find out how to do it. Note, this change is not enabled and it might never be enabled. But it might be useful in the future. Building the dynamic linker as an executable should make it start up faster, because it won't have any relocations. But in practice I suspect the difference is negligible.
1999-04-09 00:28:43 +00:00
if (!obj->rtld) {
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
Needed_Entry *nep = NEW(Needed_Entry);
nep->name = dynp->d_un.d_val;
nep->obj = NULL;
nep->next = NULL;
*needed_tail = nep;
needed_tail = &nep->next;
}
break;
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
case DT_FILTER:
if (!obj->rtld) {
Needed_Entry *nep = NEW(Needed_Entry);
nep->name = dynp->d_un.d_val;
nep->obj = NULL;
nep->next = NULL;
*needed_filtees_tail = nep;
needed_filtees_tail = &nep->next;
}
break;
case DT_AUXILIARY:
if (!obj->rtld) {
Needed_Entry *nep = NEW(Needed_Entry);
nep->name = dynp->d_un.d_val;
nep->obj = NULL;
nep->next = NULL;
*needed_aux_filtees_tail = nep;
needed_aux_filtees_tail = &nep->next;
}
break;
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
case DT_PLTGOT:
Eliminate all machine-dependent code from the main source body and the Makefile, and move it down into the architecture-specific subdirectories. Eliminate an asm() statement for the i386. Make the dynamic linker work if it is built as an executable instead of as a shared library. See i386/Makefile.inc to find out how to do it. Note, this change is not enabled and it might never be enabled. But it might be useful in the future. Building the dynamic linker as an executable should make it start up faster, because it won't have any relocations. But in practice I suspect the difference is negligible.
1999-04-09 00:28:43 +00:00
obj->pltgot = (Elf_Addr *) (obj->relocbase + dynp->d_un.d_ptr);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
break;
case DT_TEXTREL:
obj->textrel = true;
break;
case DT_SYMBOLIC:
obj->symbolic = true;
break;
case DT_RPATH:
/*
* We have to wait until later to process this, because we
* might not have gotten the address of the string table yet.
*/
Use the newly provided aux vectors to get pagesize and osreldate information. Use local version of getpagesize(), rtld_getpagesize() in private allocator. Override the __getosreldate() previously fetched from libc_pic.a with local version that uses aux value if present. Note that __getosreldate() is used by rtld indirectly, by mmap(2) libc wrapper. To be able to utilize aux, split digest_dynamic() for use by init_rtld() into two parts, where the first one does not call malloc(), and the second part uses it. init_rtld() is able to initialize global variables before digest_dynamic2() calls. In particular, pagesize and osreldate are set up from the aux values. Now, rtld avoids (two) sysctl calls in startup. Tested by: marius (sparc64) MFC after: 1 month
2010-08-17 09:05:39 +00:00
*dyn_rpath = dynp;
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
break;
case DT_SONAME:
Use the newly provided aux vectors to get pagesize and osreldate information. Use local version of getpagesize(), rtld_getpagesize() in private allocator. Override the __getosreldate() previously fetched from libc_pic.a with local version that uses aux value if present. Note that __getosreldate() is used by rtld indirectly, by mmap(2) libc wrapper. To be able to utilize aux, split digest_dynamic() for use by init_rtld() into two parts, where the first one does not call malloc(), and the second part uses it. init_rtld() is able to initialize global variables before digest_dynamic2() calls. In particular, pagesize and osreldate are set up from the aux values. Now, rtld avoids (two) sysctl calls in startup. Tested by: marius (sparc64) MFC after: 1 month
2010-08-17 09:05:39 +00:00
*dyn_soname = dynp;
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
break;
Import the DragonFly BSD commit 4f0bc915b65fcf5a23214f6d221d65c80be68ad4 by John Marino <draco@marino.st>, with the following (edited) commit message Date: Sat, 24 Mar 2012 06:40:50 +0100 Subject: [PATCH 1/1] rtld: Implement DT_RUNPATH and -z nodefaultlib DT_RUNPATH is incorrectly being considered as an alias of DT_RPATH. The purpose of DT_RUNPATH is to have two different types of rpath: one that can be overridden by the environment variable LD_LIBRARY_PATH and one that can't. With the currently implementation, LD_LIBRARY_PATH will always trump any embedded rpath or runpath tags. Current path search order by rtld: ================================== LD_LIBRARY_PATH DT_RPATH / DT_RUNPATH (always the same) ldconfig hints file (default: /var/run/ld-elf.so.hints) /usr/lib New path search order by rtld: ============================== DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file /usr/lib The new path search matches how the linux runtime loader works. The other major added feature is support for linker flag "-z nodefaultlib". When this flag is passed to the linker, rtld will skip all references to the standard library search path ("/usr/lib" in this case but it could handle more color delimited paths) except in DT_RPATH and DT_RUNPATH. New path search order by rtld with -z nodefaultlib flag set: ============================================================ DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file (skips all references to /usr/lib) FreeBSD notes: - we fixed some bugs which were submitted to DragonFly and merged there as commit 1ff8a2bd3eb6e5587174c6a983303ea3a79e0002; - we added LD_LIBRARY_PATH_RPATH environment variable to switch to the previous behaviour of considering DT_RPATH a synonym for DT_RUNPATH; - the FreeBSD default search path is /lib:/usr/lib and not /usr/lib. Reviewed by: kan MFC after: 1 month MFC note: flip the ld_library_path_rpath default value for stable/9
2012-07-15 10:53:48 +00:00
case DT_RUNPATH:
*dyn_runpath = dynp;
break;
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
case DT_INIT:
Update rtld for the "new" ia64 ABI. In the old toolchain, the DT_INIT and DT_FINI tags pointed to fptr records. In 2.11.2, it points to the actuall address of the function. On IA64 you cannot just take an address of a function, store it in a function pointer variable and call it.. the function pointers point to a fptr data block that has the target gp and address in it. This is absolutely necessary for using the in-tree binutils toolchain, but (unfortunately) will not work with old shared libraries. Save your old ld-elf.so.1 if you want to use old ones still. Do not mix-and-match. This is a no-op change for i386 and alpha. Reviewed by: dfr
2001-10-29 10:10:10 +00:00
obj->init = (Elf_Addr) (obj->relocbase + dynp->d_un.d_ptr);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
break;
Add support for preinit, init and fini arrays. Some ABIs, in particular on ARM, do require working init arrays. Traditional FreeBSD crt1 calls _init and _fini of the binary, instead of allowing runtime linker to arrange the calls. This was probably done to have the same crt code serve both statically and dynamically linked binaries. Since ABI mandates that first is called preinit array functions, then init, and then init array functions, the init have to be called from rtld now. To provide binary compatibility to old FreeBSD crt1, which calls _init itself, rtld only calls intializers and finalizers for main binary if binary has a note indicating that new crt was used for linking. Add parsing of ELF notes to rtld, and cache p_osrel value since we parsed it anyway. The patch is inspired by init_array support for DragonflyBSD, written by John Marino. Reviewed by: kan Tested by: andrew (arm, previous version), flo (sparc64, previous version) MFC after: 3 weeks
2012-03-11 20:03:09 +00:00
case DT_PREINIT_ARRAY:
obj->preinit_array = (Elf_Addr)(obj->relocbase + dynp->d_un.d_ptr);
break;
case DT_PREINIT_ARRAYSZ:
obj->preinit_array_num = dynp->d_un.d_val / sizeof(Elf_Addr);
break;
case DT_INIT_ARRAY:
obj->init_array = (Elf_Addr)(obj->relocbase + dynp->d_un.d_ptr);
break;
case DT_INIT_ARRAYSZ:
obj->init_array_num = dynp->d_un.d_val / sizeof(Elf_Addr);
break;
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
case DT_FINI:
Update rtld for the "new" ia64 ABI. In the old toolchain, the DT_INIT and DT_FINI tags pointed to fptr records. In 2.11.2, it points to the actuall address of the function. On IA64 you cannot just take an address of a function, store it in a function pointer variable and call it.. the function pointers point to a fptr data block that has the target gp and address in it. This is absolutely necessary for using the in-tree binutils toolchain, but (unfortunately) will not work with old shared libraries. Save your old ld-elf.so.1 if you want to use old ones still. Do not mix-and-match. This is a no-op change for i386 and alpha. Reviewed by: dfr
2001-10-29 10:10:10 +00:00
obj->fini = (Elf_Addr) (obj->relocbase + dynp->d_un.d_ptr);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
break;
Add support for preinit, init and fini arrays. Some ABIs, in particular on ARM, do require working init arrays. Traditional FreeBSD crt1 calls _init and _fini of the binary, instead of allowing runtime linker to arrange the calls. This was probably done to have the same crt code serve both statically and dynamically linked binaries. Since ABI mandates that first is called preinit array functions, then init, and then init array functions, the init have to be called from rtld now. To provide binary compatibility to old FreeBSD crt1, which calls _init itself, rtld only calls intializers and finalizers for main binary if binary has a note indicating that new crt was used for linking. Add parsing of ELF notes to rtld, and cache p_osrel value since we parsed it anyway. The patch is inspired by init_array support for DragonflyBSD, written by John Marino. Reviewed by: kan Tested by: andrew (arm, previous version), flo (sparc64, previous version) MFC after: 3 weeks
2012-03-11 20:03:09 +00:00
case DT_FINI_ARRAY:
obj->fini_array = (Elf_Addr)(obj->relocbase + dynp->d_un.d_ptr);
break;
case DT_FINI_ARRAYSZ:
obj->fini_array_num = dynp->d_un.d_val / sizeof(Elf_Addr);
break;
MFp4: Add mips support for dynamic linking. This code came from the merged mips2 and Juniper mips repositories. Warner Losh, Randall Seager, Oleksandr Tymoshenko and Olivier Houchard worked to merge, debug and integrate this code. This code may also contain code derived from NetBSD.
2008-04-04 20:59:26 +00:00
/*
* Don't process DT_DEBUG on MIPS as the dynamic section
* is mapped read-only. DT_MIPS_RLD_MAP is used instead.
*/
Revert r293201, r293202 (rtld: populate DT_DEBUG iff DYNAMIC segment is writable) It turns out MIPS binaries may have other oddities that can trigger a fault at startup. PR: 206017 Reported by: ray
2016-01-08 00:56:41 +00:00
#ifndef __mips__
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
case DT_DEBUG:
Minor changes to make this work on sparc64. Approved by: jdp Tested on: alpha, i386, sparc64
2002-04-02 02:19:02 +00:00
if (!early)
dbg("Filling in DT_DEBUG entry");
Add alpha support. Submitted by: John Birrell <jb@cimlogic.com.au> (with extra hacks by me) Obtained from: Probably NetBSD
1998-09-04 19:03:57 +00:00
((Elf_Dyn*)dynp)->d_un.d_ptr = (Elf_Addr) &r_debug;
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
break;
Revert r293201, r293202 (rtld: populate DT_DEBUG iff DYNAMIC segment is writable) It turns out MIPS binaries may have other oddities that can trigger a fault at startup. PR: 206017 Reported by: ray
2016-01-08 00:56:41 +00:00
#endif
Add alpha support. Submitted by: John Birrell <jb@cimlogic.com.au> (with extra hacks by me) Obtained from: Probably NetBSD
1998-09-04 19:03:57 +00:00
- Add support for DT_FLAGS. - Define various things from the most recent ELF spec.
2003-06-18 03:34:29 +00:00
case DT_FLAGS:
Change interpretation of the DF_ORIGIN and DF_1_ORIGIN flags. According to standard, the presence of the flags only means that the object path must be resolved at the time object loading, instead of my reading that the flag is required to enable token substitution at all. The consequence is that -z origin linker flag is no longer required for the token substitution in the run/rpath or the needed library soname. It is only recommended if token substition is needed at dlopen(3) time, since namecache might drop the required entries at the time of resolution. Found, reviewed and tested by: emaste Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-04-27 18:41:31 +00:00
if (dynp->d_un.d_val & DF_ORIGIN)
Implement the dynamic string token substitution in the rpath and soneeded pathes. The $ORIGIN, $OSNAME, $OSREL and $PLATFORM tokens are supported. Enabling the substitution requires DF_ORIGIN flag in DT_FLAGS or DF_1_ORIGIN if DF_FLAGS_1, that may be set with -z origin gnu ld flag. Translation is unconditionally disabled for setuid/setgid processes. The $ORIGIN translation relies on the AT_EXECPATH auxinfo supplied by kernel. Requested by: maho Tested by: maho, pho Reviewed by: kan
2009-03-18 13:40:37 +00:00
obj->z_origin = true;
- Add support for DT_FLAGS. - Define various things from the most recent ELF spec.
2003-06-18 03:34:29 +00:00
if (dynp->d_un.d_val & DF_SYMBOLIC)
obj->symbolic = true;
if (dynp->d_un.d_val & DF_TEXTREL)
obj->textrel = true;
if (dynp->d_un.d_val & DF_BIND_NOW)
obj->bind_now = true;
Fix clang warnings. Approved by: philip (mentor)
2011-06-18 13:56:33 +00:00
/*if (dynp->d_un.d_val & DF_STATIC_TLS)
;*/
- Add support for DT_FLAGS. - Define various things from the most recent ELF spec.
2003-06-18 03:34:29 +00:00
break;
MFp4: Add mips support for dynamic linking. This code came from the merged mips2 and Juniper mips repositories. Warner Losh, Randall Seager, Oleksandr Tymoshenko and Olivier Houchard worked to merge, debug and integrate this code. This code may also contain code derived from NetBSD.
2008-04-04 20:59:26 +00:00
#ifdef __mips__
case DT_MIPS_LOCAL_GOTNO:
obj->local_gotno = dynp->d_un.d_val;
Provide support for ELFv2 userland if using a newer compiler (recent clang or gcc) and binutils >= 2.24. Not enabled by default.
2015-12-03 00:10:57 +00:00
break;
MFp4: Add mips support for dynamic linking. This code came from the merged mips2 and Juniper mips repositories. Warner Losh, Randall Seager, Oleksandr Tymoshenko and Olivier Houchard worked to merge, debug and integrate this code. This code may also contain code derived from NetBSD.
2008-04-04 20:59:26 +00:00
case DT_MIPS_SYMTABNO:
obj->symtabno = dynp->d_un.d_val;
break;
case DT_MIPS_GOTSYM:
obj->gotsym = dynp->d_un.d_val;
break;
case DT_MIPS_RLD_MAP:
Populate .rld_map on MIPS for debuggers On MIPS the .dynamic section is read-only, so the pointer to rtld information for debuggers cannot be stored there (in DT_DEBUG). Instead, a special section .rld_map is used. Sponsored by: DARPA, AFRL Approved by: re (delphij)
2013-10-02 02:32:58 +00:00
*((Elf_Addr *)(dynp->d_un.d_ptr)) = (Elf_Addr) &r_debug;
MFp4: Add mips support for dynamic linking. This code came from the merged mips2 and Juniper mips repositories. Warner Losh, Randall Seager, Oleksandr Tymoshenko and Olivier Houchard worked to merge, debug and integrate this code. This code may also contain code derived from NetBSD.
2008-04-04 20:59:26 +00:00
break;
#endif
- Add support for DT_FLAGS. - Define various things from the most recent ELF spec.
2003-06-18 03:34:29 +00:00
Provide support for ELFv2 userland if using a newer compiler (recent clang or gcc) and binutils >= 2.24. Not enabled by default.
2015-12-03 00:10:57 +00:00
#ifdef __powerpc64__
case DT_PPC64_GLINK:
obj->glink = (Elf_Addr) (obj->relocbase + dynp->d_un.d_ptr);
break;
#endif
Implement the dynamic string token substitution in the rpath and soneeded pathes. The $ORIGIN, $OSNAME, $OSREL and $PLATFORM tokens are supported. Enabling the substitution requires DF_ORIGIN flag in DT_FLAGS or DF_1_ORIGIN if DF_FLAGS_1, that may be set with -z origin gnu ld flag. Translation is unconditionally disabled for setuid/setgid processes. The $ORIGIN translation relies on the AT_EXECPATH auxinfo supplied by kernel. Requested by: maho Tested by: maho, pho Reviewed by: kan
2009-03-18 13:40:37 +00:00
case DT_FLAGS_1:
Implement rtld part of the support for -z nodlopen (see ld(1)). Reviewed by: kan MFC after: 3 weeks
2009-11-26 13:57:20 +00:00
if (dynp->d_un.d_val & DF_1_NOOPEN)
obj->z_noopen = true;
Change interpretation of the DF_ORIGIN and DF_1_ORIGIN flags. According to standard, the presence of the flags only means that the object path must be resolved at the time object loading, instead of my reading that the flag is required to enable token substitution at all. The consequence is that -z origin linker flag is no longer required for the token substitution in the run/rpath or the needed library soname. It is only recommended if token substition is needed at dlopen(3) time, since namecache might drop the required entries at the time of resolution. Found, reviewed and tested by: emaste Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-04-27 18:41:31 +00:00
if (dynp->d_un.d_val & DF_1_ORIGIN)
Implement the dynamic string token substitution in the rpath and soneeded pathes. The $ORIGIN, $OSNAME, $OSREL and $PLATFORM tokens are supported. Enabling the substitution requires DF_ORIGIN flag in DT_FLAGS or DF_1_ORIGIN if DF_FLAGS_1, that may be set with -z origin gnu ld flag. Translation is unconditionally disabled for setuid/setgid processes. The $ORIGIN translation relies on the AT_EXECPATH auxinfo supplied by kernel. Requested by: maho Tested by: maho, pho Reviewed by: kan
2009-03-18 13:40:37 +00:00
obj->z_origin = true;
Implement support -z global linker option. It marks the shared object as always participating in the global symbols namespace, regardless of the way the object was brought into the process address space. Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-04-15 08:16:34 +00:00
if (dynp->d_un.d_val & DF_1_GLOBAL)
obj->z_global = true;
Implement the dynamic string token substitution in the rpath and soneeded pathes. The $ORIGIN, $OSNAME, $OSREL and $PLATFORM tokens are supported. Enabling the substitution requires DF_ORIGIN flag in DT_FLAGS or DF_1_ORIGIN if DF_FLAGS_1, that may be set with -z origin gnu ld flag. Translation is unconditionally disabled for setuid/setgid processes. The $ORIGIN translation relies on the AT_EXECPATH auxinfo supplied by kernel. Requested by: maho Tested by: maho, pho Reviewed by: kan
2009-03-18 13:40:37 +00:00
if (dynp->d_un.d_val & DF_1_BIND_NOW)
obj->bind_now = true;
Implement support for RTLD_NODELETE flag for dlopen() and -z nodelete static linker option. Do it by incrementing reference count on the loaded object and its dependencies. Reviewed by: davidxu, kan
2009-03-30 08:47:28 +00:00
if (dynp->d_un.d_val & DF_1_NODELETE)
obj->z_nodelete = true;
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
if (dynp->d_un.d_val & DF_1_LOADFLTR)
obj->z_loadfltr = true;
Implement support for the interpose dso flag. Requested by: bf Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 1 week Approved by: re (glebius)
2013-10-07 08:19:30 +00:00
if (dynp->d_un.d_val & DF_1_INTERPOSE)
obj->z_interpose = true;
Import the DragonFly BSD commit 4f0bc915b65fcf5a23214f6d221d65c80be68ad4 by John Marino <draco@marino.st>, with the following (edited) commit message Date: Sat, 24 Mar 2012 06:40:50 +0100 Subject: [PATCH 1/1] rtld: Implement DT_RUNPATH and -z nodefaultlib DT_RUNPATH is incorrectly being considered as an alias of DT_RPATH. The purpose of DT_RUNPATH is to have two different types of rpath: one that can be overridden by the environment variable LD_LIBRARY_PATH and one that can't. With the currently implementation, LD_LIBRARY_PATH will always trump any embedded rpath or runpath tags. Current path search order by rtld: ================================== LD_LIBRARY_PATH DT_RPATH / DT_RUNPATH (always the same) ldconfig hints file (default: /var/run/ld-elf.so.hints) /usr/lib New path search order by rtld: ============================== DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file /usr/lib The new path search matches how the linux runtime loader works. The other major added feature is support for linker flag "-z nodefaultlib". When this flag is passed to the linker, rtld will skip all references to the standard library search path ("/usr/lib" in this case but it could handle more color delimited paths) except in DT_RPATH and DT_RUNPATH. New path search order by rtld with -z nodefaultlib flag set: ============================================================ DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file (skips all references to /usr/lib) FreeBSD notes: - we fixed some bugs which were submitted to DragonFly and merged there as commit 1ff8a2bd3eb6e5587174c6a983303ea3a79e0002; - we added LD_LIBRARY_PATH_RPATH environment variable to switch to the previous behaviour of considering DT_RPATH a synonym for DT_RUNPATH; - the FreeBSD default search path is /lib:/usr/lib and not /usr/lib. Reviewed by: kan MFC after: 1 month MFC note: flip the ld_library_path_rpath default value for stable/9
2012-07-15 10:53:48 +00:00
if (dynp->d_un.d_val & DF_1_NODEFLIB)
obj->z_nodeflib = true;
Implement the dynamic string token substitution in the rpath and soneeded pathes. The $ORIGIN, $OSNAME, $OSREL and $PLATFORM tokens are supported. Enabling the substitution requires DF_ORIGIN flag in DT_FLAGS or DF_1_ORIGIN if DF_FLAGS_1, that may be set with -z origin gnu ld flag. Translation is unconditionally disabled for setuid/setgid processes. The $ORIGIN translation relies on the AT_EXECPATH auxinfo supplied by kernel. Requested by: maho Tested by: maho, pho Reviewed by: kan
2009-03-18 13:40:37 +00:00
break;
Add alpha support. Submitted by: John Birrell <jb@cimlogic.com.au> (with extra hacks by me) Obtained from: Probably NetBSD
1998-09-04 19:03:57 +00:00
default:
Minor changes to make this work on sparc64. Approved by: jdp Tested on: alpha, i386, sparc64
2002-04-02 02:19:02 +00:00
if (!early) {
dbg("Ignoring d_tag %ld = %#lx", (long)dynp->d_tag,
(long)dynp->d_tag);
}
Change the warning about unrecognized entries in the dynamic table to a debug message which is disabled in production builds of the dynamic linker. The condition warned about is normally harmless. PR: bin/12849
1999-09-04 20:14:48 +00:00
break;
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
}
}
Add alpha support. Submitted by: John Birrell <jb@cimlogic.com.au> (with extra hacks by me) Obtained from: Probably NetBSD
1998-09-04 19:03:57 +00:00
obj->traced = false;
if (plttype == DT_RELA) {
obj->pltrela = (const Elf_Rela *) obj->pltrel;
obj->pltrel = NULL;
obj->pltrelasize = obj->pltrelsize;
obj->pltrelsize = 0;
}
Add GNU hash support for rtld. Based on dragonflybsd support for GNU hash by John Marino <draco marino st> Reviewed by: kan Tested by: bapt MFC after: 2 weeks
2012-04-30 13:31:10 +00:00
/* Determine size of dynsym table (equal to nchains of sysv hash) */
if (obj->valid_hash_sysv)
obj->dynsymcount = obj->nchains;
else if (obj->valid_hash_gnu) {
obj->dynsymcount = 0;
for (bkt = 0; bkt < obj->nbuckets_gnu; bkt++) {
if (obj->buckets_gnu[bkt] == 0)
continue;
hashval = &obj->chain_zero_gnu[obj->buckets_gnu[bkt]];
do
obj->dynsymcount++;
while ((*hashval++ & 1u) == 0);
}
obj->dynsymcount += obj->symndx_gnu;
}
Use the newly provided aux vectors to get pagesize and osreldate information. Use local version of getpagesize(), rtld_getpagesize() in private allocator. Override the __getosreldate() previously fetched from libc_pic.a with local version that uses aux value if present. Note that __getosreldate() is used by rtld indirectly, by mmap(2) libc wrapper. To be able to utilize aux, split digest_dynamic() for use by init_rtld() into two parts, where the first one does not call malloc(), and the second part uses it. init_rtld() is able to initialize global variables before digest_dynamic2() calls. In particular, pagesize and osreldate are set up from the aux values. Now, rtld avoids (two) sysctl calls in startup. Tested by: marius (sparc64) MFC after: 1 month
2010-08-17 09:05:39 +00:00
}
Change interpretation of the DF_ORIGIN and DF_1_ORIGIN flags. According to standard, the presence of the flags only means that the object path must be resolved at the time object loading, instead of my reading that the flag is required to enable token substitution at all. The consequence is that -z origin linker flag is no longer required for the token substitution in the run/rpath or the needed library soname. It is only recommended if token substition is needed at dlopen(3) time, since namecache might drop the required entries at the time of resolution. Found, reviewed and tested by: emaste Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-04-27 18:41:31 +00:00
static bool
obj_resolve_origin(Obj_Entry *obj)
{
if (obj->origin_path != NULL)
return (true);
obj->origin_path = xmalloc(PATH_MAX);
return (rtld_dirname_abs(obj->path, obj->origin_path) != -1);
}
Use the newly provided aux vectors to get pagesize and osreldate information. Use local version of getpagesize(), rtld_getpagesize() in private allocator. Override the __getosreldate() previously fetched from libc_pic.a with local version that uses aux value if present. Note that __getosreldate() is used by rtld indirectly, by mmap(2) libc wrapper. To be able to utilize aux, split digest_dynamic() for use by init_rtld() into two parts, where the first one does not call malloc(), and the second part uses it. init_rtld() is able to initialize global variables before digest_dynamic2() calls. In particular, pagesize and osreldate are set up from the aux values. Now, rtld avoids (two) sysctl calls in startup. Tested by: marius (sparc64) MFC after: 1 month
2010-08-17 09:05:39 +00:00
static void
digest_dynamic2(Obj_Entry *obj, const Elf_Dyn *dyn_rpath,
Import the DragonFly BSD commit 4f0bc915b65fcf5a23214f6d221d65c80be68ad4 by John Marino <draco@marino.st>, with the following (edited) commit message Date: Sat, 24 Mar 2012 06:40:50 +0100 Subject: [PATCH 1/1] rtld: Implement DT_RUNPATH and -z nodefaultlib DT_RUNPATH is incorrectly being considered as an alias of DT_RPATH. The purpose of DT_RUNPATH is to have two different types of rpath: one that can be overridden by the environment variable LD_LIBRARY_PATH and one that can't. With the currently implementation, LD_LIBRARY_PATH will always trump any embedded rpath or runpath tags. Current path search order by rtld: ================================== LD_LIBRARY_PATH DT_RPATH / DT_RUNPATH (always the same) ldconfig hints file (default: /var/run/ld-elf.so.hints) /usr/lib New path search order by rtld: ============================== DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file /usr/lib The new path search matches how the linux runtime loader works. The other major added feature is support for linker flag "-z nodefaultlib". When this flag is passed to the linker, rtld will skip all references to the standard library search path ("/usr/lib" in this case but it could handle more color delimited paths) except in DT_RPATH and DT_RUNPATH. New path search order by rtld with -z nodefaultlib flag set: ============================================================ DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file (skips all references to /usr/lib) FreeBSD notes: - we fixed some bugs which were submitted to DragonFly and merged there as commit 1ff8a2bd3eb6e5587174c6a983303ea3a79e0002; - we added LD_LIBRARY_PATH_RPATH environment variable to switch to the previous behaviour of considering DT_RPATH a synonym for DT_RUNPATH; - the FreeBSD default search path is /lib:/usr/lib and not /usr/lib. Reviewed by: kan MFC after: 1 month MFC note: flip the ld_library_path_rpath default value for stable/9
2012-07-15 10:53:48 +00:00
const Elf_Dyn *dyn_soname, const Elf_Dyn *dyn_runpath)
Use the newly provided aux vectors to get pagesize and osreldate information. Use local version of getpagesize(), rtld_getpagesize() in private allocator. Override the __getosreldate() previously fetched from libc_pic.a with local version that uses aux value if present. Note that __getosreldate() is used by rtld indirectly, by mmap(2) libc wrapper. To be able to utilize aux, split digest_dynamic() for use by init_rtld() into two parts, where the first one does not call malloc(), and the second part uses it. init_rtld() is able to initialize global variables before digest_dynamic2() calls. In particular, pagesize and osreldate are set up from the aux values. Now, rtld avoids (two) sysctl calls in startup. Tested by: marius (sparc64) MFC after: 1 month
2010-08-17 09:05:39 +00:00
{
Add alpha support. Submitted by: John Birrell <jb@cimlogic.com.au> (with extra hacks by me) Obtained from: Probably NetBSD
1998-09-04 19:03:57 +00:00
Change interpretation of the DF_ORIGIN and DF_1_ORIGIN flags. According to standard, the presence of the flags only means that the object path must be resolved at the time object loading, instead of my reading that the flag is required to enable token substitution at all. The consequence is that -z origin linker flag is no longer required for the token substitution in the run/rpath or the needed library soname. It is only recommended if token substition is needed at dlopen(3) time, since namecache might drop the required entries at the time of resolution. Found, reviewed and tested by: emaste Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-04-27 18:41:31 +00:00
if (obj->z_origin && !obj_resolve_origin(obj))
rtld_die();
Implement ELF symbol versioning using GNU semantics. This code aims to be compatible with symbol versioning support as implemented by GNU libc and documented by http://people.redhat.com/~drepper/symbol-versioning and LSB 3.0. Implement dlvsym() function to allow lookups for a specific version of a given symbol.
2005-12-18 19:43:33 +00:00
Change interpretation of the DF_ORIGIN and DF_1_ORIGIN flags. According to standard, the presence of the flags only means that the object path must be resolved at the time object loading, instead of my reading that the flag is required to enable token substitution at all. The consequence is that -z origin linker flag is no longer required for the token substitution in the run/rpath or the needed library soname. It is only recommended if token substition is needed at dlopen(3) time, since namecache might drop the required entries at the time of resolution. Found, reviewed and tested by: emaste Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-04-27 18:41:31 +00:00
if (dyn_runpath != NULL) {
obj->runpath = (char *)obj->strtab + dyn_runpath->d_un.d_val;
obj->runpath = origin_subst(obj, obj->runpath);
} else if (dyn_rpath != NULL) {
obj->rpath = (char *)obj->strtab + dyn_rpath->d_un.d_val;
obj->rpath = origin_subst(obj, obj->rpath);
}
if (dyn_soname != NULL)
object_add_name(obj, obj->strtab + dyn_soname->d_un.d_val);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
}
Use the newly provided aux vectors to get pagesize and osreldate information. Use local version of getpagesize(), rtld_getpagesize() in private allocator. Override the __getosreldate() previously fetched from libc_pic.a with local version that uses aux value if present. Note that __getosreldate() is used by rtld indirectly, by mmap(2) libc wrapper. To be able to utilize aux, split digest_dynamic() for use by init_rtld() into two parts, where the first one does not call malloc(), and the second part uses it. init_rtld() is able to initialize global variables before digest_dynamic2() calls. In particular, pagesize and osreldate are set up from the aux values. Now, rtld avoids (two) sysctl calls in startup. Tested by: marius (sparc64) MFC after: 1 month
2010-08-17 09:05:39 +00:00
static void
digest_dynamic(Obj_Entry *obj, int early)
{
const Elf_Dyn *dyn_rpath;
const Elf_Dyn *dyn_soname;
Import the DragonFly BSD commit 4f0bc915b65fcf5a23214f6d221d65c80be68ad4 by John Marino <draco@marino.st>, with the following (edited) commit message Date: Sat, 24 Mar 2012 06:40:50 +0100 Subject: [PATCH 1/1] rtld: Implement DT_RUNPATH and -z nodefaultlib DT_RUNPATH is incorrectly being considered as an alias of DT_RPATH. The purpose of DT_RUNPATH is to have two different types of rpath: one that can be overridden by the environment variable LD_LIBRARY_PATH and one that can't. With the currently implementation, LD_LIBRARY_PATH will always trump any embedded rpath or runpath tags. Current path search order by rtld: ================================== LD_LIBRARY_PATH DT_RPATH / DT_RUNPATH (always the same) ldconfig hints file (default: /var/run/ld-elf.so.hints) /usr/lib New path search order by rtld: ============================== DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file /usr/lib The new path search matches how the linux runtime loader works. The other major added feature is support for linker flag "-z nodefaultlib". When this flag is passed to the linker, rtld will skip all references to the standard library search path ("/usr/lib" in this case but it could handle more color delimited paths) except in DT_RPATH and DT_RUNPATH. New path search order by rtld with -z nodefaultlib flag set: ============================================================ DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file (skips all references to /usr/lib) FreeBSD notes: - we fixed some bugs which were submitted to DragonFly and merged there as commit 1ff8a2bd3eb6e5587174c6a983303ea3a79e0002; - we added LD_LIBRARY_PATH_RPATH environment variable to switch to the previous behaviour of considering DT_RPATH a synonym for DT_RUNPATH; - the FreeBSD default search path is /lib:/usr/lib and not /usr/lib. Reviewed by: kan MFC after: 1 month MFC note: flip the ld_library_path_rpath default value for stable/9
2012-07-15 10:53:48 +00:00
const Elf_Dyn *dyn_runpath;
Use the newly provided aux vectors to get pagesize and osreldate information. Use local version of getpagesize(), rtld_getpagesize() in private allocator. Override the __getosreldate() previously fetched from libc_pic.a with local version that uses aux value if present. Note that __getosreldate() is used by rtld indirectly, by mmap(2) libc wrapper. To be able to utilize aux, split digest_dynamic() for use by init_rtld() into two parts, where the first one does not call malloc(), and the second part uses it. init_rtld() is able to initialize global variables before digest_dynamic2() calls. In particular, pagesize and osreldate are set up from the aux values. Now, rtld avoids (two) sysctl calls in startup. Tested by: marius (sparc64) MFC after: 1 month
2010-08-17 09:05:39 +00:00
Import the DragonFly BSD commit 4f0bc915b65fcf5a23214f6d221d65c80be68ad4 by John Marino <draco@marino.st>, with the following (edited) commit message Date: Sat, 24 Mar 2012 06:40:50 +0100 Subject: [PATCH 1/1] rtld: Implement DT_RUNPATH and -z nodefaultlib DT_RUNPATH is incorrectly being considered as an alias of DT_RPATH. The purpose of DT_RUNPATH is to have two different types of rpath: one that can be overridden by the environment variable LD_LIBRARY_PATH and one that can't. With the currently implementation, LD_LIBRARY_PATH will always trump any embedded rpath or runpath tags. Current path search order by rtld: ================================== LD_LIBRARY_PATH DT_RPATH / DT_RUNPATH (always the same) ldconfig hints file (default: /var/run/ld-elf.so.hints) /usr/lib New path search order by rtld: ============================== DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file /usr/lib The new path search matches how the linux runtime loader works. The other major added feature is support for linker flag "-z nodefaultlib". When this flag is passed to the linker, rtld will skip all references to the standard library search path ("/usr/lib" in this case but it could handle more color delimited paths) except in DT_RPATH and DT_RUNPATH. New path search order by rtld with -z nodefaultlib flag set: ============================================================ DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file (skips all references to /usr/lib) FreeBSD notes: - we fixed some bugs which were submitted to DragonFly and merged there as commit 1ff8a2bd3eb6e5587174c6a983303ea3a79e0002; - we added LD_LIBRARY_PATH_RPATH environment variable to switch to the previous behaviour of considering DT_RPATH a synonym for DT_RUNPATH; - the FreeBSD default search path is /lib:/usr/lib and not /usr/lib. Reviewed by: kan MFC after: 1 month MFC note: flip the ld_library_path_rpath default value for stable/9
2012-07-15 10:53:48 +00:00
digest_dynamic1(obj, early, &dyn_rpath, &dyn_soname, &dyn_runpath);
digest_dynamic2(obj, dyn_rpath, dyn_soname, dyn_runpath);
Use the newly provided aux vectors to get pagesize and osreldate information. Use local version of getpagesize(), rtld_getpagesize() in private allocator. Override the __getosreldate() previously fetched from libc_pic.a with local version that uses aux value if present. Note that __getosreldate() is used by rtld indirectly, by mmap(2) libc wrapper. To be able to utilize aux, split digest_dynamic() for use by init_rtld() into two parts, where the first one does not call malloc(), and the second part uses it. init_rtld() is able to initialize global variables before digest_dynamic2() calls. In particular, pagesize and osreldate are set up from the aux values. Now, rtld avoids (two) sysctl calls in startup. Tested by: marius (sparc64) MFC after: 1 month
2010-08-17 09:05:39 +00:00
}
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
/*
* Process a shared object's program header. This is used only for the
* main program, when the kernel has already loaded the main program
* into memory before calling the dynamic linker. It creates and
* returns an Obj_Entry structure.
*/
static Obj_Entry *
Change many asserts into normal errors. They were all for conditions caused by invalid shared objects rather than by internal errors. Enable format string mismatch checking for _rtld_error().
1999-07-18 00:02:19 +00:00
digest_phdr(const Elf_Phdr *phdr, int phnum, caddr_t entry, const char *path)
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
{
Revamp the symbol lookup algorithm to cope better with objects loaded separately by dlopen that have global symbols with identical names. Viewing each dlopened object as a DAG which is linked by its DT_NEEDED entries in the dynamic table, the search order is as follows: * If the referencing object was linked with -Bsymbolic, search it internally. * Search all dlopened DAGs containing the referencing object. * Search all objects loaded at program start up. * Search all objects which were dlopened() using the RTLD_GLOBAL flag (which is now supported too). The search terminates as soon as a strong definition is found. Lacking that, the first weak definition is used. These rules match those of Solaris, as best I could determine them from its vague manual pages and the results of experiments I performed. PR: misc/12438
1999-08-30 01:48:19 +00:00
Obj_Entry *obj;
Add alpha support. Submitted by: John Birrell <jb@cimlogic.com.au> (with extra hacks by me) Obtained from: Probably NetBSD
1998-09-04 19:03:57 +00:00
const Elf_Phdr *phlimit = phdr + phnum;
const Elf_Phdr *ph;
Add support for preinit, init and fini arrays. Some ABIs, in particular on ARM, do require working init arrays. Traditional FreeBSD crt1 calls _init and _fini of the binary, instead of allowing runtime linker to arrange the calls. This was probably done to have the same crt code serve both statically and dynamically linked binaries. Since ABI mandates that first is called preinit array functions, then init, and then init array functions, the init have to be called from rtld now. To provide binary compatibility to old FreeBSD crt1, which calls _init itself, rtld only calls intializers and finalizers for main binary if binary has a note indicating that new crt was used for linking. Add parsing of ELF notes to rtld, and cache p_osrel value since we parsed it anyway. The patch is inspired by init_array support for DragonflyBSD, written by John Marino. Reviewed by: kan Tested by: andrew (arm, previous version), flo (sparc64, previous version) MFC after: 3 weeks
2012-03-11 20:03:09 +00:00
Elf_Addr note_start, note_end;
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
int nsegs = 0;
Revamp the symbol lookup algorithm to cope better with objects loaded separately by dlopen that have global symbols with identical names. Viewing each dlopened object as a DAG which is linked by its DT_NEEDED entries in the dynamic table, the search order is as follows: * If the referencing object was linked with -Bsymbolic, search it internally. * Search all dlopened DAGs containing the referencing object. * Search all objects loaded at program start up. * Search all objects which were dlopened() using the RTLD_GLOBAL flag (which is now supported too). The search terminates as soon as a strong definition is found. Lacking that, the first weak definition is used. These rules match those of Solaris, as best I could determine them from its vague manual pages and the results of experiments I performed. PR: misc/12438
1999-08-30 01:48:19 +00:00
obj = obj_new();
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
for (ph = phdr; ph < phlimit; ph++) {
Calculate relocation base for the main object, and apply the relocation adjustment for all virtual addresses encoded into the ELF structures of it. PIE binary could and should be loaded at non-zero mapbase. For sym_zero pseudosymbol used as a return value from find_symdef() for undefined weak symbols, st_value also should be adjusted, since _rtld_bind corrects symbol values by relocbase. Discussed with: bz Reviewed by: kan Tested by: bz (i386, amd64), bsam (linux) MFC after: some time
2009-10-10 15:27:10 +00:00
if (ph->p_type != PT_PHDR)
continue;
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
Calculate relocation base for the main object, and apply the relocation adjustment for all virtual addresses encoded into the ELF structures of it. PIE binary could and should be loaded at non-zero mapbase. For sym_zero pseudosymbol used as a return value from find_symdef() for undefined weak symbols, st_value also should be adjusted, since _rtld_bind corrects symbol values by relocbase. Discussed with: bz Reviewed by: kan Tested by: bz (i386, amd64), bsam (linux) MFC after: some time
2009-10-10 15:27:10 +00:00
obj->phdr = phdr;
obj->phsize = ph->p_memsz;
obj->relocbase = (caddr_t)phdr - ph->p_vaddr;
break;
}
In rtld, read the initial stack access mode from AT_STACKPROT as set by kernel, and parse PT_GNU_STACK phdr from linked and loaded dsos. If the loaded dso requires executable stack, as specified by PF_X bit of p_flags of PT_GNU_STACK phdr, but current stack protection does not permit execution, the __pthread_map_stacks_exec symbol is looked up and called. It should be implemented in libc or threading library and change the protection mode of all thread stacks to be executable. Provide a private interface _rtld_get_stack_prot() to export the stack access mode as calculated by rtld. Reviewed by: kan
2011-01-08 17:11:49 +00:00
obj->stack_flags = PF_X | PF_R | PF_W;
Calculate relocation base for the main object, and apply the relocation adjustment for all virtual addresses encoded into the ELF structures of it. PIE binary could and should be loaded at non-zero mapbase. For sym_zero pseudosymbol used as a return value from find_symdef() for undefined weak symbols, st_value also should be adjusted, since _rtld_bind corrects symbol values by relocbase. Discussed with: bz Reviewed by: kan Tested by: bz (i386, amd64), bsam (linux) MFC after: some time
2009-10-10 15:27:10 +00:00
for (ph = phdr; ph < phlimit; ph++) {
switch (ph->p_type) {
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
Get the actual pathname of the dynamic linker from the executable's PT_INTERP program header entry, to ensure that gdb always finds the right dynamic linker. Use obj->relocbase to simplify a few calculations where appropriate.
1999-08-30 01:54:13 +00:00
case PT_INTERP:
Calculate relocation base for the main object, and apply the relocation adjustment for all virtual addresses encoded into the ELF structures of it. PIE binary could and should be loaded at non-zero mapbase. For sym_zero pseudosymbol used as a return value from find_symdef() for undefined weak symbols, st_value also should be adjusted, since _rtld_bind corrects symbol values by relocbase. Discussed with: bz Reviewed by: kan Tested by: bz (i386, amd64), bsam (linux) MFC after: some time
2009-10-10 15:27:10 +00:00
obj->interp = (const char *)(ph->p_vaddr + obj->relocbase);
Get the actual pathname of the dynamic linker from the executable's PT_INTERP program header entry, to ensure that gdb always finds the right dynamic linker. Use obj->relocbase to simplify a few calculations where appropriate.
1999-08-30 01:54:13 +00:00
break;
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
case PT_LOAD:
if (nsegs == 0) { /* First load segment */
obj->vaddrbase = trunc_page(ph->p_vaddr);
Calculate relocation base for the main object, and apply the relocation adjustment for all virtual addresses encoded into the ELF structures of it. PIE binary could and should be loaded at non-zero mapbase. For sym_zero pseudosymbol used as a return value from find_symdef() for undefined weak symbols, st_value also should be adjusted, since _rtld_bind corrects symbol values by relocbase. Discussed with: bz Reviewed by: kan Tested by: bz (i386, amd64), bsam (linux) MFC after: some time
2009-10-10 15:27:10 +00:00
obj->mapbase = obj->vaddrbase + obj->relocbase;
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
obj->textsize = round_page(ph->p_vaddr + ph->p_memsz) -
obj->vaddrbase;
} else { /* Last load segment */
obj->mapsize = round_page(ph->p_vaddr + ph->p_memsz) -
obj->vaddrbase;
}
nsegs++;
break;
case PT_DYNAMIC:
Calculate relocation base for the main object, and apply the relocation adjustment for all virtual addresses encoded into the ELF structures of it. PIE binary could and should be loaded at non-zero mapbase. For sym_zero pseudosymbol used as a return value from find_symdef() for undefined weak symbols, st_value also should be adjusted, since _rtld_bind corrects symbol values by relocbase. Discussed with: bz Reviewed by: kan Tested by: bz (i386, amd64), bsam (linux) MFC after: some time
2009-10-10 15:27:10 +00:00
obj->dynamic = (const Elf_Dyn *)(ph->p_vaddr + obj->relocbase);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
break;
Add support for Thread Local Storage.
2004-08-03 08:51:00 +00:00
case PT_TLS:
obj->tlsindex = 1;
obj->tlssize = ph->p_memsz;
obj->tlsalign = ph->p_align;
obj->tlsinitsize = ph->p_filesz;
Calculate relocation base for the main object, and apply the relocation adjustment for all virtual addresses encoded into the ELF structures of it. PIE binary could and should be loaded at non-zero mapbase. For sym_zero pseudosymbol used as a return value from find_symdef() for undefined weak symbols, st_value also should be adjusted, since _rtld_bind corrects symbol values by relocbase. Discussed with: bz Reviewed by: kan Tested by: bz (i386, amd64), bsam (linux) MFC after: some time
2009-10-10 15:27:10 +00:00
obj->tlsinit = (void*)(ph->p_vaddr + obj->relocbase);
Add support for Thread Local Storage.
2004-08-03 08:51:00 +00:00
break;
In rtld, read the initial stack access mode from AT_STACKPROT as set by kernel, and parse PT_GNU_STACK phdr from linked and loaded dsos. If the loaded dso requires executable stack, as specified by PF_X bit of p_flags of PT_GNU_STACK phdr, but current stack protection does not permit execution, the __pthread_map_stacks_exec symbol is looked up and called. It should be implemented in libc or threading library and change the protection mode of all thread stacks to be executable. Provide a private interface _rtld_get_stack_prot() to export the stack access mode as calculated by rtld. Reviewed by: kan
2011-01-08 17:11:49 +00:00
case PT_GNU_STACK:
obj->stack_flags = ph->p_flags;
break;
Add support for GNU RELRO. Submitted by: John Marino <draco marino st> MFC after: 2 weeks
2012-01-30 19:52:17 +00:00
case PT_GNU_RELRO:
obj->relro_page = obj->relocbase + trunc_page(ph->p_vaddr);
obj->relro_size = round_page(ph->p_memsz);
break;
Add support for preinit, init and fini arrays. Some ABIs, in particular on ARM, do require working init arrays. Traditional FreeBSD crt1 calls _init and _fini of the binary, instead of allowing runtime linker to arrange the calls. This was probably done to have the same crt code serve both statically and dynamically linked binaries. Since ABI mandates that first is called preinit array functions, then init, and then init array functions, the init have to be called from rtld now. To provide binary compatibility to old FreeBSD crt1, which calls _init itself, rtld only calls intializers and finalizers for main binary if binary has a note indicating that new crt was used for linking. Add parsing of ELF notes to rtld, and cache p_osrel value since we parsed it anyway. The patch is inspired by init_array support for DragonflyBSD, written by John Marino. Reviewed by: kan Tested by: andrew (arm, previous version), flo (sparc64, previous version) MFC after: 3 weeks
2012-03-11 20:03:09 +00:00
case PT_NOTE:
note_start = (Elf_Addr)obj->relocbase + ph->p_vaddr;
note_end = note_start + ph->p_filesz;
digest_notes(obj, note_start, note_end);
break;
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
}
}
Put back a test for binaries with no PT_LOAD entries I over-jealosly removed in r1.69. Apploved by: re (rwatson)
2002-11-29 16:41:31 +00:00
if (nsegs < 1) {
_rtld_error("%s: too few PT_LOAD segments", path);
return NULL;
}
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
obj->entry = entry;
return obj;
}
Add support for preinit, init and fini arrays. Some ABIs, in particular on ARM, do require working init arrays. Traditional FreeBSD crt1 calls _init and _fini of the binary, instead of allowing runtime linker to arrange the calls. This was probably done to have the same crt code serve both statically and dynamically linked binaries. Since ABI mandates that first is called preinit array functions, then init, and then init array functions, the init have to be called from rtld now. To provide binary compatibility to old FreeBSD crt1, which calls _init itself, rtld only calls intializers and finalizers for main binary if binary has a note indicating that new crt was used for linking. Add parsing of ELF notes to rtld, and cache p_osrel value since we parsed it anyway. The patch is inspired by init_array support for DragonflyBSD, written by John Marino. Reviewed by: kan Tested by: andrew (arm, previous version), flo (sparc64, previous version) MFC after: 3 weeks
2012-03-11 20:03:09 +00:00
void
digest_notes(Obj_Entry *obj, Elf_Addr note_start, Elf_Addr note_end)
{
const Elf_Note *note;
const char *note_name;
uintptr_t p;
for (note = (const Elf_Note *)note_start; (Elf_Addr)note < note_end;
note = (const Elf_Note *)((const char *)(note + 1) +
roundup2(note->n_namesz, sizeof(Elf32_Addr)) +
roundup2(note->n_descsz, sizeof(Elf32_Addr)))) {
if (note->n_namesz != sizeof(NOTE_FREEBSD_VENDOR) ||
note->n_descsz != sizeof(int32_t))
continue;
rtld: Use common NT_FREEBSD_* note types introduced in r291909 Sponsored by: The FreeBSD Foundation
2015-12-21 22:40:29 +00:00
if (note->n_type != NT_FREEBSD_ABI_TAG &&
note->n_type != NT_FREEBSD_NOINIT_TAG)
Add support for preinit, init and fini arrays. Some ABIs, in particular on ARM, do require working init arrays. Traditional FreeBSD crt1 calls _init and _fini of the binary, instead of allowing runtime linker to arrange the calls. This was probably done to have the same crt code serve both statically and dynamically linked binaries. Since ABI mandates that first is called preinit array functions, then init, and then init array functions, the init have to be called from rtld now. To provide binary compatibility to old FreeBSD crt1, which calls _init itself, rtld only calls intializers and finalizers for main binary if binary has a note indicating that new crt was used for linking. Add parsing of ELF notes to rtld, and cache p_osrel value since we parsed it anyway. The patch is inspired by init_array support for DragonflyBSD, written by John Marino. Reviewed by: kan Tested by: andrew (arm, previous version), flo (sparc64, previous version) MFC after: 3 weeks
2012-03-11 20:03:09 +00:00
continue;
note_name = (const char *)(note + 1);
if (strncmp(NOTE_FREEBSD_VENDOR, note_name,
sizeof(NOTE_FREEBSD_VENDOR)) != 0)
continue;
switch (note->n_type) {
rtld: Use common NT_FREEBSD_* note types introduced in r291909 Sponsored by: The FreeBSD Foundation
2015-12-21 22:40:29 +00:00
case NT_FREEBSD_ABI_TAG:
Add support for preinit, init and fini arrays. Some ABIs, in particular on ARM, do require working init arrays. Traditional FreeBSD crt1 calls _init and _fini of the binary, instead of allowing runtime linker to arrange the calls. This was probably done to have the same crt code serve both statically and dynamically linked binaries. Since ABI mandates that first is called preinit array functions, then init, and then init array functions, the init have to be called from rtld now. To provide binary compatibility to old FreeBSD crt1, which calls _init itself, rtld only calls intializers and finalizers for main binary if binary has a note indicating that new crt was used for linking. Add parsing of ELF notes to rtld, and cache p_osrel value since we parsed it anyway. The patch is inspired by init_array support for DragonflyBSD, written by John Marino. Reviewed by: kan Tested by: andrew (arm, previous version), flo (sparc64, previous version) MFC after: 3 weeks
2012-03-11 20:03:09 +00:00
/* FreeBSD osrel note */
p = (uintptr_t)(note + 1);
p += roundup2(note->n_namesz, sizeof(Elf32_Addr));
obj->osrel = *(const int32_t *)(p);
dbg("note osrel %d", obj->osrel);
break;
rtld: Use common NT_FREEBSD_* note types introduced in r291909 Sponsored by: The FreeBSD Foundation
2015-12-21 22:40:29 +00:00
case NT_FREEBSD_NOINIT_TAG:
Add support for preinit, init and fini arrays. Some ABIs, in particular on ARM, do require working init arrays. Traditional FreeBSD crt1 calls _init and _fini of the binary, instead of allowing runtime linker to arrange the calls. This was probably done to have the same crt code serve both statically and dynamically linked binaries. Since ABI mandates that first is called preinit array functions, then init, and then init array functions, the init have to be called from rtld now. To provide binary compatibility to old FreeBSD crt1, which calls _init itself, rtld only calls intializers and finalizers for main binary if binary has a note indicating that new crt was used for linking. Add parsing of ELF notes to rtld, and cache p_osrel value since we parsed it anyway. The patch is inspired by init_array support for DragonflyBSD, written by John Marino. Reviewed by: kan Tested by: andrew (arm, previous version), flo (sparc64, previous version) MFC after: 3 weeks
2012-03-11 20:03:09 +00:00
/* FreeBSD 'crt does not call init' note */
obj->crt_no_init = true;
dbg("note crt_no_init");
break;
}
}
}
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
static Obj_Entry *
dlcheck(void *handle)
{
Obj_Entry *obj;
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
TAILQ_FOREACH(obj, &obj_list, next) {
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
if (obj == (Obj_Entry *) handle)
break;
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
}
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
Fix a bug in which a program called dlclose from a destructor and got an assert failure in the dynamic linker.
2001-01-05 04:36:17 +00:00
if (obj == NULL || obj->refcount == 0 || obj->dl_refcount == 0) {
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
_rtld_error("Invalid shared object handle %p", handle);
return NULL;
}
return obj;
}
Solve the dynamic linker's problems with multithreaded programs once and for all (I hope). Packages such as wine, JDK, and linuxthreads should no longer have any problems with re-entering the dynamic linker. This commit replaces the locking used in the dynamic linker with a new spinlock-based reader/writer lock implementation. Brian Fundakowski Feldman <green> argued for this from the very beginning, but it took me a long time to come around to his point of view. Spinlocks are the only kinds of locks that work with all thread packages. But on uniprocessor systems they can be inefficient, because while a contender for the lock is spinning the holder of the lock cannot make any progress toward releasing it. To alleviate this disadvantage I have borrowed a trick from Sleepycat's Berkeley DB implementation. When spinning for a lock, the requester does a nanosleep() call for 1 usec. each time around the loop. This will generally yield the CPU to other threads, allowing the lock holder to finish its business and release the lock. I chose 1 usec. as the minimum sleep which would with reasonable certainty not be rounded down to 0. The formerly machine-independent file "lockdflt.c" has been moved into the architecture-specific subdirectories by repository copy. It now contains the machine-dependent spinlocking code. For the spinlocks I used the very nifty "simple, non-scalable reader-preference lock" which I found at <http://www.cs.rochester.edu/u/scott/synchronization/pseudocode/rw.html> on all CPUs except the 80386 (the specific CPU model, not the architecture). The 80386 CPU doesn't support the necessary "cmpxchg" instruction, so on that CPU a simple exclusive test-and-set lock is used instead. 80386 CPUs are detected at initialization time by trying to execute "cmpxchg" and catching the resulting SIGILL signal. To reduce contention for the locks, I have revamped a couple of key data structures, permitting all common operations to be done under non-exclusive (reader) locking. The only operations that require exclusive locking now are the rare intrusive operations such as dlopen() and dlclose(). The dllockinit() interface is now deprecated. It still exists, but only as a do-nothing stub. I plan to remove it as soon as is reasonably possible. (From the very beginning it was clearly labeled as experimental and subject to change.) As far as I know, only the linuxthreads port uses dllockinit(). This interface turned out to have several problems. As one example, when the dynamic linker called a client-supplied locking function, that function sometimes needed lazy binding, causing re-entry into the dynamic linker and a big looping mess. And in any case, it turned out to be too burdensome to require threads packages to register themselves with the dynamic linker.
2000-07-08 04:10:38 +00:00
/*
* If the given object is already in the donelist, return true. Otherwise
* add the object to the list and return false.
*/
static bool
Add support for dlsym(RTLD_DEFAULT, ...).
2000-09-19 04:27:16 +00:00
donelist_check(DoneList *dlp, const Obj_Entry *obj)
Solve the dynamic linker's problems with multithreaded programs once and for all (I hope). Packages such as wine, JDK, and linuxthreads should no longer have any problems with re-entering the dynamic linker. This commit replaces the locking used in the dynamic linker with a new spinlock-based reader/writer lock implementation. Brian Fundakowski Feldman <green> argued for this from the very beginning, but it took me a long time to come around to his point of view. Spinlocks are the only kinds of locks that work with all thread packages. But on uniprocessor systems they can be inefficient, because while a contender for the lock is spinning the holder of the lock cannot make any progress toward releasing it. To alleviate this disadvantage I have borrowed a trick from Sleepycat's Berkeley DB implementation. When spinning for a lock, the requester does a nanosleep() call for 1 usec. each time around the loop. This will generally yield the CPU to other threads, allowing the lock holder to finish its business and release the lock. I chose 1 usec. as the minimum sleep which would with reasonable certainty not be rounded down to 0. The formerly machine-independent file "lockdflt.c" has been moved into the architecture-specific subdirectories by repository copy. It now contains the machine-dependent spinlocking code. For the spinlocks I used the very nifty "simple, non-scalable reader-preference lock" which I found at <http://www.cs.rochester.edu/u/scott/synchronization/pseudocode/rw.html> on all CPUs except the 80386 (the specific CPU model, not the architecture). The 80386 CPU doesn't support the necessary "cmpxchg" instruction, so on that CPU a simple exclusive test-and-set lock is used instead. 80386 CPUs are detected at initialization time by trying to execute "cmpxchg" and catching the resulting SIGILL signal. To reduce contention for the locks, I have revamped a couple of key data structures, permitting all common operations to be done under non-exclusive (reader) locking. The only operations that require exclusive locking now are the rare intrusive operations such as dlopen() and dlclose(). The dllockinit() interface is now deprecated. It still exists, but only as a do-nothing stub. I plan to remove it as soon as is reasonably possible. (From the very beginning it was clearly labeled as experimental and subject to change.) As far as I know, only the linuxthreads port uses dllockinit(). This interface turned out to have several problems. As one example, when the dynamic linker called a client-supplied locking function, that function sometimes needed lazy binding, causing re-entry into the dynamic linker and a big looping mess. And in any case, it turned out to be too burdensome to require threads packages to register themselves with the dynamic linker.
2000-07-08 04:10:38 +00:00
{
unsigned int i;
for (i = 0; i < dlp->num_used; i++)
if (dlp->objs[i] == obj)
return true;
/*
* Our donelist allocation should always be sufficient. But if
* our threads locking isn't working properly, more shared objects
* could have been loaded since we allocated the list. That should
* never happen, but we'll handle it properly just in case it does.
*/
if (dlp->num_used < dlp->num_alloc)
dlp->objs[dlp->num_used++] = obj;
return false;
}
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
/*
* Hash function for symbol table lookup. Don't even think about changing
* this. It is specified by the System V ABI.
*/
Add alpha support. Submitted by: John Birrell <jb@cimlogic.com.au> (with extra hacks by me) Obtained from: Probably NetBSD
1998-09-04 19:03:57 +00:00
unsigned long
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
elf_hash(const char *name)
{
const unsigned char *p = (const unsigned char *) name;
unsigned long h = 0;
unsigned long g;
while (*p != '\0') {
h = (h << 4) + *p++;
if ((g = h & 0xf0000000) != 0)
h ^= g >> 24;
h &= ~g;
}
return h;
}
Add GNU hash support for rtld. Based on dragonflybsd support for GNU hash by John Marino <draco marino st> Reviewed by: kan Tested by: bapt MFC after: 2 weeks
2012-04-30 13:31:10 +00:00
/*
* The GNU hash function is the Daniel J. Bernstein hash clipped to 32 bits
* unsigned in case it's implemented with a wider type.
*/
static uint32_t
gnu_hash(const char *s)
{
uint32_t h;
unsigned char c;
h = 5381;
for (c = *s; c != '\0'; c = *++s)
h = h * 33 + c;
return (h & 0xffffffff);
}
Add the LD_LIBRARY_PATH_FDS environmental variable. This variable allows the loading of shared libraries via directory descriptors rather than via library paths. If LD_LIBRARY_PATH_FDS=3:4:12, the directories represented by file descriptors 3, 4 and 12 will searched for shared libraries before the normal path-based mechanisms are used. This allows us to execute unprivileged binaries from within a Capsicum sandbox even if they require shared libraries. Approved by: rwatson (mentor) Reviewed by: kib MFC after: 3 weeks Sponsored by: DARPA/AFRL
2014-06-20 17:08:32 +00:00
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
/*
* Find the library with the given name, and return its full pathname.
* The returned string is dynamically allocated. Generates an error
* message and returns NULL if the library cannot be found.
*
* If the second argument is non-NULL, then it refers to an already-
* loaded shared object, whose library search path will be searched.
Implement ldconfig functionality for ELF. The hints are stored in a different file than the a.out hints, namely, "/var/run/ld-elf.so.hints". These hints consist only of the directory search path. There is no hash table as in the a.out hints, because ELF doesn't have to search for the file with the highest minor version number. (It doesn't have minor version numbers at all.) A single run of ldconfig updates either the a.out hints or the ELF hints, but not both. The set of hints to process is selected in the usual way, via /etc/objformat, or ${OBJFORMAT}, or the "-aout" or "-elf" command line option. The rationale is that you probably want to search different directories for ELF than for a.out. "ldconfig -r" is faked up to produce output like we are used to, except that for ELF there are no minor version numbers. This should enable "ldconfig -r" to be used for checking LIB_DEPENDS in ports even for ELF. I implemented the ELF functionality in a new source file, with an eye toward eliminating the a.out code entirely at some point in the future.
1998-09-05 03:31:00 +00:00
*
Add the LD_LIBRARY_PATH_FDS environmental variable. This variable allows the loading of shared libraries via directory descriptors rather than via library paths. If LD_LIBRARY_PATH_FDS=3:4:12, the directories represented by file descriptors 3, 4 and 12 will searched for shared libraries before the normal path-based mechanisms are used. This allows us to execute unprivileged binaries from within a Capsicum sandbox even if they require shared libraries. Approved by: rwatson (mentor) Reviewed by: kib MFC after: 3 weeks Sponsored by: DARPA/AFRL
2014-06-20 17:08:32 +00:00
* If a library is successfully located via LD_LIBRARY_PATH_FDS, its
* descriptor (which is close-on-exec) will be passed out via the third
* argument.
*
Implement ldconfig functionality for ELF. The hints are stored in a different file than the a.out hints, namely, "/var/run/ld-elf.so.hints". These hints consist only of the directory search path. There is no hash table as in the a.out hints, because ELF doesn't have to search for the file with the highest minor version number. (It doesn't have minor version numbers at all.) A single run of ldconfig updates either the a.out hints or the ELF hints, but not both. The set of hints to process is selected in the usual way, via /etc/objformat, or ${OBJFORMAT}, or the "-aout" or "-elf" command line option. The rationale is that you probably want to search different directories for ELF than for a.out. "ldconfig -r" is faked up to produce output like we are used to, except that for ELF there are no minor version numbers. This should enable "ldconfig -r" to be used for checking LIB_DEPENDS in ports even for ELF. I implemented the ELF functionality in a new source file, with an eye toward eliminating the a.out code entirely at some point in the future.
1998-09-05 03:31:00 +00:00
* The search order is:
Import the DragonFly BSD commit 4f0bc915b65fcf5a23214f6d221d65c80be68ad4 by John Marino <draco@marino.st>, with the following (edited) commit message Date: Sat, 24 Mar 2012 06:40:50 +0100 Subject: [PATCH 1/1] rtld: Implement DT_RUNPATH and -z nodefaultlib DT_RUNPATH is incorrectly being considered as an alias of DT_RPATH. The purpose of DT_RUNPATH is to have two different types of rpath: one that can be overridden by the environment variable LD_LIBRARY_PATH and one that can't. With the currently implementation, LD_LIBRARY_PATH will always trump any embedded rpath or runpath tags. Current path search order by rtld: ================================== LD_LIBRARY_PATH DT_RPATH / DT_RUNPATH (always the same) ldconfig hints file (default: /var/run/ld-elf.so.hints) /usr/lib New path search order by rtld: ============================== DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file /usr/lib The new path search matches how the linux runtime loader works. The other major added feature is support for linker flag "-z nodefaultlib". When this flag is passed to the linker, rtld will skip all references to the standard library search path ("/usr/lib" in this case but it could handle more color delimited paths) except in DT_RPATH and DT_RUNPATH. New path search order by rtld with -z nodefaultlib flag set: ============================================================ DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file (skips all references to /usr/lib) FreeBSD notes: - we fixed some bugs which were submitted to DragonFly and merged there as commit 1ff8a2bd3eb6e5587174c6a983303ea3a79e0002; - we added LD_LIBRARY_PATH_RPATH environment variable to switch to the previous behaviour of considering DT_RPATH a synonym for DT_RUNPATH; - the FreeBSD default search path is /lib:/usr/lib and not /usr/lib. Reviewed by: kan MFC after: 1 month MFC note: flip the ld_library_path_rpath default value for stable/9
2012-07-15 10:53:48 +00:00
* DT_RPATH in the referencing file _unless_ DT_RUNPATH is present (1)
* DT_RPATH of the main object if DSO without defined DT_RUNPATH (1)
Implement ldconfig functionality for ELF. The hints are stored in a different file than the a.out hints, namely, "/var/run/ld-elf.so.hints". These hints consist only of the directory search path. There is no hash table as in the a.out hints, because ELF doesn't have to search for the file with the highest minor version number. (It doesn't have minor version numbers at all.) A single run of ldconfig updates either the a.out hints or the ELF hints, but not both. The set of hints to process is selected in the usual way, via /etc/objformat, or ${OBJFORMAT}, or the "-aout" or "-elf" command line option. The rationale is that you probably want to search different directories for ELF than for a.out. "ldconfig -r" is faked up to produce output like we are used to, except that for ELF there are no minor version numbers. This should enable "ldconfig -r" to be used for checking LIB_DEPENDS in ports even for ELF. I implemented the ELF functionality in a new source file, with an eye toward eliminating the a.out code entirely at some point in the future.
1998-09-05 03:31:00 +00:00
* LD_LIBRARY_PATH
Import the DragonFly BSD commit 4f0bc915b65fcf5a23214f6d221d65c80be68ad4 by John Marino <draco@marino.st>, with the following (edited) commit message Date: Sat, 24 Mar 2012 06:40:50 +0100 Subject: [PATCH 1/1] rtld: Implement DT_RUNPATH and -z nodefaultlib DT_RUNPATH is incorrectly being considered as an alias of DT_RPATH. The purpose of DT_RUNPATH is to have two different types of rpath: one that can be overridden by the environment variable LD_LIBRARY_PATH and one that can't. With the currently implementation, LD_LIBRARY_PATH will always trump any embedded rpath or runpath tags. Current path search order by rtld: ================================== LD_LIBRARY_PATH DT_RPATH / DT_RUNPATH (always the same) ldconfig hints file (default: /var/run/ld-elf.so.hints) /usr/lib New path search order by rtld: ============================== DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file /usr/lib The new path search matches how the linux runtime loader works. The other major added feature is support for linker flag "-z nodefaultlib". When this flag is passed to the linker, rtld will skip all references to the standard library search path ("/usr/lib" in this case but it could handle more color delimited paths) except in DT_RPATH and DT_RUNPATH. New path search order by rtld with -z nodefaultlib flag set: ============================================================ DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file (skips all references to /usr/lib) FreeBSD notes: - we fixed some bugs which were submitted to DragonFly and merged there as commit 1ff8a2bd3eb6e5587174c6a983303ea3a79e0002; - we added LD_LIBRARY_PATH_RPATH environment variable to switch to the previous behaviour of considering DT_RPATH a synonym for DT_RUNPATH; - the FreeBSD default search path is /lib:/usr/lib and not /usr/lib. Reviewed by: kan MFC after: 1 month MFC note: flip the ld_library_path_rpath default value for stable/9
2012-07-15 10:53:48 +00:00
* DT_RUNPATH in the referencing file
* ldconfig hints (if -z nodefaultlib, filter out default library directories
* from list)
* /lib:/usr/lib _unless_ the referencing file is linked with -z nodefaultlib
*
* (1) Handled in digest_dynamic2 - rpath left NULL if runpath defined.
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
*/
static char *
Add the LD_LIBRARY_PATH_FDS environmental variable. This variable allows the loading of shared libraries via directory descriptors rather than via library paths. If LD_LIBRARY_PATH_FDS=3:4:12, the directories represented by file descriptors 3, 4 and 12 will searched for shared libraries before the normal path-based mechanisms are used. This allows us to execute unprivileged binaries from within a Capsicum sandbox even if they require shared libraries. Approved by: rwatson (mentor) Reviewed by: kib MFC after: 3 weeks Sponsored by: DARPA/AFRL
2014-06-20 17:08:32 +00:00
find_library(const char *xname, const Obj_Entry *refobj, int *fdp)
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
{
char *pathname;
Dynamic object dependency mapping: libmap. This is an optional feature, disabled by default. This will be useful to people testing the various POSIX threading libraries under -CURRENT but can easily serve other needs.
2003-04-07 16:21:26 +00:00
char *name;
Do not reference z_nodeflib for !objgiven case, thus fixing LD_PRELOAD for a non-absolute path. PR: bin/171604 MFC after: 3 days
2012-09-19 05:11:25 +00:00
bool nodeflib, objgiven;
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
Import the DragonFly BSD commit 4f0bc915b65fcf5a23214f6d221d65c80be68ad4 by John Marino <draco@marino.st>, with the following (edited) commit message Date: Sat, 24 Mar 2012 06:40:50 +0100 Subject: [PATCH 1/1] rtld: Implement DT_RUNPATH and -z nodefaultlib DT_RUNPATH is incorrectly being considered as an alias of DT_RPATH. The purpose of DT_RUNPATH is to have two different types of rpath: one that can be overridden by the environment variable LD_LIBRARY_PATH and one that can't. With the currently implementation, LD_LIBRARY_PATH will always trump any embedded rpath or runpath tags. Current path search order by rtld: ================================== LD_LIBRARY_PATH DT_RPATH / DT_RUNPATH (always the same) ldconfig hints file (default: /var/run/ld-elf.so.hints) /usr/lib New path search order by rtld: ============================== DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file /usr/lib The new path search matches how the linux runtime loader works. The other major added feature is support for linker flag "-z nodefaultlib". When this flag is passed to the linker, rtld will skip all references to the standard library search path ("/usr/lib" in this case but it could handle more color delimited paths) except in DT_RPATH and DT_RUNPATH. New path search order by rtld with -z nodefaultlib flag set: ============================================================ DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file (skips all references to /usr/lib) FreeBSD notes: - we fixed some bugs which were submitted to DragonFly and merged there as commit 1ff8a2bd3eb6e5587174c6a983303ea3a79e0002; - we added LD_LIBRARY_PATH_RPATH environment variable to switch to the previous behaviour of considering DT_RPATH a synonym for DT_RUNPATH; - the FreeBSD default search path is /lib:/usr/lib and not /usr/lib. Reviewed by: kan MFC after: 1 month MFC note: flip the ld_library_path_rpath default value for stable/9
2012-07-15 10:53:48 +00:00
objgiven = refobj != NULL;
Allow to specify targets by absolute paths in libmap.conf. Submitted by: Tatu Kilappa <tatu.kilappa@iki.fi> PR: 221032 MFC after: 2 weeks
2017-07-27 08:33:31 +00:00
if (libmap_disable || !objgiven ||
(name = lm_find(refobj->path, xname)) == NULL)
name = (char *)xname;
if (strchr(name, '/') != NULL) { /* Hard coded pathname */
if (name[0] != '/' && !trust) {
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
_rtld_error("Absolute pathname required for shared object \"%s\"",
Allow to specify targets by absolute paths in libmap.conf. Submitted by: Tatu Kilappa <tatu.kilappa@iki.fi> PR: 221032 MFC after: 2 weeks
2017-07-27 08:33:31 +00:00
name);
return (NULL);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
}
Change interpretation of the DF_ORIGIN and DF_1_ORIGIN flags. According to standard, the presence of the flags only means that the object path must be resolved at the time object loading, instead of my reading that the flag is required to enable token substitution at all. The consequence is that -z origin linker flag is no longer required for the token substitution in the run/rpath or the needed library soname. It is only recommended if token substition is needed at dlopen(3) time, since namecache might drop the required entries at the time of resolution. Found, reviewed and tested by: emaste Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-04-27 18:41:31 +00:00
return (origin_subst(__DECONST(Obj_Entry *, refobj),
Allow to specify targets by absolute paths in libmap.conf. Submitted by: Tatu Kilappa <tatu.kilappa@iki.fi> PR: 221032 MFC after: 2 weeks
2017-07-27 08:33:31 +00:00
__DECONST(char *, name)));
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
}
dbg(" Searching for \"%s\"", name);
Import the DragonFly BSD commit 4f0bc915b65fcf5a23214f6d221d65c80be68ad4 by John Marino <draco@marino.st>, with the following (edited) commit message Date: Sat, 24 Mar 2012 06:40:50 +0100 Subject: [PATCH 1/1] rtld: Implement DT_RUNPATH and -z nodefaultlib DT_RUNPATH is incorrectly being considered as an alias of DT_RPATH. The purpose of DT_RUNPATH is to have two different types of rpath: one that can be overridden by the environment variable LD_LIBRARY_PATH and one that can't. With the currently implementation, LD_LIBRARY_PATH will always trump any embedded rpath or runpath tags. Current path search order by rtld: ================================== LD_LIBRARY_PATH DT_RPATH / DT_RUNPATH (always the same) ldconfig hints file (default: /var/run/ld-elf.so.hints) /usr/lib New path search order by rtld: ============================== DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file /usr/lib The new path search matches how the linux runtime loader works. The other major added feature is support for linker flag "-z nodefaultlib". When this flag is passed to the linker, rtld will skip all references to the standard library search path ("/usr/lib" in this case but it could handle more color delimited paths) except in DT_RPATH and DT_RUNPATH. New path search order by rtld with -z nodefaultlib flag set: ============================================================ DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file (skips all references to /usr/lib) FreeBSD notes: - we fixed some bugs which were submitted to DragonFly and merged there as commit 1ff8a2bd3eb6e5587174c6a983303ea3a79e0002; - we added LD_LIBRARY_PATH_RPATH environment variable to switch to the previous behaviour of considering DT_RPATH a synonym for DT_RUNPATH; - the FreeBSD default search path is /lib:/usr/lib and not /usr/lib. Reviewed by: kan MFC after: 1 month MFC note: flip the ld_library_path_rpath default value for stable/9
2012-07-15 10:53:48 +00:00
/*
* If refobj->rpath != NULL, then refobj->runpath is NULL. Fall
* back to pre-conforming behaviour if user requested so with
* LD_LIBRARY_PATH_RPATH environment variable and ignore -z
* nodeflib.
*/
if (objgiven && refobj->rpath != NULL && ld_library_path_rpath) {
if ((pathname = search_library_path(name, ld_library_path)) != NULL ||
(refobj != NULL &&
(pathname = search_library_path(name, refobj->rpath)) != NULL) ||
Add the LD_LIBRARY_PATH_FDS environmental variable. This variable allows the loading of shared libraries via directory descriptors rather than via library paths. If LD_LIBRARY_PATH_FDS=3:4:12, the directories represented by file descriptors 3, 4 and 12 will searched for shared libraries before the normal path-based mechanisms are used. This allows us to execute unprivileged binaries from within a Capsicum sandbox even if they require shared libraries. Approved by: rwatson (mentor) Reviewed by: kib MFC after: 3 weeks Sponsored by: DARPA/AFRL
2014-06-20 17:08:32 +00:00
(pathname = search_library_pathfds(name, ld_library_dirs, fdp)) != NULL ||
Import the DragonFly BSD commit 4f0bc915b65fcf5a23214f6d221d65c80be68ad4 by John Marino <draco@marino.st>, with the following (edited) commit message Date: Sat, 24 Mar 2012 06:40:50 +0100 Subject: [PATCH 1/1] rtld: Implement DT_RUNPATH and -z nodefaultlib DT_RUNPATH is incorrectly being considered as an alias of DT_RPATH. The purpose of DT_RUNPATH is to have two different types of rpath: one that can be overridden by the environment variable LD_LIBRARY_PATH and one that can't. With the currently implementation, LD_LIBRARY_PATH will always trump any embedded rpath or runpath tags. Current path search order by rtld: ================================== LD_LIBRARY_PATH DT_RPATH / DT_RUNPATH (always the same) ldconfig hints file (default: /var/run/ld-elf.so.hints) /usr/lib New path search order by rtld: ============================== DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file /usr/lib The new path search matches how the linux runtime loader works. The other major added feature is support for linker flag "-z nodefaultlib". When this flag is passed to the linker, rtld will skip all references to the standard library search path ("/usr/lib" in this case but it could handle more color delimited paths) except in DT_RPATH and DT_RUNPATH. New path search order by rtld with -z nodefaultlib flag set: ============================================================ DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file (skips all references to /usr/lib) FreeBSD notes: - we fixed some bugs which were submitted to DragonFly and merged there as commit 1ff8a2bd3eb6e5587174c6a983303ea3a79e0002; - we added LD_LIBRARY_PATH_RPATH environment variable to switch to the previous behaviour of considering DT_RPATH a synonym for DT_RUNPATH; - the FreeBSD default search path is /lib:/usr/lib and not /usr/lib. Reviewed by: kan MFC after: 1 month MFC note: flip the ld_library_path_rpath default value for stable/9
2012-07-15 10:53:48 +00:00
(pathname = search_library_path(name, gethints(false))) != NULL ||
Rather than using the #define for path names, indirect through a char * variable that could change for different executable types detected.
2015-10-31 04:39:55 +00:00
(pathname = search_library_path(name, ld_standard_library_path)) != NULL)
Import the DragonFly BSD commit 4f0bc915b65fcf5a23214f6d221d65c80be68ad4 by John Marino <draco@marino.st>, with the following (edited) commit message Date: Sat, 24 Mar 2012 06:40:50 +0100 Subject: [PATCH 1/1] rtld: Implement DT_RUNPATH and -z nodefaultlib DT_RUNPATH is incorrectly being considered as an alias of DT_RPATH. The purpose of DT_RUNPATH is to have two different types of rpath: one that can be overridden by the environment variable LD_LIBRARY_PATH and one that can't. With the currently implementation, LD_LIBRARY_PATH will always trump any embedded rpath or runpath tags. Current path search order by rtld: ================================== LD_LIBRARY_PATH DT_RPATH / DT_RUNPATH (always the same) ldconfig hints file (default: /var/run/ld-elf.so.hints) /usr/lib New path search order by rtld: ============================== DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file /usr/lib The new path search matches how the linux runtime loader works. The other major added feature is support for linker flag "-z nodefaultlib". When this flag is passed to the linker, rtld will skip all references to the standard library search path ("/usr/lib" in this case but it could handle more color delimited paths) except in DT_RPATH and DT_RUNPATH. New path search order by rtld with -z nodefaultlib flag set: ============================================================ DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file (skips all references to /usr/lib) FreeBSD notes: - we fixed some bugs which were submitted to DragonFly and merged there as commit 1ff8a2bd3eb6e5587174c6a983303ea3a79e0002; - we added LD_LIBRARY_PATH_RPATH environment variable to switch to the previous behaviour of considering DT_RPATH a synonym for DT_RUNPATH; - the FreeBSD default search path is /lib:/usr/lib and not /usr/lib. Reviewed by: kan MFC after: 1 month MFC note: flip the ld_library_path_rpath default value for stable/9
2012-07-15 10:53:48 +00:00
return (pathname);
} else {
Do not reference z_nodeflib for !objgiven case, thus fixing LD_PRELOAD for a non-absolute path. PR: bin/171604 MFC after: 3 days
2012-09-19 05:11:25 +00:00
nodeflib = objgiven ? refobj->z_nodeflib : false;
Import the DragonFly BSD commit 4f0bc915b65fcf5a23214f6d221d65c80be68ad4 by John Marino <draco@marino.st>, with the following (edited) commit message Date: Sat, 24 Mar 2012 06:40:50 +0100 Subject: [PATCH 1/1] rtld: Implement DT_RUNPATH and -z nodefaultlib DT_RUNPATH is incorrectly being considered as an alias of DT_RPATH. The purpose of DT_RUNPATH is to have two different types of rpath: one that can be overridden by the environment variable LD_LIBRARY_PATH and one that can't. With the currently implementation, LD_LIBRARY_PATH will always trump any embedded rpath or runpath tags. Current path search order by rtld: ================================== LD_LIBRARY_PATH DT_RPATH / DT_RUNPATH (always the same) ldconfig hints file (default: /var/run/ld-elf.so.hints) /usr/lib New path search order by rtld: ============================== DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file /usr/lib The new path search matches how the linux runtime loader works. The other major added feature is support for linker flag "-z nodefaultlib". When this flag is passed to the linker, rtld will skip all references to the standard library search path ("/usr/lib" in this case but it could handle more color delimited paths) except in DT_RPATH and DT_RUNPATH. New path search order by rtld with -z nodefaultlib flag set: ============================================================ DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file (skips all references to /usr/lib) FreeBSD notes: - we fixed some bugs which were submitted to DragonFly and merged there as commit 1ff8a2bd3eb6e5587174c6a983303ea3a79e0002; - we added LD_LIBRARY_PATH_RPATH environment variable to switch to the previous behaviour of considering DT_RPATH a synonym for DT_RUNPATH; - the FreeBSD default search path is /lib:/usr/lib and not /usr/lib. Reviewed by: kan MFC after: 1 month MFC note: flip the ld_library_path_rpath default value for stable/9
2012-07-15 10:53:48 +00:00
if ((objgiven &&
(pathname = search_library_path(name, refobj->rpath)) != NULL) ||
(objgiven && refobj->runpath == NULL && refobj != obj_main &&
(pathname = search_library_path(name, obj_main->rpath)) != NULL) ||
(pathname = search_library_path(name, ld_library_path)) != NULL ||
(objgiven &&
(pathname = search_library_path(name, refobj->runpath)) != NULL) ||
Add the LD_LIBRARY_PATH_FDS environmental variable. This variable allows the loading of shared libraries via directory descriptors rather than via library paths. If LD_LIBRARY_PATH_FDS=3:4:12, the directories represented by file descriptors 3, 4 and 12 will searched for shared libraries before the normal path-based mechanisms are used. This allows us to execute unprivileged binaries from within a Capsicum sandbox even if they require shared libraries. Approved by: rwatson (mentor) Reviewed by: kib MFC after: 3 weeks Sponsored by: DARPA/AFRL
2014-06-20 17:08:32 +00:00
(pathname = search_library_pathfds(name, ld_library_dirs, fdp)) != NULL ||
Do not reference z_nodeflib for !objgiven case, thus fixing LD_PRELOAD for a non-absolute path. PR: bin/171604 MFC after: 3 days
2012-09-19 05:11:25 +00:00
(pathname = search_library_path(name, gethints(nodeflib))) != NULL ||
(objgiven && !nodeflib &&
Rather than using the #define for path names, indirect through a char * variable that could change for different executable types detected.
2015-10-31 04:39:55 +00:00
(pathname = search_library_path(name, ld_standard_library_path)) != NULL))
Import the DragonFly BSD commit 4f0bc915b65fcf5a23214f6d221d65c80be68ad4 by John Marino <draco@marino.st>, with the following (edited) commit message Date: Sat, 24 Mar 2012 06:40:50 +0100 Subject: [PATCH 1/1] rtld: Implement DT_RUNPATH and -z nodefaultlib DT_RUNPATH is incorrectly being considered as an alias of DT_RPATH. The purpose of DT_RUNPATH is to have two different types of rpath: one that can be overridden by the environment variable LD_LIBRARY_PATH and one that can't. With the currently implementation, LD_LIBRARY_PATH will always trump any embedded rpath or runpath tags. Current path search order by rtld: ================================== LD_LIBRARY_PATH DT_RPATH / DT_RUNPATH (always the same) ldconfig hints file (default: /var/run/ld-elf.so.hints) /usr/lib New path search order by rtld: ============================== DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file /usr/lib The new path search matches how the linux runtime loader works. The other major added feature is support for linker flag "-z nodefaultlib". When this flag is passed to the linker, rtld will skip all references to the standard library search path ("/usr/lib" in this case but it could handle more color delimited paths) except in DT_RPATH and DT_RUNPATH. New path search order by rtld with -z nodefaultlib flag set: ============================================================ DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file (skips all references to /usr/lib) FreeBSD notes: - we fixed some bugs which were submitted to DragonFly and merged there as commit 1ff8a2bd3eb6e5587174c6a983303ea3a79e0002; - we added LD_LIBRARY_PATH_RPATH environment variable to switch to the previous behaviour of considering DT_RPATH a synonym for DT_RUNPATH; - the FreeBSD default search path is /lib:/usr/lib and not /usr/lib. Reviewed by: kan MFC after: 1 month MFC note: flip the ld_library_path_rpath default value for stable/9
2012-07-15 10:53:48 +00:00
return (pathname);
}
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
Import the DragonFly BSD commit 4f0bc915b65fcf5a23214f6d221d65c80be68ad4 by John Marino <draco@marino.st>, with the following (edited) commit message Date: Sat, 24 Mar 2012 06:40:50 +0100 Subject: [PATCH 1/1] rtld: Implement DT_RUNPATH and -z nodefaultlib DT_RUNPATH is incorrectly being considered as an alias of DT_RPATH. The purpose of DT_RUNPATH is to have two different types of rpath: one that can be overridden by the environment variable LD_LIBRARY_PATH and one that can't. With the currently implementation, LD_LIBRARY_PATH will always trump any embedded rpath or runpath tags. Current path search order by rtld: ================================== LD_LIBRARY_PATH DT_RPATH / DT_RUNPATH (always the same) ldconfig hints file (default: /var/run/ld-elf.so.hints) /usr/lib New path search order by rtld: ============================== DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file /usr/lib The new path search matches how the linux runtime loader works. The other major added feature is support for linker flag "-z nodefaultlib". When this flag is passed to the linker, rtld will skip all references to the standard library search path ("/usr/lib" in this case but it could handle more color delimited paths) except in DT_RPATH and DT_RUNPATH. New path search order by rtld with -z nodefaultlib flag set: ============================================================ DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file (skips all references to /usr/lib) FreeBSD notes: - we fixed some bugs which were submitted to DragonFly and merged there as commit 1ff8a2bd3eb6e5587174c6a983303ea3a79e0002; - we added LD_LIBRARY_PATH_RPATH environment variable to switch to the previous behaviour of considering DT_RPATH a synonym for DT_RUNPATH; - the FreeBSD default search path is /lib:/usr/lib and not /usr/lib. Reviewed by: kan MFC after: 1 month MFC note: flip the ld_library_path_rpath default value for stable/9
2012-07-15 10:53:48 +00:00
if (objgiven && refobj->path != NULL) {
give out a little more information in case of a missing dependency PR: 56549 Submitted by: edwin Reviewed by: joerg, ru Approved by: joerg MFC after: 2 weeks
2004-05-28 00:05:28 +00:00
_rtld_error("Shared object \"%s\" not found, required by \"%s\"",
name, basename(refobj->path));
} else {
_rtld_error("Shared object \"%s\" not found", name);
}
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
return NULL;
}
/*
* Given a symbol number in a referencing object, find the corresponding
* definition of the symbol. Returns a pointer to the symbol, or NULL if
* no definition was found. Returns a pointer to the Obj_Entry of the
* defining object via the reference parameter DEFOBJ_OUT.
*/
Add alpha support. Submitted by: John Birrell <jb@cimlogic.com.au> (with extra hacks by me) Obtained from: Probably NetBSD
1998-09-04 19:03:57 +00:00
const Elf_Sym *
Add support for dlsym(RTLD_DEFAULT, ...).
2000-09-19 04:27:16 +00:00
find_symdef(unsigned long symnum, const Obj_Entry *refobj,
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
const Obj_Entry **defobj_out, int flags, SymCache *cache,
RtldLockState *lockstate)
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
{
Add alpha support. Submitted by: John Birrell <jb@cimlogic.com.au> (with extra hacks by me) Obtained from: Probably NetBSD
1998-09-04 19:03:57 +00:00
const Elf_Sym *ref;
Simplify the logic in find_symdef().
1999-08-30 01:24:08 +00:00
const Elf_Sym *def;
const Obj_Entry *defobj;
When reporting undefined symbol, note the version, if specified. Use the standard syntax of name@version, I do not expect a confusion due to unlikely possibility of the name containing the '@' character. Requested by: emaste Sponsored by: The FreeBSD Foundation MFC after: 1 week
2017-07-04 20:19:36 +00:00
const Ver_Entry *ve;
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
SymLook req;
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
const char *name;
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
int res;
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
Performance improvements for the ELF dynamic linker. These particularly help programs which load many shared libraries with a lot of relocations. Large C++ programs such as are found in KDE are a prime example. While relocating a shared object, maintain a vector of symbols which have already been looked up, directly indexed by symbol number. Typically, symbols which are referenced by a relocation entry are referenced by many of them. This is the same optimization I made to the a.out dynamic linker in 1995 (rtld.c revision 1.30). Also, compare the first character of a sought-after symbol with its symbol table entry before calling strcmp(). On a PII/400 these changes reduce the start-up time of a typical KDE program from 833 msec (elapsed) to 370 msec. MFC after: 5 days
2001-05-05 23:21:05 +00:00
/*
* If we have already found this symbol, get the information from
* the cache.
*/
Add GNU hash support for rtld. Based on dragonflybsd support for GNU hash by John Marino <draco marino st> Reviewed by: kan Tested by: bapt MFC after: 2 weeks
2012-04-30 13:31:10 +00:00
if (symnum >= refobj->dynsymcount)
Performance improvements for the ELF dynamic linker. These particularly help programs which load many shared libraries with a lot of relocations. Large C++ programs such as are found in KDE are a prime example. While relocating a shared object, maintain a vector of symbols which have already been looked up, directly indexed by symbol number. Typically, symbols which are referenced by a relocation entry are referenced by many of them. This is the same optimization I made to the a.out dynamic linker in 1995 (rtld.c revision 1.30). Also, compare the first character of a sought-after symbol with its symbol table entry before calling strcmp(). On a PII/400 these changes reduce the start-up time of a typical KDE program from 833 msec (elapsed) to 370 msec. MFC after: 5 days
2001-05-05 23:21:05 +00:00
return NULL; /* Bad object */
if (cache != NULL && cache[symnum].sym != NULL) {
*defobj_out = cache[symnum].obj;
return cache[symnum].sym;
}
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
ref = refobj->symtab + symnum;
name = refobj->strtab + ref->st_name;
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
def = NULL;
When honoring -Bsymbolic, still keep searching if only a weak definition was found in the referencing object.
1999-08-30 01:25:38 +00:00
defobj = NULL;
When reporting undefined symbol, note the version, if specified. Use the standard syntax of name@version, I do not expect a confusion due to unlikely possibility of the name containing the '@' character. Requested by: emaste Sponsored by: The FreeBSD Foundation MFC after: 1 week
2017-07-04 20:19:36 +00:00
ve = NULL;
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
Don't do symbol lookups for local symbols. The symbol index in the relocation identifies the symbol to which we need to bind. This solves a problem seen on ia64 where the symbol hash table does not contain local symbols and thus resulted in unresolved symbols. Tested on: alpha, i386, ia64
2002-04-27 02:48:29 +00:00
/*
* We don't have to do a full scale lookup if the symbol is local.
* We know it will bind to the instance in this load module; to
* which we already have a pointer (ie ref). By not doing a lookup,
* we not only improve performance, but it also avoids unresolvable
* symbols when local symbols are not in the hash table. This has
* been seen with the ia64 toolchain.
*/
if (ELF_ST_BIND(ref->st_info) != STB_LOCAL) {
if (ELF_ST_TYPE(ref->st_info) == STT_SECTION) {
Minor changes to make this work on sparc64. Approved by: jdp Tested on: alpha, i386, sparc64
2002-04-02 02:19:02 +00:00
_rtld_error("%s: Bogus symbol table entry %lu", refobj->path,
symnum);
}
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
symlook_init(&req, name);
req.flags = flags;
When reporting undefined symbol, note the version, if specified. Use the standard syntax of name@version, I do not expect a confusion due to unlikely possibility of the name containing the '@' character. Requested by: emaste Sponsored by: The FreeBSD Foundation MFC after: 1 week
2017-07-04 20:19:36 +00:00
ve = req.ventry = fetch_ventry(refobj, symnum);
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
req.lockstate = lockstate;
res = symlook_default(&req, refobj);
if (res == 0) {
def = req.sym_out;
defobj = req.defobj_out;
}
Don't do symbol lookups for local symbols. The symbol index in the relocation identifies the symbol to which we need to bind. This solves a problem seen on ia64 where the symbol hash table does not contain local symbols and thus resulted in unresolved symbols. Tested on: alpha, i386, ia64
2002-04-27 02:48:29 +00:00
} else {
Minor changes to make this work on sparc64. Approved by: jdp Tested on: alpha, i386, sparc64
2002-04-02 02:19:02 +00:00
def = ref;
defobj = refobj;
Don't do symbol lookups for local symbols. The symbol index in the relocation identifies the symbol to which we need to bind. This solves a problem seen on ia64 where the symbol hash table does not contain local symbols and thus resulted in unresolved symbols. Tested on: alpha, i386, ia64
2002-04-27 02:48:29 +00:00
}
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
Simplify the logic in find_symdef().
1999-08-30 01:24:08 +00:00
/*
Revamp the symbol lookup algorithm to cope better with objects loaded separately by dlopen that have global symbols with identical names. Viewing each dlopened object as a DAG which is linked by its DT_NEEDED entries in the dynamic table, the search order is as follows: * If the referencing object was linked with -Bsymbolic, search it internally. * Search all dlopened DAGs containing the referencing object. * Search all objects loaded at program start up. * Search all objects which were dlopened() using the RTLD_GLOBAL flag (which is now supported too). The search terminates as soon as a strong definition is found. Lacking that, the first weak definition is used. These rules match those of Solaris, as best I could determine them from its vague manual pages and the results of experiments I performed. PR: misc/12438
1999-08-30 01:48:19 +00:00
* If we found no definition and the reference is weak, treat the
Simplify the logic in find_symdef().
1999-08-30 01:24:08 +00:00
* symbol as having the value zero.
*/
Revamp the symbol lookup algorithm to cope better with objects loaded separately by dlopen that have global symbols with identical names. Viewing each dlopened object as a DAG which is linked by its DT_NEEDED entries in the dynamic table, the search order is as follows: * If the referencing object was linked with -Bsymbolic, search it internally. * Search all dlopened DAGs containing the referencing object. * Search all objects loaded at program start up. * Search all objects which were dlopened() using the RTLD_GLOBAL flag (which is now supported too). The search terminates as soon as a strong definition is found. Lacking that, the first weak definition is used. These rules match those of Solaris, as best I could determine them from its vague manual pages and the results of experiments I performed. PR: misc/12438
1999-08-30 01:48:19 +00:00
if (def == NULL && ELF_ST_BIND(ref->st_info) == STB_WEAK) {
def = &sym_zero;
defobj = obj_main;
Resolve undefined weak references to a value of 0. This solves the "__deregister_frame_info" problem that was seen when combining a program linked using the old gcc with shared libraries that were built using egcs.
1999-04-05 02:36:40 +00:00
}
Performance improvements for the ELF dynamic linker. These particularly help programs which load many shared libraries with a lot of relocations. Large C++ programs such as are found in KDE are a prime example. While relocating a shared object, maintain a vector of symbols which have already been looked up, directly indexed by symbol number. Typically, symbols which are referenced by a relocation entry are referenced by many of them. This is the same optimization I made to the a.out dynamic linker in 1995 (rtld.c revision 1.30). Also, compare the first character of a sought-after symbol with its symbol table entry before calling strcmp(). On a PII/400 these changes reduce the start-up time of a typical KDE program from 833 msec (elapsed) to 370 msec. MFC after: 5 days
2001-05-05 23:21:05 +00:00
if (def != NULL) {
Revamp the symbol lookup algorithm to cope better with objects loaded separately by dlopen that have global symbols with identical names. Viewing each dlopened object as a DAG which is linked by its DT_NEEDED entries in the dynamic table, the search order is as follows: * If the referencing object was linked with -Bsymbolic, search it internally. * Search all dlopened DAGs containing the referencing object. * Search all objects loaded at program start up. * Search all objects which were dlopened() using the RTLD_GLOBAL flag (which is now supported too). The search terminates as soon as a strong definition is found. Lacking that, the first weak definition is used. These rules match those of Solaris, as best I could determine them from its vague manual pages and the results of experiments I performed. PR: misc/12438
1999-08-30 01:48:19 +00:00
*defobj_out = defobj;
Performance improvements for the ELF dynamic linker. These particularly help programs which load many shared libraries with a lot of relocations. Large C++ programs such as are found in KDE are a prime example. While relocating a shared object, maintain a vector of symbols which have already been looked up, directly indexed by symbol number. Typically, symbols which are referenced by a relocation entry are referenced by many of them. This is the same optimization I made to the a.out dynamic linker in 1995 (rtld.c revision 1.30). Also, compare the first character of a sought-after symbol with its symbol table entry before calling strcmp(). On a PII/400 these changes reduce the start-up time of a typical KDE program from 833 msec (elapsed) to 370 msec. MFC after: 5 days
2001-05-05 23:21:05 +00:00
/* Record the information in the cache to avoid subsequent lookups. */
if (cache != NULL) {
cache[symnum].sym = def;
cache[symnum].obj = defobj;
}
Add ia64 support. Various adjustments were made to existing targets to cope with a few interface changes required by the ia64. In particular, function pointers on ia64 need special treatment in rtld.
2001-10-15 18:48:42 +00:00
} else {
if (refobj != &obj_rtld)
When reporting undefined symbol, note the version, if specified. Use the standard syntax of name@version, I do not expect a confusion due to unlikely possibility of the name containing the '@' character. Requested by: emaste Sponsored by: The FreeBSD Foundation MFC after: 1 week
2017-07-04 20:19:36 +00:00
_rtld_error("%s: Undefined symbol \"%s%s%s\"", refobj->path, name,
ve != NULL ? "@" : "", ve != NULL ? ve->name : "");
Add ia64 support. Various adjustments were made to existing targets to cope with a few interface changes required by the ia64. In particular, function pointers on ia64 need special treatment in rtld.
2001-10-15 18:48:42 +00:00
}
Revamp the symbol lookup algorithm to cope better with objects loaded separately by dlopen that have global symbols with identical names. Viewing each dlopened object as a DAG which is linked by its DT_NEEDED entries in the dynamic table, the search order is as follows: * If the referencing object was linked with -Bsymbolic, search it internally. * Search all dlopened DAGs containing the referencing object. * Search all objects loaded at program start up. * Search all objects which were dlopened() using the RTLD_GLOBAL flag (which is now supported too). The search terminates as soon as a strong definition is found. Lacking that, the first weak definition is used. These rules match those of Solaris, as best I could determine them from its vague manual pages and the results of experiments I performed. PR: misc/12438
1999-08-30 01:48:19 +00:00
return def;
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
}
Implement ldconfig functionality for ELF. The hints are stored in a different file than the a.out hints, namely, "/var/run/ld-elf.so.hints". These hints consist only of the directory search path. There is no hash table as in the a.out hints, because ELF doesn't have to search for the file with the highest minor version number. (It doesn't have minor version numbers at all.) A single run of ldconfig updates either the a.out hints or the ELF hints, but not both. The set of hints to process is selected in the usual way, via /etc/objformat, or ${OBJFORMAT}, or the "-aout" or "-elf" command line option. The rationale is that you probably want to search different directories for ELF than for a.out. "ldconfig -r" is faked up to produce output like we are used to, except that for ELF there are no minor version numbers. This should enable "ldconfig -r" to be used for checking LIB_DEPENDS in ports even for ELF. I implemented the ELF functionality in a new source file, with an eye toward eliminating the a.out code entirely at some point in the future.
1998-09-05 03:31:00 +00:00
/*
* Return the search path from the ldconfig hints file, reading it if
Import the DragonFly BSD commit 4f0bc915b65fcf5a23214f6d221d65c80be68ad4 by John Marino <draco@marino.st>, with the following (edited) commit message Date: Sat, 24 Mar 2012 06:40:50 +0100 Subject: [PATCH 1/1] rtld: Implement DT_RUNPATH and -z nodefaultlib DT_RUNPATH is incorrectly being considered as an alias of DT_RPATH. The purpose of DT_RUNPATH is to have two different types of rpath: one that can be overridden by the environment variable LD_LIBRARY_PATH and one that can't. With the currently implementation, LD_LIBRARY_PATH will always trump any embedded rpath or runpath tags. Current path search order by rtld: ================================== LD_LIBRARY_PATH DT_RPATH / DT_RUNPATH (always the same) ldconfig hints file (default: /var/run/ld-elf.so.hints) /usr/lib New path search order by rtld: ============================== DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file /usr/lib The new path search matches how the linux runtime loader works. The other major added feature is support for linker flag "-z nodefaultlib". When this flag is passed to the linker, rtld will skip all references to the standard library search path ("/usr/lib" in this case but it could handle more color delimited paths) except in DT_RPATH and DT_RUNPATH. New path search order by rtld with -z nodefaultlib flag set: ============================================================ DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file (skips all references to /usr/lib) FreeBSD notes: - we fixed some bugs which were submitted to DragonFly and merged there as commit 1ff8a2bd3eb6e5587174c6a983303ea3a79e0002; - we added LD_LIBRARY_PATH_RPATH environment variable to switch to the previous behaviour of considering DT_RPATH a synonym for DT_RUNPATH; - the FreeBSD default search path is /lib:/usr/lib and not /usr/lib. Reviewed by: kan MFC after: 1 month MFC note: flip the ld_library_path_rpath default value for stable/9
2012-07-15 10:53:48 +00:00
* necessary. If nostdlib is true, then the default search paths are
* not added to result.
*
* Returns NULL if there are problems with the hints file,
Implement ldconfig functionality for ELF. The hints are stored in a different file than the a.out hints, namely, "/var/run/ld-elf.so.hints". These hints consist only of the directory search path. There is no hash table as in the a.out hints, because ELF doesn't have to search for the file with the highest minor version number. (It doesn't have minor version numbers at all.) A single run of ldconfig updates either the a.out hints or the ELF hints, but not both. The set of hints to process is selected in the usual way, via /etc/objformat, or ${OBJFORMAT}, or the "-aout" or "-elf" command line option. The rationale is that you probably want to search different directories for ELF than for a.out. "ldconfig -r" is faked up to produce output like we are used to, except that for ELF there are no minor version numbers. This should enable "ldconfig -r" to be used for checking LIB_DEPENDS in ports even for ELF. I implemented the ELF functionality in a new source file, with an eye toward eliminating the a.out code entirely at some point in the future.
1998-09-05 03:31:00 +00:00
* or if the search path there is empty.
*/
static const char *
Import the DragonFly BSD commit 4f0bc915b65fcf5a23214f6d221d65c80be68ad4 by John Marino <draco@marino.st>, with the following (edited) commit message Date: Sat, 24 Mar 2012 06:40:50 +0100 Subject: [PATCH 1/1] rtld: Implement DT_RUNPATH and -z nodefaultlib DT_RUNPATH is incorrectly being considered as an alias of DT_RPATH. The purpose of DT_RUNPATH is to have two different types of rpath: one that can be overridden by the environment variable LD_LIBRARY_PATH and one that can't. With the currently implementation, LD_LIBRARY_PATH will always trump any embedded rpath or runpath tags. Current path search order by rtld: ================================== LD_LIBRARY_PATH DT_RPATH / DT_RUNPATH (always the same) ldconfig hints file (default: /var/run/ld-elf.so.hints) /usr/lib New path search order by rtld: ============================== DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file /usr/lib The new path search matches how the linux runtime loader works. The other major added feature is support for linker flag "-z nodefaultlib". When this flag is passed to the linker, rtld will skip all references to the standard library search path ("/usr/lib" in this case but it could handle more color delimited paths) except in DT_RPATH and DT_RUNPATH. New path search order by rtld with -z nodefaultlib flag set: ============================================================ DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file (skips all references to /usr/lib) FreeBSD notes: - we fixed some bugs which were submitted to DragonFly and merged there as commit 1ff8a2bd3eb6e5587174c6a983303ea3a79e0002; - we added LD_LIBRARY_PATH_RPATH environment variable to switch to the previous behaviour of considering DT_RPATH a synonym for DT_RUNPATH; - the FreeBSD default search path is /lib:/usr/lib and not /usr/lib. Reviewed by: kan MFC after: 1 month MFC note: flip the ld_library_path_rpath default value for stable/9
2012-07-15 10:53:48 +00:00
gethints(bool nostdlib)
Implement ldconfig functionality for ELF. The hints are stored in a different file than the a.out hints, namely, "/var/run/ld-elf.so.hints". These hints consist only of the directory search path. There is no hash table as in the a.out hints, because ELF doesn't have to search for the file with the highest minor version number. (It doesn't have minor version numbers at all.) A single run of ldconfig updates either the a.out hints or the ELF hints, but not both. The set of hints to process is selected in the usual way, via /etc/objformat, or ${OBJFORMAT}, or the "-aout" or "-elf" command line option. The rationale is that you probably want to search different directories for ELF than for a.out. "ldconfig -r" is faked up to produce output like we are used to, except that for ELF there are no minor version numbers. This should enable "ldconfig -r" to be used for checking LIB_DEPENDS in ports even for ELF. I implemented the ELF functionality in a new source file, with an eye toward eliminating the a.out code entirely at some point in the future.
1998-09-05 03:31:00 +00:00
{
Import the DragonFly BSD commit 4f0bc915b65fcf5a23214f6d221d65c80be68ad4 by John Marino <draco@marino.st>, with the following (edited) commit message Date: Sat, 24 Mar 2012 06:40:50 +0100 Subject: [PATCH 1/1] rtld: Implement DT_RUNPATH and -z nodefaultlib DT_RUNPATH is incorrectly being considered as an alias of DT_RPATH. The purpose of DT_RUNPATH is to have two different types of rpath: one that can be overridden by the environment variable LD_LIBRARY_PATH and one that can't. With the currently implementation, LD_LIBRARY_PATH will always trump any embedded rpath or runpath tags. Current path search order by rtld: ================================== LD_LIBRARY_PATH DT_RPATH / DT_RUNPATH (always the same) ldconfig hints file (default: /var/run/ld-elf.so.hints) /usr/lib New path search order by rtld: ============================== DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file /usr/lib The new path search matches how the linux runtime loader works. The other major added feature is support for linker flag "-z nodefaultlib". When this flag is passed to the linker, rtld will skip all references to the standard library search path ("/usr/lib" in this case but it could handle more color delimited paths) except in DT_RPATH and DT_RUNPATH. New path search order by rtld with -z nodefaultlib flag set: ============================================================ DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file (skips all references to /usr/lib) FreeBSD notes: - we fixed some bugs which were submitted to DragonFly and merged there as commit 1ff8a2bd3eb6e5587174c6a983303ea3a79e0002; - we added LD_LIBRARY_PATH_RPATH environment variable to switch to the previous behaviour of considering DT_RPATH a synonym for DT_RUNPATH; - the FreeBSD default search path is /lib:/usr/lib and not /usr/lib. Reviewed by: kan MFC after: 1 month MFC note: flip the ld_library_path_rpath default value for stable/9
2012-07-15 10:53:48 +00:00
static char *hints, *filtered_path;
Fix issues found by Coverity in the rtld-elf.c:gethints(). Check that the dirlist path string specification does not cause overflow and is fully contained in the hints file. Check that the dirlist string is nul-terminated. Make 'hdr' static variable, so that hdr.dirlistlen is available when hints cached value is used on next function calls. Reset hdr.dirlistlen to zero if error was detected, so that allocations use reasonable size. Use 'hints', and not 'p' in the body, since p is only initialized on the first call. Reported and reviewed by: truckman (previous version) Sponsored by: The FreeBSD Foundation CIDs: 1006503, 1006504, 1006676, 1008488, 1007263 MFC after: 2 weeks
2016-05-25 18:10:44 +00:00
static struct elfhints_hdr hdr;
Import the DragonFly BSD commit 4f0bc915b65fcf5a23214f6d221d65c80be68ad4 by John Marino <draco@marino.st>, with the following (edited) commit message Date: Sat, 24 Mar 2012 06:40:50 +0100 Subject: [PATCH 1/1] rtld: Implement DT_RUNPATH and -z nodefaultlib DT_RUNPATH is incorrectly being considered as an alias of DT_RPATH. The purpose of DT_RUNPATH is to have two different types of rpath: one that can be overridden by the environment variable LD_LIBRARY_PATH and one that can't. With the currently implementation, LD_LIBRARY_PATH will always trump any embedded rpath or runpath tags. Current path search order by rtld: ================================== LD_LIBRARY_PATH DT_RPATH / DT_RUNPATH (always the same) ldconfig hints file (default: /var/run/ld-elf.so.hints) /usr/lib New path search order by rtld: ============================== DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file /usr/lib The new path search matches how the linux runtime loader works. The other major added feature is support for linker flag "-z nodefaultlib". When this flag is passed to the linker, rtld will skip all references to the standard library search path ("/usr/lib" in this case but it could handle more color delimited paths) except in DT_RPATH and DT_RUNPATH. New path search order by rtld with -z nodefaultlib flag set: ============================================================ DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file (skips all references to /usr/lib) FreeBSD notes: - we fixed some bugs which were submitted to DragonFly and merged there as commit 1ff8a2bd3eb6e5587174c6a983303ea3a79e0002; - we added LD_LIBRARY_PATH_RPATH environment variable to switch to the previous behaviour of considering DT_RPATH a synonym for DT_RUNPATH; - the FreeBSD default search path is /lib:/usr/lib and not /usr/lib. Reviewed by: kan MFC after: 1 month MFC note: flip the ld_library_path_rpath default value for stable/9
2012-07-15 10:53:48 +00:00
struct fill_search_info_args sargs, hargs;
struct dl_serinfo smeta, hmeta, *SLPinfo, *hintinfo;
struct dl_serpath *SLPpath, *hintpath;
Implement ldconfig functionality for ELF. The hints are stored in a different file than the a.out hints, namely, "/var/run/ld-elf.so.hints". These hints consist only of the directory search path. There is no hash table as in the a.out hints, because ELF doesn't have to search for the file with the highest minor version number. (It doesn't have minor version numbers at all.) A single run of ldconfig updates either the a.out hints or the ELF hints, but not both. The set of hints to process is selected in the usual way, via /etc/objformat, or ${OBJFORMAT}, or the "-aout" or "-elf" command line option. The rationale is that you probably want to search different directories for ELF than for a.out. "ldconfig -r" is faked up to produce output like we are used to, except that for ELF there are no minor version numbers. This should enable "ldconfig -r" to be used for checking LIB_DEPENDS in ports even for ELF. I implemented the ELF functionality in a new source file, with an eye toward eliminating the a.out code entirely at some point in the future.
1998-09-05 03:31:00 +00:00
char *p;
Fix issues found by Coverity in the rtld-elf.c:gethints(). Check that the dirlist path string specification does not cause overflow and is fully contained in the hints file. Check that the dirlist string is nul-terminated. Make 'hdr' static variable, so that hdr.dirlistlen is available when hints cached value is used on next function calls. Reset hdr.dirlistlen to zero if error was detected, so that allocations use reasonable size. Use 'hints', and not 'p' in the body, since p is only initialized on the first call. Reported and reviewed by: truckman (previous version) Sponsored by: The FreeBSD Foundation CIDs: 1006503, 1006504, 1006676, 1008488, 1007263 MFC after: 2 weeks
2016-05-25 18:10:44 +00:00
struct stat hint_stat;
Import the DragonFly BSD commit 4f0bc915b65fcf5a23214f6d221d65c80be68ad4 by John Marino <draco@marino.st>, with the following (edited) commit message Date: Sat, 24 Mar 2012 06:40:50 +0100 Subject: [PATCH 1/1] rtld: Implement DT_RUNPATH and -z nodefaultlib DT_RUNPATH is incorrectly being considered as an alias of DT_RPATH. The purpose of DT_RUNPATH is to have two different types of rpath: one that can be overridden by the environment variable LD_LIBRARY_PATH and one that can't. With the currently implementation, LD_LIBRARY_PATH will always trump any embedded rpath or runpath tags. Current path search order by rtld: ================================== LD_LIBRARY_PATH DT_RPATH / DT_RUNPATH (always the same) ldconfig hints file (default: /var/run/ld-elf.so.hints) /usr/lib New path search order by rtld: ============================== DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file /usr/lib The new path search matches how the linux runtime loader works. The other major added feature is support for linker flag "-z nodefaultlib". When this flag is passed to the linker, rtld will skip all references to the standard library search path ("/usr/lib" in this case but it could handle more color delimited paths) except in DT_RPATH and DT_RUNPATH. New path search order by rtld with -z nodefaultlib flag set: ============================================================ DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file (skips all references to /usr/lib) FreeBSD notes: - we fixed some bugs which were submitted to DragonFly and merged there as commit 1ff8a2bd3eb6e5587174c6a983303ea3a79e0002; - we added LD_LIBRARY_PATH_RPATH environment variable to switch to the previous behaviour of considering DT_RPATH a synonym for DT_RUNPATH; - the FreeBSD default search path is /lib:/usr/lib and not /usr/lib. Reviewed by: kan MFC after: 1 month MFC note: flip the ld_library_path_rpath default value for stable/9
2012-07-15 10:53:48 +00:00
unsigned int SLPndx, hintndx, fndx, fcount;
int fd;
size_t flen;
Fix issues found by Coverity in the rtld-elf.c:gethints(). Check that the dirlist path string specification does not cause overflow and is fully contained in the hints file. Check that the dirlist string is nul-terminated. Make 'hdr' static variable, so that hdr.dirlistlen is available when hints cached value is used on next function calls. Reset hdr.dirlistlen to zero if error was detected, so that allocations use reasonable size. Use 'hints', and not 'p' in the body, since p is only initialized on the first call. Reported and reviewed by: truckman (previous version) Sponsored by: The FreeBSD Foundation CIDs: 1006503, 1006504, 1006676, 1008488, 1007263 MFC after: 2 weeks
2016-05-25 18:10:44 +00:00
uint32_t dl;
Import the DragonFly BSD commit 4f0bc915b65fcf5a23214f6d221d65c80be68ad4 by John Marino <draco@marino.st>, with the following (edited) commit message Date: Sat, 24 Mar 2012 06:40:50 +0100 Subject: [PATCH 1/1] rtld: Implement DT_RUNPATH and -z nodefaultlib DT_RUNPATH is incorrectly being considered as an alias of DT_RPATH. The purpose of DT_RUNPATH is to have two different types of rpath: one that can be overridden by the environment variable LD_LIBRARY_PATH and one that can't. With the currently implementation, LD_LIBRARY_PATH will always trump any embedded rpath or runpath tags. Current path search order by rtld: ================================== LD_LIBRARY_PATH DT_RPATH / DT_RUNPATH (always the same) ldconfig hints file (default: /var/run/ld-elf.so.hints) /usr/lib New path search order by rtld: ============================== DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file /usr/lib The new path search matches how the linux runtime loader works. The other major added feature is support for linker flag "-z nodefaultlib". When this flag is passed to the linker, rtld will skip all references to the standard library search path ("/usr/lib" in this case but it could handle more color delimited paths) except in DT_RPATH and DT_RUNPATH. New path search order by rtld with -z nodefaultlib flag set: ============================================================ DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file (skips all references to /usr/lib) FreeBSD notes: - we fixed some bugs which were submitted to DragonFly and merged there as commit 1ff8a2bd3eb6e5587174c6a983303ea3a79e0002; - we added LD_LIBRARY_PATH_RPATH environment variable to switch to the previous behaviour of considering DT_RPATH a synonym for DT_RUNPATH; - the FreeBSD default search path is /lib:/usr/lib and not /usr/lib. Reviewed by: kan MFC after: 1 month MFC note: flip the ld_library_path_rpath default value for stable/9
2012-07-15 10:53:48 +00:00
bool skip;
/* First call, read the hints file */
if (hints == NULL) {
/* Keep from trying again in case the hints file is bad. */
hints = "";
rtld: Fix fd leak with parallel dlopen and fork/exec. Rtld did not set FD_CLOEXEC on its internal file descriptors; therefore, such a file descriptor may be passed to a process created by another thread running in parallel to dlopen() or fdlopen(). No other threads are expected to be running during parsing of the hints and libmap files but the file descriptors need not be passed to child processes so add O_CLOEXEC there as well. This change will break fdlopen() (as used by OpenPAM) on kernels without F_DUPFD_CLOEXEC (added in July). Note that running new userland on old kernels is not supported. Reviewed by: kib
2012-11-04 21:42:24 +00:00
if ((fd = open(ld_elf_hints_path, O_RDONLY | O_CLOEXEC)) == -1)
Import the DragonFly BSD commit 4f0bc915b65fcf5a23214f6d221d65c80be68ad4 by John Marino <draco@marino.st>, with the following (edited) commit message Date: Sat, 24 Mar 2012 06:40:50 +0100 Subject: [PATCH 1/1] rtld: Implement DT_RUNPATH and -z nodefaultlib DT_RUNPATH is incorrectly being considered as an alias of DT_RPATH. The purpose of DT_RUNPATH is to have two different types of rpath: one that can be overridden by the environment variable LD_LIBRARY_PATH and one that can't. With the currently implementation, LD_LIBRARY_PATH will always trump any embedded rpath or runpath tags. Current path search order by rtld: ================================== LD_LIBRARY_PATH DT_RPATH / DT_RUNPATH (always the same) ldconfig hints file (default: /var/run/ld-elf.so.hints) /usr/lib New path search order by rtld: ============================== DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file /usr/lib The new path search matches how the linux runtime loader works. The other major added feature is support for linker flag "-z nodefaultlib". When this flag is passed to the linker, rtld will skip all references to the standard library search path ("/usr/lib" in this case but it could handle more color delimited paths) except in DT_RPATH and DT_RUNPATH. New path search order by rtld with -z nodefaultlib flag set: ============================================================ DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file (skips all references to /usr/lib) FreeBSD notes: - we fixed some bugs which were submitted to DragonFly and merged there as commit 1ff8a2bd3eb6e5587174c6a983303ea3a79e0002; - we added LD_LIBRARY_PATH_RPATH environment variable to switch to the previous behaviour of considering DT_RPATH a synonym for DT_RUNPATH; - the FreeBSD default search path is /lib:/usr/lib and not /usr/lib. Reviewed by: kan MFC after: 1 month MFC note: flip the ld_library_path_rpath default value for stable/9
2012-07-15 10:53:48 +00:00
return (NULL);
Fix issues found by Coverity in the rtld-elf.c:gethints(). Check that the dirlist path string specification does not cause overflow and is fully contained in the hints file. Check that the dirlist string is nul-terminated. Make 'hdr' static variable, so that hdr.dirlistlen is available when hints cached value is used on next function calls. Reset hdr.dirlistlen to zero if error was detected, so that allocations use reasonable size. Use 'hints', and not 'p' in the body, since p is only initialized on the first call. Reported and reviewed by: truckman (previous version) Sponsored by: The FreeBSD Foundation CIDs: 1006503, 1006504, 1006676, 1008488, 1007263 MFC after: 2 weeks
2016-05-25 18:10:44 +00:00
/*
* Check of hdr.dirlistlen value against type limit
* intends to pacify static analyzers. Further
* paranoia leads to checks that dirlist is fully
* contained in the file range.
*/
Import the DragonFly BSD commit 4f0bc915b65fcf5a23214f6d221d65c80be68ad4 by John Marino <draco@marino.st>, with the following (edited) commit message Date: Sat, 24 Mar 2012 06:40:50 +0100 Subject: [PATCH 1/1] rtld: Implement DT_RUNPATH and -z nodefaultlib DT_RUNPATH is incorrectly being considered as an alias of DT_RPATH. The purpose of DT_RUNPATH is to have two different types of rpath: one that can be overridden by the environment variable LD_LIBRARY_PATH and one that can't. With the currently implementation, LD_LIBRARY_PATH will always trump any embedded rpath or runpath tags. Current path search order by rtld: ================================== LD_LIBRARY_PATH DT_RPATH / DT_RUNPATH (always the same) ldconfig hints file (default: /var/run/ld-elf.so.hints) /usr/lib New path search order by rtld: ============================== DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file /usr/lib The new path search matches how the linux runtime loader works. The other major added feature is support for linker flag "-z nodefaultlib". When this flag is passed to the linker, rtld will skip all references to the standard library search path ("/usr/lib" in this case but it could handle more color delimited paths) except in DT_RPATH and DT_RUNPATH. New path search order by rtld with -z nodefaultlib flag set: ============================================================ DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file (skips all references to /usr/lib) FreeBSD notes: - we fixed some bugs which were submitted to DragonFly and merged there as commit 1ff8a2bd3eb6e5587174c6a983303ea3a79e0002; - we added LD_LIBRARY_PATH_RPATH environment variable to switch to the previous behaviour of considering DT_RPATH a synonym for DT_RUNPATH; - the FreeBSD default search path is /lib:/usr/lib and not /usr/lib. Reviewed by: kan MFC after: 1 month MFC note: flip the ld_library_path_rpath default value for stable/9
2012-07-15 10:53:48 +00:00
if (read(fd, &hdr, sizeof hdr) != sizeof hdr ||
hdr.magic != ELFHINTS_MAGIC ||
Fix issues found by Coverity in the rtld-elf.c:gethints(). Check that the dirlist path string specification does not cause overflow and is fully contained in the hints file. Check that the dirlist string is nul-terminated. Make 'hdr' static variable, so that hdr.dirlistlen is available when hints cached value is used on next function calls. Reset hdr.dirlistlen to zero if error was detected, so that allocations use reasonable size. Use 'hints', and not 'p' in the body, since p is only initialized on the first call. Reported and reviewed by: truckman (previous version) Sponsored by: The FreeBSD Foundation CIDs: 1006503, 1006504, 1006676, 1008488, 1007263 MFC after: 2 weeks
2016-05-25 18:10:44 +00:00
hdr.version != 1 || hdr.dirlistlen > UINT_MAX / 2 ||
fstat(fd, &hint_stat) == -1) {
cleanup1:
Import the DragonFly BSD commit 4f0bc915b65fcf5a23214f6d221d65c80be68ad4 by John Marino <draco@marino.st>, with the following (edited) commit message Date: Sat, 24 Mar 2012 06:40:50 +0100 Subject: [PATCH 1/1] rtld: Implement DT_RUNPATH and -z nodefaultlib DT_RUNPATH is incorrectly being considered as an alias of DT_RPATH. The purpose of DT_RUNPATH is to have two different types of rpath: one that can be overridden by the environment variable LD_LIBRARY_PATH and one that can't. With the currently implementation, LD_LIBRARY_PATH will always trump any embedded rpath or runpath tags. Current path search order by rtld: ================================== LD_LIBRARY_PATH DT_RPATH / DT_RUNPATH (always the same) ldconfig hints file (default: /var/run/ld-elf.so.hints) /usr/lib New path search order by rtld: ============================== DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file /usr/lib The new path search matches how the linux runtime loader works. The other major added feature is support for linker flag "-z nodefaultlib". When this flag is passed to the linker, rtld will skip all references to the standard library search path ("/usr/lib" in this case but it could handle more color delimited paths) except in DT_RPATH and DT_RUNPATH. New path search order by rtld with -z nodefaultlib flag set: ============================================================ DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file (skips all references to /usr/lib) FreeBSD notes: - we fixed some bugs which were submitted to DragonFly and merged there as commit 1ff8a2bd3eb6e5587174c6a983303ea3a79e0002; - we added LD_LIBRARY_PATH_RPATH environment variable to switch to the previous behaviour of considering DT_RPATH a synonym for DT_RUNPATH; - the FreeBSD default search path is /lib:/usr/lib and not /usr/lib. Reviewed by: kan MFC after: 1 month MFC note: flip the ld_library_path_rpath default value for stable/9
2012-07-15 10:53:48 +00:00
close(fd);
Fix issues found by Coverity in the rtld-elf.c:gethints(). Check that the dirlist path string specification does not cause overflow and is fully contained in the hints file. Check that the dirlist string is nul-terminated. Make 'hdr' static variable, so that hdr.dirlistlen is available when hints cached value is used on next function calls. Reset hdr.dirlistlen to zero if error was detected, so that allocations use reasonable size. Use 'hints', and not 'p' in the body, since p is only initialized on the first call. Reported and reviewed by: truckman (previous version) Sponsored by: The FreeBSD Foundation CIDs: 1006503, 1006504, 1006676, 1008488, 1007263 MFC after: 2 weeks
2016-05-25 18:10:44 +00:00
hdr.dirlistlen = 0;
Import the DragonFly BSD commit 4f0bc915b65fcf5a23214f6d221d65c80be68ad4 by John Marino <draco@marino.st>, with the following (edited) commit message Date: Sat, 24 Mar 2012 06:40:50 +0100 Subject: [PATCH 1/1] rtld: Implement DT_RUNPATH and -z nodefaultlib DT_RUNPATH is incorrectly being considered as an alias of DT_RPATH. The purpose of DT_RUNPATH is to have two different types of rpath: one that can be overridden by the environment variable LD_LIBRARY_PATH and one that can't. With the currently implementation, LD_LIBRARY_PATH will always trump any embedded rpath or runpath tags. Current path search order by rtld: ================================== LD_LIBRARY_PATH DT_RPATH / DT_RUNPATH (always the same) ldconfig hints file (default: /var/run/ld-elf.so.hints) /usr/lib New path search order by rtld: ============================== DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file /usr/lib The new path search matches how the linux runtime loader works. The other major added feature is support for linker flag "-z nodefaultlib". When this flag is passed to the linker, rtld will skip all references to the standard library search path ("/usr/lib" in this case but it could handle more color delimited paths) except in DT_RPATH and DT_RUNPATH. New path search order by rtld with -z nodefaultlib flag set: ============================================================ DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file (skips all references to /usr/lib) FreeBSD notes: - we fixed some bugs which were submitted to DragonFly and merged there as commit 1ff8a2bd3eb6e5587174c6a983303ea3a79e0002; - we added LD_LIBRARY_PATH_RPATH environment variable to switch to the previous behaviour of considering DT_RPATH a synonym for DT_RUNPATH; - the FreeBSD default search path is /lib:/usr/lib and not /usr/lib. Reviewed by: kan MFC after: 1 month MFC note: flip the ld_library_path_rpath default value for stable/9
2012-07-15 10:53:48 +00:00
return (NULL);
}
Fix issues found by Coverity in the rtld-elf.c:gethints(). Check that the dirlist path string specification does not cause overflow and is fully contained in the hints file. Check that the dirlist string is nul-terminated. Make 'hdr' static variable, so that hdr.dirlistlen is available when hints cached value is used on next function calls. Reset hdr.dirlistlen to zero if error was detected, so that allocations use reasonable size. Use 'hints', and not 'p' in the body, since p is only initialized on the first call. Reported and reviewed by: truckman (previous version) Sponsored by: The FreeBSD Foundation CIDs: 1006503, 1006504, 1006676, 1008488, 1007263 MFC after: 2 weeks
2016-05-25 18:10:44 +00:00
dl = hdr.strtab;
if (dl + hdr.dirlist < dl)
goto cleanup1;
dl += hdr.dirlist;
if (dl + hdr.dirlistlen < dl)
goto cleanup1;
dl += hdr.dirlistlen;
if (dl > hint_stat.st_size)
goto cleanup1;
Import the DragonFly BSD commit 4f0bc915b65fcf5a23214f6d221d65c80be68ad4 by John Marino <draco@marino.st>, with the following (edited) commit message Date: Sat, 24 Mar 2012 06:40:50 +0100 Subject: [PATCH 1/1] rtld: Implement DT_RUNPATH and -z nodefaultlib DT_RUNPATH is incorrectly being considered as an alias of DT_RPATH. The purpose of DT_RUNPATH is to have two different types of rpath: one that can be overridden by the environment variable LD_LIBRARY_PATH and one that can't. With the currently implementation, LD_LIBRARY_PATH will always trump any embedded rpath or runpath tags. Current path search order by rtld: ================================== LD_LIBRARY_PATH DT_RPATH / DT_RUNPATH (always the same) ldconfig hints file (default: /var/run/ld-elf.so.hints) /usr/lib New path search order by rtld: ============================== DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file /usr/lib The new path search matches how the linux runtime loader works. The other major added feature is support for linker flag "-z nodefaultlib". When this flag is passed to the linker, rtld will skip all references to the standard library search path ("/usr/lib" in this case but it could handle more color delimited paths) except in DT_RPATH and DT_RUNPATH. New path search order by rtld with -z nodefaultlib flag set: ============================================================ DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file (skips all references to /usr/lib) FreeBSD notes: - we fixed some bugs which were submitted to DragonFly and merged there as commit 1ff8a2bd3eb6e5587174c6a983303ea3a79e0002; - we added LD_LIBRARY_PATH_RPATH environment variable to switch to the previous behaviour of considering DT_RPATH a synonym for DT_RUNPATH; - the FreeBSD default search path is /lib:/usr/lib and not /usr/lib. Reviewed by: kan MFC after: 1 month MFC note: flip the ld_library_path_rpath default value for stable/9
2012-07-15 10:53:48 +00:00
p = xmalloc(hdr.dirlistlen + 1);
Fix issues found by Coverity in the rtld-elf.c:gethints(). Check that the dirlist path string specification does not cause overflow and is fully contained in the hints file. Check that the dirlist string is nul-terminated. Make 'hdr' static variable, so that hdr.dirlistlen is available when hints cached value is used on next function calls. Reset hdr.dirlistlen to zero if error was detected, so that allocations use reasonable size. Use 'hints', and not 'p' in the body, since p is only initialized on the first call. Reported and reviewed by: truckman (previous version) Sponsored by: The FreeBSD Foundation CIDs: 1006503, 1006504, 1006676, 1008488, 1007263 MFC after: 2 weeks
2016-05-25 18:10:44 +00:00
Import the DragonFly BSD commit 4f0bc915b65fcf5a23214f6d221d65c80be68ad4 by John Marino <draco@marino.st>, with the following (edited) commit message Date: Sat, 24 Mar 2012 06:40:50 +0100 Subject: [PATCH 1/1] rtld: Implement DT_RUNPATH and -z nodefaultlib DT_RUNPATH is incorrectly being considered as an alias of DT_RPATH. The purpose of DT_RUNPATH is to have two different types of rpath: one that can be overridden by the environment variable LD_LIBRARY_PATH and one that can't. With the currently implementation, LD_LIBRARY_PATH will always trump any embedded rpath or runpath tags. Current path search order by rtld: ================================== LD_LIBRARY_PATH DT_RPATH / DT_RUNPATH (always the same) ldconfig hints file (default: /var/run/ld-elf.so.hints) /usr/lib New path search order by rtld: ============================== DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file /usr/lib The new path search matches how the linux runtime loader works. The other major added feature is support for linker flag "-z nodefaultlib". When this flag is passed to the linker, rtld will skip all references to the standard library search path ("/usr/lib" in this case but it could handle more color delimited paths) except in DT_RPATH and DT_RUNPATH. New path search order by rtld with -z nodefaultlib flag set: ============================================================ DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file (skips all references to /usr/lib) FreeBSD notes: - we fixed some bugs which were submitted to DragonFly and merged there as commit 1ff8a2bd3eb6e5587174c6a983303ea3a79e0002; - we added LD_LIBRARY_PATH_RPATH environment variable to switch to the previous behaviour of considering DT_RPATH a synonym for DT_RUNPATH; - the FreeBSD default search path is /lib:/usr/lib and not /usr/lib. Reviewed by: kan MFC after: 1 month MFC note: flip the ld_library_path_rpath default value for stable/9
2012-07-15 10:53:48 +00:00
if (lseek(fd, hdr.strtab + hdr.dirlist, SEEK_SET) == -1 ||
read(fd, p, hdr.dirlistlen + 1) !=
Fix issues found by Coverity in the rtld-elf.c:gethints(). Check that the dirlist path string specification does not cause overflow and is fully contained in the hints file. Check that the dirlist string is nul-terminated. Make 'hdr' static variable, so that hdr.dirlistlen is available when hints cached value is used on next function calls. Reset hdr.dirlistlen to zero if error was detected, so that allocations use reasonable size. Use 'hints', and not 'p' in the body, since p is only initialized on the first call. Reported and reviewed by: truckman (previous version) Sponsored by: The FreeBSD Foundation CIDs: 1006503, 1006504, 1006676, 1008488, 1007263 MFC after: 2 weeks
2016-05-25 18:10:44 +00:00
(ssize_t)hdr.dirlistlen + 1 || p[hdr.dirlistlen] != '\0') {
Import the DragonFly BSD commit 4f0bc915b65fcf5a23214f6d221d65c80be68ad4 by John Marino <draco@marino.st>, with the following (edited) commit message Date: Sat, 24 Mar 2012 06:40:50 +0100 Subject: [PATCH 1/1] rtld: Implement DT_RUNPATH and -z nodefaultlib DT_RUNPATH is incorrectly being considered as an alias of DT_RPATH. The purpose of DT_RUNPATH is to have two different types of rpath: one that can be overridden by the environment variable LD_LIBRARY_PATH and one that can't. With the currently implementation, LD_LIBRARY_PATH will always trump any embedded rpath or runpath tags. Current path search order by rtld: ================================== LD_LIBRARY_PATH DT_RPATH / DT_RUNPATH (always the same) ldconfig hints file (default: /var/run/ld-elf.so.hints) /usr/lib New path search order by rtld: ============================== DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file /usr/lib The new path search matches how the linux runtime loader works. The other major added feature is support for linker flag "-z nodefaultlib". When this flag is passed to the linker, rtld will skip all references to the standard library search path ("/usr/lib" in this case but it could handle more color delimited paths) except in DT_RPATH and DT_RUNPATH. New path search order by rtld with -z nodefaultlib flag set: ============================================================ DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file (skips all references to /usr/lib) FreeBSD notes: - we fixed some bugs which were submitted to DragonFly and merged there as commit 1ff8a2bd3eb6e5587174c6a983303ea3a79e0002; - we added LD_LIBRARY_PATH_RPATH environment variable to switch to the previous behaviour of considering DT_RPATH a synonym for DT_RUNPATH; - the FreeBSD default search path is /lib:/usr/lib and not /usr/lib. Reviewed by: kan MFC after: 1 month MFC note: flip the ld_library_path_rpath default value for stable/9
2012-07-15 10:53:48 +00:00
free(p);
Fix issues found by Coverity in the rtld-elf.c:gethints(). Check that the dirlist path string specification does not cause overflow and is fully contained in the hints file. Check that the dirlist string is nul-terminated. Make 'hdr' static variable, so that hdr.dirlistlen is available when hints cached value is used on next function calls. Reset hdr.dirlistlen to zero if error was detected, so that allocations use reasonable size. Use 'hints', and not 'p' in the body, since p is only initialized on the first call. Reported and reviewed by: truckman (previous version) Sponsored by: The FreeBSD Foundation CIDs: 1006503, 1006504, 1006676, 1008488, 1007263 MFC after: 2 weeks
2016-05-25 18:10:44 +00:00
goto cleanup1;
Import the DragonFly BSD commit 4f0bc915b65fcf5a23214f6d221d65c80be68ad4 by John Marino <draco@marino.st>, with the following (edited) commit message Date: Sat, 24 Mar 2012 06:40:50 +0100 Subject: [PATCH 1/1] rtld: Implement DT_RUNPATH and -z nodefaultlib DT_RUNPATH is incorrectly being considered as an alias of DT_RPATH. The purpose of DT_RUNPATH is to have two different types of rpath: one that can be overridden by the environment variable LD_LIBRARY_PATH and one that can't. With the currently implementation, LD_LIBRARY_PATH will always trump any embedded rpath or runpath tags. Current path search order by rtld: ================================== LD_LIBRARY_PATH DT_RPATH / DT_RUNPATH (always the same) ldconfig hints file (default: /var/run/ld-elf.so.hints) /usr/lib New path search order by rtld: ============================== DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file /usr/lib The new path search matches how the linux runtime loader works. The other major added feature is support for linker flag "-z nodefaultlib". When this flag is passed to the linker, rtld will skip all references to the standard library search path ("/usr/lib" in this case but it could handle more color delimited paths) except in DT_RPATH and DT_RUNPATH. New path search order by rtld with -z nodefaultlib flag set: ============================================================ DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file (skips all references to /usr/lib) FreeBSD notes: - we fixed some bugs which were submitted to DragonFly and merged there as commit 1ff8a2bd3eb6e5587174c6a983303ea3a79e0002; - we added LD_LIBRARY_PATH_RPATH environment variable to switch to the previous behaviour of considering DT_RPATH a synonym for DT_RUNPATH; - the FreeBSD default search path is /lib:/usr/lib and not /usr/lib. Reviewed by: kan MFC after: 1 month MFC note: flip the ld_library_path_rpath default value for stable/9
2012-07-15 10:53:48 +00:00
}
hints = p;
close(fd);
}
Implement ldconfig functionality for ELF. The hints are stored in a different file than the a.out hints, namely, "/var/run/ld-elf.so.hints". These hints consist only of the directory search path. There is no hash table as in the a.out hints, because ELF doesn't have to search for the file with the highest minor version number. (It doesn't have minor version numbers at all.) A single run of ldconfig updates either the a.out hints or the ELF hints, but not both. The set of hints to process is selected in the usual way, via /etc/objformat, or ${OBJFORMAT}, or the "-aout" or "-elf" command line option. The rationale is that you probably want to search different directories for ELF than for a.out. "ldconfig -r" is faked up to produce output like we are used to, except that for ELF there are no minor version numbers. This should enable "ldconfig -r" to be used for checking LIB_DEPENDS in ports even for ELF. I implemented the ELF functionality in a new source file, with an eye toward eliminating the a.out code entirely at some point in the future.
1998-09-05 03:31:00 +00:00
Import the DragonFly BSD commit 4f0bc915b65fcf5a23214f6d221d65c80be68ad4 by John Marino <draco@marino.st>, with the following (edited) commit message Date: Sat, 24 Mar 2012 06:40:50 +0100 Subject: [PATCH 1/1] rtld: Implement DT_RUNPATH and -z nodefaultlib DT_RUNPATH is incorrectly being considered as an alias of DT_RPATH. The purpose of DT_RUNPATH is to have two different types of rpath: one that can be overridden by the environment variable LD_LIBRARY_PATH and one that can't. With the currently implementation, LD_LIBRARY_PATH will always trump any embedded rpath or runpath tags. Current path search order by rtld: ================================== LD_LIBRARY_PATH DT_RPATH / DT_RUNPATH (always the same) ldconfig hints file (default: /var/run/ld-elf.so.hints) /usr/lib New path search order by rtld: ============================== DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file /usr/lib The new path search matches how the linux runtime loader works. The other major added feature is support for linker flag "-z nodefaultlib". When this flag is passed to the linker, rtld will skip all references to the standard library search path ("/usr/lib" in this case but it could handle more color delimited paths) except in DT_RPATH and DT_RUNPATH. New path search order by rtld with -z nodefaultlib flag set: ============================================================ DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file (skips all references to /usr/lib) FreeBSD notes: - we fixed some bugs which were submitted to DragonFly and merged there as commit 1ff8a2bd3eb6e5587174c6a983303ea3a79e0002; - we added LD_LIBRARY_PATH_RPATH environment variable to switch to the previous behaviour of considering DT_RPATH a synonym for DT_RUNPATH; - the FreeBSD default search path is /lib:/usr/lib and not /usr/lib. Reviewed by: kan MFC after: 1 month MFC note: flip the ld_library_path_rpath default value for stable/9
2012-07-15 10:53:48 +00:00
/*
* If caller agreed to receive list which includes the default
* paths, we are done. Otherwise, if we still did not
* calculated filtered result, do it now.
*/
if (!nostdlib)
return (hints[0] != '\0' ? hints : NULL);
if (filtered_path != NULL)
goto filt_ret;
Implement ldconfig functionality for ELF. The hints are stored in a different file than the a.out hints, namely, "/var/run/ld-elf.so.hints". These hints consist only of the directory search path. There is no hash table as in the a.out hints, because ELF doesn't have to search for the file with the highest minor version number. (It doesn't have minor version numbers at all.) A single run of ldconfig updates either the a.out hints or the ELF hints, but not both. The set of hints to process is selected in the usual way, via /etc/objformat, or ${OBJFORMAT}, or the "-aout" or "-elf" command line option. The rationale is that you probably want to search different directories for ELF than for a.out. "ldconfig -r" is faked up to produce output like we are used to, except that for ELF there are no minor version numbers. This should enable "ldconfig -r" to be used for checking LIB_DEPENDS in ports even for ELF. I implemented the ELF functionality in a new source file, with an eye toward eliminating the a.out code entirely at some point in the future.
1998-09-05 03:31:00 +00:00
Import the DragonFly BSD commit 4f0bc915b65fcf5a23214f6d221d65c80be68ad4 by John Marino <draco@marino.st>, with the following (edited) commit message Date: Sat, 24 Mar 2012 06:40:50 +0100 Subject: [PATCH 1/1] rtld: Implement DT_RUNPATH and -z nodefaultlib DT_RUNPATH is incorrectly being considered as an alias of DT_RPATH. The purpose of DT_RUNPATH is to have two different types of rpath: one that can be overridden by the environment variable LD_LIBRARY_PATH and one that can't. With the currently implementation, LD_LIBRARY_PATH will always trump any embedded rpath or runpath tags. Current path search order by rtld: ================================== LD_LIBRARY_PATH DT_RPATH / DT_RUNPATH (always the same) ldconfig hints file (default: /var/run/ld-elf.so.hints) /usr/lib New path search order by rtld: ============================== DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file /usr/lib The new path search matches how the linux runtime loader works. The other major added feature is support for linker flag "-z nodefaultlib". When this flag is passed to the linker, rtld will skip all references to the standard library search path ("/usr/lib" in this case but it could handle more color delimited paths) except in DT_RPATH and DT_RUNPATH. New path search order by rtld with -z nodefaultlib flag set: ============================================================ DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file (skips all references to /usr/lib) FreeBSD notes: - we fixed some bugs which were submitted to DragonFly and merged there as commit 1ff8a2bd3eb6e5587174c6a983303ea3a79e0002; - we added LD_LIBRARY_PATH_RPATH environment variable to switch to the previous behaviour of considering DT_RPATH a synonym for DT_RUNPATH; - the FreeBSD default search path is /lib:/usr/lib and not /usr/lib. Reviewed by: kan MFC after: 1 month MFC note: flip the ld_library_path_rpath default value for stable/9
2012-07-15 10:53:48 +00:00
/*
* Obtain the list of all configured search paths, and the
* list of the default paths.
*
* First estimate the size of the results.
*/
smeta.dls_size = __offsetof(struct dl_serinfo, dls_serpath);
smeta.dls_cnt = 0;
hmeta.dls_size = __offsetof(struct dl_serinfo, dls_serpath);
hmeta.dls_cnt = 0;
sargs.request = RTLD_DI_SERINFOSIZE;
sargs.serinfo = &smeta;
hargs.request = RTLD_DI_SERINFOSIZE;
hargs.serinfo = &hmeta;
Rather than using the #define for path names, indirect through a char * variable that could change for different executable types detected.
2015-10-31 04:39:55 +00:00
path_enumerate(ld_standard_library_path, fill_search_info, &sargs);
Fix issues found by Coverity in the rtld-elf.c:gethints(). Check that the dirlist path string specification does not cause overflow and is fully contained in the hints file. Check that the dirlist string is nul-terminated. Make 'hdr' static variable, so that hdr.dirlistlen is available when hints cached value is used on next function calls. Reset hdr.dirlistlen to zero if error was detected, so that allocations use reasonable size. Use 'hints', and not 'p' in the body, since p is only initialized on the first call. Reported and reviewed by: truckman (previous version) Sponsored by: The FreeBSD Foundation CIDs: 1006503, 1006504, 1006676, 1008488, 1007263 MFC after: 2 weeks
2016-05-25 18:10:44 +00:00
path_enumerate(hints, fill_search_info, &hargs);
Import the DragonFly BSD commit 4f0bc915b65fcf5a23214f6d221d65c80be68ad4 by John Marino <draco@marino.st>, with the following (edited) commit message Date: Sat, 24 Mar 2012 06:40:50 +0100 Subject: [PATCH 1/1] rtld: Implement DT_RUNPATH and -z nodefaultlib DT_RUNPATH is incorrectly being considered as an alias of DT_RPATH. The purpose of DT_RUNPATH is to have two different types of rpath: one that can be overridden by the environment variable LD_LIBRARY_PATH and one that can't. With the currently implementation, LD_LIBRARY_PATH will always trump any embedded rpath or runpath tags. Current path search order by rtld: ================================== LD_LIBRARY_PATH DT_RPATH / DT_RUNPATH (always the same) ldconfig hints file (default: /var/run/ld-elf.so.hints) /usr/lib New path search order by rtld: ============================== DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file /usr/lib The new path search matches how the linux runtime loader works. The other major added feature is support for linker flag "-z nodefaultlib". When this flag is passed to the linker, rtld will skip all references to the standard library search path ("/usr/lib" in this case but it could handle more color delimited paths) except in DT_RPATH and DT_RUNPATH. New path search order by rtld with -z nodefaultlib flag set: ============================================================ DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file (skips all references to /usr/lib) FreeBSD notes: - we fixed some bugs which were submitted to DragonFly and merged there as commit 1ff8a2bd3eb6e5587174c6a983303ea3a79e0002; - we added LD_LIBRARY_PATH_RPATH environment variable to switch to the previous behaviour of considering DT_RPATH a synonym for DT_RUNPATH; - the FreeBSD default search path is /lib:/usr/lib and not /usr/lib. Reviewed by: kan MFC after: 1 month MFC note: flip the ld_library_path_rpath default value for stable/9
2012-07-15 10:53:48 +00:00
SLPinfo = xmalloc(smeta.dls_size);
hintinfo = xmalloc(hmeta.dls_size);
/*
* Next fetch both sets of paths.
*/
sargs.request = RTLD_DI_SERINFO;
sargs.serinfo = SLPinfo;
sargs.serpath = &SLPinfo->dls_serpath[0];
sargs.strspace = (char *)&SLPinfo->dls_serpath[smeta.dls_cnt];
hargs.request = RTLD_DI_SERINFO;
hargs.serinfo = hintinfo;
hargs.serpath = &hintinfo->dls_serpath[0];
hargs.strspace = (char *)&hintinfo->dls_serpath[hmeta.dls_cnt];
Rather than using the #define for path names, indirect through a char * variable that could change for different executable types detected.
2015-10-31 04:39:55 +00:00
path_enumerate(ld_standard_library_path, fill_search_info, &sargs);
Fix issues found by Coverity in the rtld-elf.c:gethints(). Check that the dirlist path string specification does not cause overflow and is fully contained in the hints file. Check that the dirlist string is nul-terminated. Make 'hdr' static variable, so that hdr.dirlistlen is available when hints cached value is used on next function calls. Reset hdr.dirlistlen to zero if error was detected, so that allocations use reasonable size. Use 'hints', and not 'p' in the body, since p is only initialized on the first call. Reported and reviewed by: truckman (previous version) Sponsored by: The FreeBSD Foundation CIDs: 1006503, 1006504, 1006676, 1008488, 1007263 MFC after: 2 weeks
2016-05-25 18:10:44 +00:00
path_enumerate(hints, fill_search_info, &hargs);
Import the DragonFly BSD commit 4f0bc915b65fcf5a23214f6d221d65c80be68ad4 by John Marino <draco@marino.st>, with the following (edited) commit message Date: Sat, 24 Mar 2012 06:40:50 +0100 Subject: [PATCH 1/1] rtld: Implement DT_RUNPATH and -z nodefaultlib DT_RUNPATH is incorrectly being considered as an alias of DT_RPATH. The purpose of DT_RUNPATH is to have two different types of rpath: one that can be overridden by the environment variable LD_LIBRARY_PATH and one that can't. With the currently implementation, LD_LIBRARY_PATH will always trump any embedded rpath or runpath tags. Current path search order by rtld: ================================== LD_LIBRARY_PATH DT_RPATH / DT_RUNPATH (always the same) ldconfig hints file (default: /var/run/ld-elf.so.hints) /usr/lib New path search order by rtld: ============================== DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file /usr/lib The new path search matches how the linux runtime loader works. The other major added feature is support for linker flag "-z nodefaultlib". When this flag is passed to the linker, rtld will skip all references to the standard library search path ("/usr/lib" in this case but it could handle more color delimited paths) except in DT_RPATH and DT_RUNPATH. New path search order by rtld with -z nodefaultlib flag set: ============================================================ DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file (skips all references to /usr/lib) FreeBSD notes: - we fixed some bugs which were submitted to DragonFly and merged there as commit 1ff8a2bd3eb6e5587174c6a983303ea3a79e0002; - we added LD_LIBRARY_PATH_RPATH environment variable to switch to the previous behaviour of considering DT_RPATH a synonym for DT_RUNPATH; - the FreeBSD default search path is /lib:/usr/lib and not /usr/lib. Reviewed by: kan MFC after: 1 month MFC note: flip the ld_library_path_rpath default value for stable/9
2012-07-15 10:53:48 +00:00
/*
* Now calculate the difference between two sets, by excluding
* standard paths from the full set.
*/
fndx = 0;
fcount = 0;
filtered_path = xmalloc(hdr.dirlistlen + 1);
hintpath = &hintinfo->dls_serpath[0];
for (hintndx = 0; hintndx < hmeta.dls_cnt; hintndx++, hintpath++) {
skip = false;
SLPpath = &SLPinfo->dls_serpath[0];
/*
* Check each standard path against current.
*/
for (SLPndx = 0; SLPndx < smeta.dls_cnt; SLPndx++, SLPpath++) {
/* matched, skip the path */
if (!strcmp(hintpath->dls_name, SLPpath->dls_name)) {
skip = true;
break;
}
}
if (skip)
continue;
/*
* Not matched against any standard path, add the path
* to result. Separate consequtive paths with ':'.
*/
if (fcount > 0) {
filtered_path[fndx] = ':';
fndx++;
}
fcount++;
flen = strlen(hintpath->dls_name);
strncpy((filtered_path + fndx), hintpath->dls_name, flen);
fndx += flen;
Implement ldconfig functionality for ELF. The hints are stored in a different file than the a.out hints, namely, "/var/run/ld-elf.so.hints". These hints consist only of the directory search path. There is no hash table as in the a.out hints, because ELF doesn't have to search for the file with the highest minor version number. (It doesn't have minor version numbers at all.) A single run of ldconfig updates either the a.out hints or the ELF hints, but not both. The set of hints to process is selected in the usual way, via /etc/objformat, or ${OBJFORMAT}, or the "-aout" or "-elf" command line option. The rationale is that you probably want to search different directories for ELF than for a.out. "ldconfig -r" is faked up to produce output like we are used to, except that for ELF there are no minor version numbers. This should enable "ldconfig -r" to be used for checking LIB_DEPENDS in ports even for ELF. I implemented the ELF functionality in a new source file, with an eye toward eliminating the a.out code entirely at some point in the future.
1998-09-05 03:31:00 +00:00
}
Import the DragonFly BSD commit 4f0bc915b65fcf5a23214f6d221d65c80be68ad4 by John Marino <draco@marino.st>, with the following (edited) commit message Date: Sat, 24 Mar 2012 06:40:50 +0100 Subject: [PATCH 1/1] rtld: Implement DT_RUNPATH and -z nodefaultlib DT_RUNPATH is incorrectly being considered as an alias of DT_RPATH. The purpose of DT_RUNPATH is to have two different types of rpath: one that can be overridden by the environment variable LD_LIBRARY_PATH and one that can't. With the currently implementation, LD_LIBRARY_PATH will always trump any embedded rpath or runpath tags. Current path search order by rtld: ================================== LD_LIBRARY_PATH DT_RPATH / DT_RUNPATH (always the same) ldconfig hints file (default: /var/run/ld-elf.so.hints) /usr/lib New path search order by rtld: ============================== DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file /usr/lib The new path search matches how the linux runtime loader works. The other major added feature is support for linker flag "-z nodefaultlib". When this flag is passed to the linker, rtld will skip all references to the standard library search path ("/usr/lib" in this case but it could handle more color delimited paths) except in DT_RPATH and DT_RUNPATH. New path search order by rtld with -z nodefaultlib flag set: ============================================================ DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file (skips all references to /usr/lib) FreeBSD notes: - we fixed some bugs which were submitted to DragonFly and merged there as commit 1ff8a2bd3eb6e5587174c6a983303ea3a79e0002; - we added LD_LIBRARY_PATH_RPATH environment variable to switch to the previous behaviour of considering DT_RPATH a synonym for DT_RUNPATH; - the FreeBSD default search path is /lib:/usr/lib and not /usr/lib. Reviewed by: kan MFC after: 1 month MFC note: flip the ld_library_path_rpath default value for stable/9
2012-07-15 10:53:48 +00:00
filtered_path[fndx] = '\0';
free(SLPinfo);
free(hintinfo);
filt_ret:
return (filtered_path[0] != '\0' ? filtered_path : NULL);
Implement ldconfig functionality for ELF. The hints are stored in a different file than the a.out hints, namely, "/var/run/ld-elf.so.hints". These hints consist only of the directory search path. There is no hash table as in the a.out hints, because ELF doesn't have to search for the file with the highest minor version number. (It doesn't have minor version numbers at all.) A single run of ldconfig updates either the a.out hints or the ELF hints, but not both. The set of hints to process is selected in the usual way, via /etc/objformat, or ${OBJFORMAT}, or the "-aout" or "-elf" command line option. The rationale is that you probably want to search different directories for ELF than for a.out. "ldconfig -r" is faked up to produce output like we are used to, except that for ELF there are no minor version numbers. This should enable "ldconfig -r" to be used for checking LIB_DEPENDS in ports even for ELF. I implemented the ELF functionality in a new source file, with an eye toward eliminating the a.out code entirely at some point in the future.
1998-09-05 03:31:00 +00:00
}
Revamp the symbol lookup algorithm to cope better with objects loaded separately by dlopen that have global symbols with identical names. Viewing each dlopened object as a DAG which is linked by its DT_NEEDED entries in the dynamic table, the search order is as follows: * If the referencing object was linked with -Bsymbolic, search it internally. * Search all dlopened DAGs containing the referencing object. * Search all objects loaded at program start up. * Search all objects which were dlopened() using the RTLD_GLOBAL flag (which is now supported too). The search terminates as soon as a strong definition is found. Lacking that, the first weak definition is used. These rules match those of Solaris, as best I could determine them from its vague manual pages and the results of experiments I performed. PR: misc/12438
1999-08-30 01:48:19 +00:00
static void
init_dag(Obj_Entry *root)
{
Eliminate the use of symlook_needed function in favor of DAGS. Place elements on DAG lists in breadth-first order. This allows us to walk pre-built list in all cases where breadth-first dependency chain enumeration is required. Fix dlsym on special handle obtained by dlopen(NULL, ...) to do what comment claims it does. Take advantage of recently added symlook_global function to iterate over main objects and global DAGs lists properly in search of a symbol. Since rtld itself provides part of the global namespace, search rtld_obj too. Remove recursion from init_dag and symlook_needed functions. Use symlook_needed for ELF filtee processing only and change lookup order used in the function to match the order used by Solaris runtime linker under same circumstances. While there, fix weak symbol handling in the loop so that we return the first weak symbol definition if no strong one was found, instead of the last one. Reviewed by: kib MFC after: 1 month
2011-01-28 23:44:57 +00:00
const Needed_Entry *needed;
const Objlist_Entry *elm;
Solve the dynamic linker's problems with multithreaded programs once and for all (I hope). Packages such as wine, JDK, and linuxthreads should no longer have any problems with re-entering the dynamic linker. This commit replaces the locking used in the dynamic linker with a new spinlock-based reader/writer lock implementation. Brian Fundakowski Feldman <green> argued for this from the very beginning, but it took me a long time to come around to his point of view. Spinlocks are the only kinds of locks that work with all thread packages. But on uniprocessor systems they can be inefficient, because while a contender for the lock is spinning the holder of the lock cannot make any progress toward releasing it. To alleviate this disadvantage I have borrowed a trick from Sleepycat's Berkeley DB implementation. When spinning for a lock, the requester does a nanosleep() call for 1 usec. each time around the loop. This will generally yield the CPU to other threads, allowing the lock holder to finish its business and release the lock. I chose 1 usec. as the minimum sleep which would with reasonable certainty not be rounded down to 0. The formerly machine-independent file "lockdflt.c" has been moved into the architecture-specific subdirectories by repository copy. It now contains the machine-dependent spinlocking code. For the spinlocks I used the very nifty "simple, non-scalable reader-preference lock" which I found at <http://www.cs.rochester.edu/u/scott/synchronization/pseudocode/rw.html> on all CPUs except the 80386 (the specific CPU model, not the architecture). The 80386 CPU doesn't support the necessary "cmpxchg" instruction, so on that CPU a simple exclusive test-and-set lock is used instead. 80386 CPUs are detected at initialization time by trying to execute "cmpxchg" and catching the resulting SIGILL signal. To reduce contention for the locks, I have revamped a couple of key data structures, permitting all common operations to be done under non-exclusive (reader) locking. The only operations that require exclusive locking now are the rare intrusive operations such as dlopen() and dlclose(). The dllockinit() interface is now deprecated. It still exists, but only as a do-nothing stub. I plan to remove it as soon as is reasonably possible. (From the very beginning it was clearly labeled as experimental and subject to change.) As far as I know, only the linuxthreads port uses dllockinit(). This interface turned out to have several problems. As one example, when the dynamic linker called a client-supplied locking function, that function sometimes needed lazy binding, causing re-entry into the dynamic linker and a big looping mess. And in any case, it turned out to be too burdensome to require threads packages to register themselves with the dynamic linker.
2000-07-08 04:10:38 +00:00
DoneList donelist;
If dlopen() is called for the dso that has been already loaded as a dependency, then the dso never has its DAG initialized. Empty DAG makes ref_dag() call in dlopen() a nop, and the dso refcount is off by one. Initialize the DAG on the first dlopen() call, using a boolean flag to prevent double initialization. From the PR (edited): Assume we have a library liba.so, containing a function a(), and a library libb.so, containing function b(). liba.so needs functionality from libb.so, so liba.so links in libb.so. An application doesn't know about the relation between these libraries, but needs to call a() and b(). It dlopen()s liba.so and obtains a pointer to a(), then it dlopen()s libb.so and obtains a pointer to b(). As soon as the application doesn't need a() anymore, it dlclose()s liba.so. Expected result: the pointer to b() is still valid and can be called Actual result: the pointer to b() has become invalid, even though the application did not dlclose() the handle to libb.so. On calling b(), the application crashes with a segmentation fault. PR: misc/151861 Based on patch by: jh Reviewed by: kan Tested by: Arjan van Leeuwen <freebsd-maintainer opera com> MFC after: 1 week
2010-11-03 09:23:08 +00:00
if (root->dag_inited)
Fix style. MFC after: 6 days
2010-11-04 09:19:14 +00:00
return;
Solve the dynamic linker's problems with multithreaded programs once and for all (I hope). Packages such as wine, JDK, and linuxthreads should no longer have any problems with re-entering the dynamic linker. This commit replaces the locking used in the dynamic linker with a new spinlock-based reader/writer lock implementation. Brian Fundakowski Feldman <green> argued for this from the very beginning, but it took me a long time to come around to his point of view. Spinlocks are the only kinds of locks that work with all thread packages. But on uniprocessor systems they can be inefficient, because while a contender for the lock is spinning the holder of the lock cannot make any progress toward releasing it. To alleviate this disadvantage I have borrowed a trick from Sleepycat's Berkeley DB implementation. When spinning for a lock, the requester does a nanosleep() call for 1 usec. each time around the loop. This will generally yield the CPU to other threads, allowing the lock holder to finish its business and release the lock. I chose 1 usec. as the minimum sleep which would with reasonable certainty not be rounded down to 0. The formerly machine-independent file "lockdflt.c" has been moved into the architecture-specific subdirectories by repository copy. It now contains the machine-dependent spinlocking code. For the spinlocks I used the very nifty "simple, non-scalable reader-preference lock" which I found at <http://www.cs.rochester.edu/u/scott/synchronization/pseudocode/rw.html> on all CPUs except the 80386 (the specific CPU model, not the architecture). The 80386 CPU doesn't support the necessary "cmpxchg" instruction, so on that CPU a simple exclusive test-and-set lock is used instead. 80386 CPUs are detected at initialization time by trying to execute "cmpxchg" and catching the resulting SIGILL signal. To reduce contention for the locks, I have revamped a couple of key data structures, permitting all common operations to be done under non-exclusive (reader) locking. The only operations that require exclusive locking now are the rare intrusive operations such as dlopen() and dlclose(). The dllockinit() interface is now deprecated. It still exists, but only as a do-nothing stub. I plan to remove it as soon as is reasonably possible. (From the very beginning it was clearly labeled as experimental and subject to change.) As far as I know, only the linuxthreads port uses dllockinit(). This interface turned out to have several problems. As one example, when the dynamic linker called a client-supplied locking function, that function sometimes needed lazy binding, causing re-entry into the dynamic linker and a big looping mess. And in any case, it turned out to be too burdensome to require threads packages to register themselves with the dynamic linker.
2000-07-08 04:10:38 +00:00
donelist_init(&donelist);
Revamp the symbol lookup algorithm to cope better with objects loaded separately by dlopen that have global symbols with identical names. Viewing each dlopened object as a DAG which is linked by its DT_NEEDED entries in the dynamic table, the search order is as follows: * If the referencing object was linked with -Bsymbolic, search it internally. * Search all dlopened DAGs containing the referencing object. * Search all objects loaded at program start up. * Search all objects which were dlopened() using the RTLD_GLOBAL flag (which is now supported too). The search terminates as soon as a strong definition is found. Lacking that, the first weak definition is used. These rules match those of Solaris, as best I could determine them from its vague manual pages and the results of experiments I performed. PR: misc/12438
1999-08-30 01:48:19 +00:00
Eliminate the use of symlook_needed function in favor of DAGS. Place elements on DAG lists in breadth-first order. This allows us to walk pre-built list in all cases where breadth-first dependency chain enumeration is required. Fix dlsym on special handle obtained by dlopen(NULL, ...) to do what comment claims it does. Take advantage of recently added symlook_global function to iterate over main objects and global DAGs lists properly in search of a symbol. Since rtld itself provides part of the global namespace, search rtld_obj too. Remove recursion from init_dag and symlook_needed functions. Use symlook_needed for ELF filtee processing only and change lookup order used in the function to match the order used by Solaris runtime linker under same circumstances. While there, fix weak symbol handling in the loop so that we return the first weak symbol definition if no strong one was found, instead of the last one. Reviewed by: kib MFC after: 1 month
2011-01-28 23:44:57 +00:00
/* Root object belongs to own DAG. */
objlist_push_tail(&root->dldags, root);
objlist_push_tail(&root->dagmembers, root);
donelist_check(&donelist, root);
Rethink the way we count module references. Simply following DT_NEEDED links is not flexible enough for cases where dynamically loaded modules form a dependency cycle. This should fix an infinite recursion problem encountered by Yahoo. Approved by: re (jhb)
2003-05-08 01:31:36 +00:00
Eliminate the use of symlook_needed function in favor of DAGS. Place elements on DAG lists in breadth-first order. This allows us to walk pre-built list in all cases where breadth-first dependency chain enumeration is required. Fix dlsym on special handle obtained by dlopen(NULL, ...) to do what comment claims it does. Take advantage of recently added symlook_global function to iterate over main objects and global DAGs lists properly in search of a symbol. Since rtld itself provides part of the global namespace, search rtld_obj too. Remove recursion from init_dag and symlook_needed functions. Use symlook_needed for ELF filtee processing only and change lookup order used in the function to match the order used by Solaris runtime linker under same circumstances. While there, fix weak symbol handling in the loop so that we return the first weak symbol definition if no strong one was found, instead of the last one. Reviewed by: kib MFC after: 1 month
2011-01-28 23:44:57 +00:00
/*
* Add dependencies of root object to DAG in breadth order
* by exploiting the fact that each new object get added
* to the tail of the dagmembers list.
*/
STAILQ_FOREACH(elm, &root->dagmembers, link) {
for (needed = elm->obj->needed; needed != NULL; needed = needed->next) {
if (needed->obj == NULL || donelist_check(&donelist, needed->obj))
continue;
objlist_push_tail(&needed->obj->dldags, root);
objlist_push_tail(&root->dagmembers, needed->obj);
}
}
root->dag_inited = true;
Revamp the symbol lookup algorithm to cope better with objects loaded separately by dlopen that have global symbols with identical names. Viewing each dlopened object as a DAG which is linked by its DT_NEEDED entries in the dynamic table, the search order is as follows: * If the referencing object was linked with -Bsymbolic, search it internally. * Search all dlopened DAGs containing the referencing object. * Search all objects loaded at program start up. * Search all objects which were dlopened() using the RTLD_GLOBAL flag (which is now supported too). The search terminates as soon as a strong definition is found. Lacking that, the first weak definition is used. These rules match those of Solaris, as best I could determine them from its vague manual pages and the results of experiments I performed. PR: misc/12438
1999-08-30 01:48:19 +00:00
}
rtld: Fix a couple of bugs around the unloading of ELF filters. - Pass the correct object to unload_filtees(). - Use a marker to restart iteration after unload_filtees() has returned. It calls dlclose() and may recursively remove entries from the global object list, so TAILQ_FOREACH_SAFE is not sufficient. Reviewed by: kib MFC after: 2 weeks Sponsored by: Dell EMC Isilon
2016-12-22 17:44:27 +00:00
static void
init_marker(Obj_Entry *marker)
{
bzero(marker, sizeof(*marker));
marker->marker = true;
}
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
Obj_Entry *
globallist_curr(const Obj_Entry *obj)
{
for (;;) {
if (obj == NULL)
return (NULL);
if (!obj->marker)
return (__DECONST(Obj_Entry *, obj));
obj = TAILQ_PREV(obj, obj_entry_q, next);
}
}
Obj_Entry *
globallist_next(const Obj_Entry *obj)
{
for (;;) {
obj = TAILQ_NEXT(obj, next);
if (obj == NULL)
return (NULL);
if (!obj->marker)
return (__DECONST(Obj_Entry *, obj));
}
}
rtld: Fix a race between dl_iterate_phdr() and dlclose(). Add a transient reference count to ensure that the phdr argument to the callback remains valid while the bind lock is dropped. Reviewed by: kib MFC after: 2 weeks Sponsored by: Dell EMC Isilon
2016-12-22 17:37:39 +00:00
/* Prevent the object from being unmapped while the bind lock is dropped. */
static void
hold_object(Obj_Entry *obj)
{
obj->holdcount++;
}
static void
unhold_object(Obj_Entry *obj)
{
assert(obj->holdcount > 0);
if (--obj->holdcount == 0 && obj->unholdfree)
release_object(obj);
}
Pospone the DF_1_NODELETE processing until object DAG is fully loaded. Trying to up the reference from the load loop risks missing dependencies that have not been loaded yet. MFC afer: 1 week Reported by: nox Reviewd by: kib
2012-08-14 13:28:30 +00:00
static void
Implement support -z global linker option. It marks the shared object as always participating in the global symbols namespace, regardless of the way the object was brought into the process address space. Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-04-15 08:16:34 +00:00
process_z(Obj_Entry *root)
Pospone the DF_1_NODELETE processing until object DAG is fully loaded. Trying to up the reference from the load loop risks missing dependencies that have not been loaded yet. MFC afer: 1 week Reported by: nox Reviewd by: kib
2012-08-14 13:28:30 +00:00
{
const Objlist_Entry *elm;
Implement support -z global linker option. It marks the shared object as always participating in the global symbols namespace, regardless of the way the object was brought into the process address space. Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-04-15 08:16:34 +00:00
Obj_Entry *obj;
Pospone the DF_1_NODELETE processing until object DAG is fully loaded. Trying to up the reference from the load loop risks missing dependencies that have not been loaded yet. MFC afer: 1 week Reported by: nox Reviewd by: kib
2012-08-14 13:28:30 +00:00
/*
Implement support -z global linker option. It marks the shared object as always participating in the global symbols namespace, regardless of the way the object was brought into the process address space. Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-04-15 08:16:34 +00:00
* Walk over object DAG and process every dependent object
* that is marked as DF_1_NODELETE or DF_1_GLOBAL. They need
* to grow their own DAG.
*
* For DF_1_GLOBAL, DAG is required for symbol lookups in
* symlook_global() to work.
*
* For DF_1_NODELETE, the DAG should have its reference upped.
Pospone the DF_1_NODELETE processing until object DAG is fully loaded. Trying to up the reference from the load loop risks missing dependencies that have not been loaded yet. MFC afer: 1 week Reported by: nox Reviewd by: kib
2012-08-14 13:28:30 +00:00
*/
STAILQ_FOREACH(elm, &root->dagmembers, link) {
Implement support -z global linker option. It marks the shared object as always participating in the global symbols namespace, regardless of the way the object was brought into the process address space. Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-04-15 08:16:34 +00:00
obj = elm->obj;
if (obj == NULL)
continue;
if (obj->z_nodelete && !obj->ref_nodel) {
dbg("obj %s -z nodelete", obj->path);
init_dag(obj);
ref_dag(obj);
obj->ref_nodel = true;
}
if (obj->z_global && objlist_find(&list_global, obj) == NULL) {
dbg("obj %s -z global", obj->path);
objlist_push_tail(&list_global, obj);
init_dag(obj);
Pospone the DF_1_NODELETE processing until object DAG is fully loaded. Trying to up the reference from the load loop risks missing dependencies that have not been loaded yet. MFC afer: 1 week Reported by: nox Reviewd by: kib
2012-08-14 13:28:30 +00:00
}
}
}
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
/*
* Initialize the dynamic linker. The argument is the address at which
* the dynamic linker has been mapped into memory. The primary task of
* this function is to relocate the dynamic linker.
*/
static void
Use the newly provided aux vectors to get pagesize and osreldate information. Use local version of getpagesize(), rtld_getpagesize() in private allocator. Override the __getosreldate() previously fetched from libc_pic.a with local version that uses aux value if present. Note that __getosreldate() is used by rtld indirectly, by mmap(2) libc wrapper. To be able to utilize aux, split digest_dynamic() for use by init_rtld() into two parts, where the first one does not call malloc(), and the second part uses it. init_rtld() is able to initialize global variables before digest_dynamic2() calls. In particular, pagesize and osreldate are set up from the aux values. Now, rtld avoids (two) sysctl calls in startup. Tested by: marius (sparc64) MFC after: 1 month
2010-08-17 09:05:39 +00:00
init_rtld(caddr_t mapbase, Elf_Auxinfo **aux_info)
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
{
Minor changes to make this work on sparc64. Approved by: jdp Tested on: alpha, i386, sparc64
2002-04-02 02:19:02 +00:00
Obj_Entry objtmp; /* Temporary rtld object */
Fill phdr and phsize for rtld object. It is needed for dl_iterate_phdr() reporting the correct values. PR: 211367 Sponsored by: The FreeBSD Foundation MFC after: 1 week
2016-08-12 18:31:44 +00:00
const Elf_Ehdr *ehdr;
Use the newly provided aux vectors to get pagesize and osreldate information. Use local version of getpagesize(), rtld_getpagesize() in private allocator. Override the __getosreldate() previously fetched from libc_pic.a with local version that uses aux value if present. Note that __getosreldate() is used by rtld indirectly, by mmap(2) libc wrapper. To be able to utilize aux, split digest_dynamic() for use by init_rtld() into two parts, where the first one does not call malloc(), and the second part uses it. init_rtld() is able to initialize global variables before digest_dynamic2() calls. In particular, pagesize and osreldate are set up from the aux values. Now, rtld avoids (two) sysctl calls in startup. Tested by: marius (sparc64) MFC after: 1 month
2010-08-17 09:05:39 +00:00
const Elf_Dyn *dyn_rpath;
const Elf_Dyn *dyn_soname;
Import the DragonFly BSD commit 4f0bc915b65fcf5a23214f6d221d65c80be68ad4 by John Marino <draco@marino.st>, with the following (edited) commit message Date: Sat, 24 Mar 2012 06:40:50 +0100 Subject: [PATCH 1/1] rtld: Implement DT_RUNPATH and -z nodefaultlib DT_RUNPATH is incorrectly being considered as an alias of DT_RPATH. The purpose of DT_RUNPATH is to have two different types of rpath: one that can be overridden by the environment variable LD_LIBRARY_PATH and one that can't. With the currently implementation, LD_LIBRARY_PATH will always trump any embedded rpath or runpath tags. Current path search order by rtld: ================================== LD_LIBRARY_PATH DT_RPATH / DT_RUNPATH (always the same) ldconfig hints file (default: /var/run/ld-elf.so.hints) /usr/lib New path search order by rtld: ============================== DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file /usr/lib The new path search matches how the linux runtime loader works. The other major added feature is support for linker flag "-z nodefaultlib". When this flag is passed to the linker, rtld will skip all references to the standard library search path ("/usr/lib" in this case but it could handle more color delimited paths) except in DT_RPATH and DT_RUNPATH. New path search order by rtld with -z nodefaultlib flag set: ============================================================ DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file (skips all references to /usr/lib) FreeBSD notes: - we fixed some bugs which were submitted to DragonFly and merged there as commit 1ff8a2bd3eb6e5587174c6a983303ea3a79e0002; - we added LD_LIBRARY_PATH_RPATH environment variable to switch to the previous behaviour of considering DT_RPATH a synonym for DT_RUNPATH; - the FreeBSD default search path is /lib:/usr/lib and not /usr/lib. Reviewed by: kan MFC after: 1 month MFC note: flip the ld_library_path_rpath default value for stable/9
2012-07-15 10:53:48 +00:00
const Elf_Dyn *dyn_runpath;
Minor changes to make this work on sparc64. Approved by: jdp Tested on: alpha, i386, sparc64
2002-04-02 02:19:02 +00:00
Fix r264346 for ia64. We need to allocate memory for the function descriptors in order to relocate RTLD itself. To allocate memory, we need the pagesizes array initialized, but that happens after RTLD is relocated. This ordering is important for amd64, but it's opposite of what ia64 needs. Handle this conflict with the define called RTLD_INIT_PAGESIZES_EARLY. When defined, obtain the page sizes before relocating rtld, otherwise do it after.
2014-07-02 22:04:12 +00:00
#ifdef RTLD_INIT_PAGESIZES_EARLY
/* The page size is required by the dynamic memory allocator. */
init_pagesizes(aux_info);
#endif
Make the pathname pointed to by the Obj_Entry structure for the dynamic linker itself dynamically allocated. All of them are supposed to be dynamically allocated, but we cheated before. It made gdb unhappy under some circumstances.
1998-09-15 21:07:52 +00:00
/*
* Conjure up an Obj_Entry structure for the dynamic linker.
*
Fix typo in comment. Submitted by: Christoph Mallon
2009-06-23 09:50:50 +00:00
* The "path" member can't be initialized yet because string constants
Fix a typo in the same comment, one line below. Submitted by: bf1783 googlemail com
2009-06-23 14:12:49 +00:00
* cannot yet be accessed. Below we will set it correctly.
Make the pathname pointed to by the Obj_Entry structure for the dynamic linker itself dynamically allocated. All of them are supposed to be dynamically allocated, but we cheated before. It made gdb unhappy under some circumstances.
1998-09-15 21:07:52 +00:00
*/
Initialise some uninitialised variables. Thanks to: valgrind
2003-12-31 15:10:41 +00:00
memset(&objtmp, 0, sizeof(objtmp));
Minor changes to make this work on sparc64. Approved by: jdp Tested on: alpha, i386, sparc64
2002-04-02 02:19:02 +00:00
objtmp.path = NULL;
objtmp.rtld = true;
objtmp.mapbase = mapbase;
Revert r274772: it is not valid on MIPS Reported by: sbruno
2014-11-25 03:50:31 +00:00
#ifdef PIC
Minor changes to make this work on sparc64. Approved by: jdp Tested on: alpha, i386, sparc64
2002-04-02 02:19:02 +00:00
objtmp.relocbase = mapbase;
Add alpha support. Submitted by: John Birrell <jb@cimlogic.com.au> (with extra hacks by me) Obtained from: Probably NetBSD
1998-09-04 19:03:57 +00:00
#endif
Retire long-broken/unused static rtld support rtld-elf has some vestigial support for building as a static executable. r45501 introduced a partial implementation with a prescient note that it "might never be enabled." r153515 introduced ELF symbol versioning support, and removed part of the unused build infrastructure for static rtld. GNU ld populates rela relocation addends and GOT entries with the same values, and rtld's run-time dynamic executable check relied on this. Alternate toolchains may not populate the GOT entries, which caused RTLD_IS_DYNAMIC to return false. Simplify rtld by just removing the unused check. If we want to restore static rtld support later on we ought to introduce a build-time #ifdef flag. PR: 214972 Reviewed by: kan MFC after: 1 month Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D8687
2016-12-02 14:23:26 +00:00
objtmp.dynamic = rtld_dynamic(&objtmp);
digest_dynamic1(&objtmp, 1, &dyn_rpath, &dyn_soname, &dyn_runpath);
assert(objtmp.needed == NULL);
MFp4: Add mips support for dynamic linking. This code came from the merged mips2 and Juniper mips repositories. Warner Losh, Randall Seager, Oleksandr Tymoshenko and Olivier Houchard worked to merge, debug and integrate this code. This code may also contain code derived from NetBSD.
2008-04-04 20:59:26 +00:00
#if !defined(__mips__)
Retire long-broken/unused static rtld support rtld-elf has some vestigial support for building as a static executable. r45501 introduced a partial implementation with a prescient note that it "might never be enabled." r153515 introduced ELF symbol versioning support, and removed part of the unused build infrastructure for static rtld. GNU ld populates rela relocation addends and GOT entries with the same values, and rtld's run-time dynamic executable check relied on this. Alternate toolchains may not populate the GOT entries, which caused RTLD_IS_DYNAMIC to return false. Simplify rtld by just removing the unused check. If we want to restore static rtld support later on we ought to introduce a build-time #ifdef flag. PR: 214972 Reviewed by: kan MFC after: 1 month Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D8687
2016-12-02 14:23:26 +00:00
/* MIPS has a bogus DT_TEXTREL. */
assert(!objtmp.textrel);
MFp4: Add mips support for dynamic linking. This code came from the merged mips2 and Juniper mips repositories. Warner Losh, Randall Seager, Oleksandr Tymoshenko and Olivier Houchard worked to merge, debug and integrate this code. This code may also contain code derived from NetBSD.
2008-04-04 20:59:26 +00:00
#endif
Retire long-broken/unused static rtld support rtld-elf has some vestigial support for building as a static executable. r45501 introduced a partial implementation with a prescient note that it "might never be enabled." r153515 introduced ELF symbol versioning support, and removed part of the unused build infrastructure for static rtld. GNU ld populates rela relocation addends and GOT entries with the same values, and rtld's run-time dynamic executable check relied on this. Alternate toolchains may not populate the GOT entries, which caused RTLD_IS_DYNAMIC to return false. Simplify rtld by just removing the unused check. If we want to restore static rtld support later on we ought to introduce a build-time #ifdef flag. PR: 214972 Reviewed by: kan MFC after: 1 month Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D8687
2016-12-02 14:23:26 +00:00
/*
* Temporarily put the dynamic linker entry into the object list, so
* that symbols can be found.
*/
relocate_objects(&objtmp, true, &objtmp, 0, NULL);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
Fill phdr and phsize for rtld object. It is needed for dl_iterate_phdr() reporting the correct values. PR: 211367 Sponsored by: The FreeBSD Foundation MFC after: 1 week
2016-08-12 18:31:44 +00:00
ehdr = (Elf_Ehdr *)mapbase;
objtmp.phdr = (Elf_Phdr *)((char *)mapbase + ehdr->e_phoff);
objtmp.phsize = ehdr->e_phnum * sizeof(objtmp.phdr[0]);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
Minor changes to make this work on sparc64. Approved by: jdp Tested on: alpha, i386, sparc64
2002-04-02 02:19:02 +00:00
/* Initialize the object list. */
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
TAILQ_INIT(&obj_list);
Minor changes to make this work on sparc64. Approved by: jdp Tested on: alpha, i386, sparc64
2002-04-02 02:19:02 +00:00
/* Now that non-local variables can be accesses, copy out obj_rtld. */
memcpy(&obj_rtld, &objtmp, sizeof(obj_rtld));
Add GDB support. The method and some of the code came from NetBSD's elf runtime linker.
1998-04-30 07:48:02 +00:00
Fix r264346 for ia64. We need to allocate memory for the function descriptors in order to relocate RTLD itself. To allocate memory, we need the pagesizes array initialized, but that happens after RTLD is relocated. This ordering is important for amd64, but it's opposite of what ia64 needs. Handle this conflict with the define called RTLD_INIT_PAGESIZES_EARLY. When defined, obtain the page sizes before relocating rtld, otherwise do it after.
2014-07-02 22:04:12 +00:00
#ifndef RTLD_INIT_PAGESIZES_EARLY
Before calling mmap() on a shared library's text and data sections, rtld first calls mmap() with the arguments PROT_NONE and MAP_ANON to reserve a single, contiguous range of virtual addresses for the entire shared library. Later, rtld calls mmap() with the the shared library's file descriptor and the argument MAP_FIXED to place the text and data sections within the reserved range. The rationale for mapping shared libraries in this way is explained in the commit message for Revision 190885. However, this approach does have an unintended, negative consequence. Since the first call to mmap() specifies MAP_ANON and not the shared library's file descriptor, the kernel has no idea what alignment the vm object backing the file prefers. As a result, the reserved range's alignment is unlikely to be the same as the vm object's, and so mapping with superpages becomes impossible. To address this problem, this revision adds the argument MAP_ALIGNED_SUPER to the first call to mmap() if the text section is larger than the smallest superpage size. To determine if the text section is larger than the smallest superpage size, rtld must always fetch the page size information. As a result, the private code for fetching the base page size in rtld's builtin malloc is redundant. Eliminate it. Requested by: kib Tested by: zbb (on arm) Reviewed by: kib (an earlier version) Discussed with: jhb
2014-04-11 16:55:25 +00:00
/* The page size is required by the dynamic memory allocator. */
init_pagesizes(aux_info);
Fix r264346 for ia64. We need to allocate memory for the function descriptors in order to relocate RTLD itself. To allocate memory, we need the pagesizes array initialized, but that happens after RTLD is relocated. This ordering is important for amd64, but it's opposite of what ia64 needs. Handle this conflict with the define called RTLD_INIT_PAGESIZES_EARLY. When defined, obtain the page sizes before relocating rtld, otherwise do it after.
2014-07-02 22:04:12 +00:00
#endif
Before calling mmap() on a shared library's text and data sections, rtld first calls mmap() with the arguments PROT_NONE and MAP_ANON to reserve a single, contiguous range of virtual addresses for the entire shared library. Later, rtld calls mmap() with the the shared library's file descriptor and the argument MAP_FIXED to place the text and data sections within the reserved range. The rationale for mapping shared libraries in this way is explained in the commit message for Revision 190885. However, this approach does have an unintended, negative consequence. Since the first call to mmap() specifies MAP_ANON and not the shared library's file descriptor, the kernel has no idea what alignment the vm object backing the file prefers. As a result, the reserved range's alignment is unlikely to be the same as the vm object's, and so mapping with superpages becomes impossible. To address this problem, this revision adds the argument MAP_ALIGNED_SUPER to the first call to mmap() if the text section is larger than the smallest superpage size. To determine if the text section is larger than the smallest superpage size, rtld must always fetch the page size information. As a result, the private code for fetching the base page size in rtld's builtin malloc is redundant. Eliminate it. Requested by: kib Tested by: zbb (on arm) Reviewed by: kib (an earlier version) Discussed with: jhb
2014-04-11 16:55:25 +00:00
Use the newly provided aux vectors to get pagesize and osreldate information. Use local version of getpagesize(), rtld_getpagesize() in private allocator. Override the __getosreldate() previously fetched from libc_pic.a with local version that uses aux value if present. Note that __getosreldate() is used by rtld indirectly, by mmap(2) libc wrapper. To be able to utilize aux, split digest_dynamic() for use by init_rtld() into two parts, where the first one does not call malloc(), and the second part uses it. init_rtld() is able to initialize global variables before digest_dynamic2() calls. In particular, pagesize and osreldate are set up from the aux values. Now, rtld avoids (two) sysctl calls in startup. Tested by: marius (sparc64) MFC after: 1 month
2010-08-17 09:05:39 +00:00
if (aux_info[AT_OSRELDATE] != NULL)
osreldate = aux_info[AT_OSRELDATE]->a_un.a_val;
Import the DragonFly BSD commit 4f0bc915b65fcf5a23214f6d221d65c80be68ad4 by John Marino <draco@marino.st>, with the following (edited) commit message Date: Sat, 24 Mar 2012 06:40:50 +0100 Subject: [PATCH 1/1] rtld: Implement DT_RUNPATH and -z nodefaultlib DT_RUNPATH is incorrectly being considered as an alias of DT_RPATH. The purpose of DT_RUNPATH is to have two different types of rpath: one that can be overridden by the environment variable LD_LIBRARY_PATH and one that can't. With the currently implementation, LD_LIBRARY_PATH will always trump any embedded rpath or runpath tags. Current path search order by rtld: ================================== LD_LIBRARY_PATH DT_RPATH / DT_RUNPATH (always the same) ldconfig hints file (default: /var/run/ld-elf.so.hints) /usr/lib New path search order by rtld: ============================== DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file /usr/lib The new path search matches how the linux runtime loader works. The other major added feature is support for linker flag "-z nodefaultlib". When this flag is passed to the linker, rtld will skip all references to the standard library search path ("/usr/lib" in this case but it could handle more color delimited paths) except in DT_RPATH and DT_RUNPATH. New path search order by rtld with -z nodefaultlib flag set: ============================================================ DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file (skips all references to /usr/lib) FreeBSD notes: - we fixed some bugs which were submitted to DragonFly and merged there as commit 1ff8a2bd3eb6e5587174c6a983303ea3a79e0002; - we added LD_LIBRARY_PATH_RPATH environment variable to switch to the previous behaviour of considering DT_RPATH a synonym for DT_RUNPATH; - the FreeBSD default search path is /lib:/usr/lib and not /usr/lib. Reviewed by: kan MFC after: 1 month MFC note: flip the ld_library_path_rpath default value for stable/9
2012-07-15 10:53:48 +00:00
digest_dynamic2(&obj_rtld, dyn_rpath, dyn_soname, dyn_runpath);
Use the newly provided aux vectors to get pagesize and osreldate information. Use local version of getpagesize(), rtld_getpagesize() in private allocator. Override the __getosreldate() previously fetched from libc_pic.a with local version that uses aux value if present. Note that __getosreldate() is used by rtld indirectly, by mmap(2) libc wrapper. To be able to utilize aux, split digest_dynamic() for use by init_rtld() into two parts, where the first one does not call malloc(), and the second part uses it. init_rtld() is able to initialize global variables before digest_dynamic2() calls. In particular, pagesize and osreldate are set up from the aux values. Now, rtld avoids (two) sysctl calls in startup. Tested by: marius (sparc64) MFC after: 1 month
2010-08-17 09:05:39 +00:00
Make the pathname pointed to by the Obj_Entry structure for the dynamic linker itself dynamically allocated. All of them are supposed to be dynamically allocated, but we cheated before. It made gdb unhappy under some circumstances.
1998-09-15 21:07:52 +00:00
/* Replace the path with a dynamically allocated copy. */
Rather than using the #define for path names, indirect through a char * variable that could change for different executable types detected.
2015-10-31 04:39:55 +00:00
obj_rtld.path = xstrdup(ld_path_rtld);
Make the pathname pointed to by the Obj_Entry structure for the dynamic linker itself dynamically allocated. All of them are supposed to be dynamically allocated, but we cheated before. It made gdb unhappy under some circumstances.
1998-09-15 21:07:52 +00:00
Add GDB support. The method and some of the code came from NetBSD's elf runtime linker.
1998-04-30 07:48:02 +00:00
r_debug.r_brk = r_debug_state;
r_debug.r_state = RT_CONSISTENT;
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
}
Before calling mmap() on a shared library's text and data sections, rtld first calls mmap() with the arguments PROT_NONE and MAP_ANON to reserve a single, contiguous range of virtual addresses for the entire shared library. Later, rtld calls mmap() with the the shared library's file descriptor and the argument MAP_FIXED to place the text and data sections within the reserved range. The rationale for mapping shared libraries in this way is explained in the commit message for Revision 190885. However, this approach does have an unintended, negative consequence. Since the first call to mmap() specifies MAP_ANON and not the shared library's file descriptor, the kernel has no idea what alignment the vm object backing the file prefers. As a result, the reserved range's alignment is unlikely to be the same as the vm object's, and so mapping with superpages becomes impossible. To address this problem, this revision adds the argument MAP_ALIGNED_SUPER to the first call to mmap() if the text section is larger than the smallest superpage size. To determine if the text section is larger than the smallest superpage size, rtld must always fetch the page size information. As a result, the private code for fetching the base page size in rtld's builtin malloc is redundant. Eliminate it. Requested by: kib Tested by: zbb (on arm) Reviewed by: kib (an earlier version) Discussed with: jhb
2014-04-11 16:55:25 +00:00
/*
* Retrieve the array of supported page sizes. The kernel provides the page
* sizes in increasing order.
*/
static void
init_pagesizes(Elf_Auxinfo **aux_info)
{
static size_t psa[MAXPAGESIZES];
int mib[2];
size_t len, size;
if (aux_info[AT_PAGESIZES] != NULL && aux_info[AT_PAGESIZESLEN] !=
NULL) {
size = aux_info[AT_PAGESIZESLEN]->a_un.a_val;
pagesizes = aux_info[AT_PAGESIZES]->a_un.a_ptr;
} else {
len = 2;
if (sysctlnametomib("hw.pagesizes", mib, &len) == 0)
size = sizeof(psa);
else {
/* As a fallback, retrieve the base page size. */
size = sizeof(psa[0]);
if (aux_info[AT_PAGESZ] != NULL) {
psa[0] = aux_info[AT_PAGESZ]->a_un.a_val;
goto psa_filled;
} else {
mib[0] = CTL_HW;
mib[1] = HW_PAGESIZE;
len = 2;
}
}
if (sysctl(mib, len, psa, &size, NULL, 0) == -1) {
_rtld_error("sysctl for hw.pagesize(s) failed");
Make die available as rtld_die for use by MD relocation code Reviewed by: kib Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D2213
2015-04-02 21:35:36 +00:00
rtld_die();
Before calling mmap() on a shared library's text and data sections, rtld first calls mmap() with the arguments PROT_NONE and MAP_ANON to reserve a single, contiguous range of virtual addresses for the entire shared library. Later, rtld calls mmap() with the the shared library's file descriptor and the argument MAP_FIXED to place the text and data sections within the reserved range. The rationale for mapping shared libraries in this way is explained in the commit message for Revision 190885. However, this approach does have an unintended, negative consequence. Since the first call to mmap() specifies MAP_ANON and not the shared library's file descriptor, the kernel has no idea what alignment the vm object backing the file prefers. As a result, the reserved range's alignment is unlikely to be the same as the vm object's, and so mapping with superpages becomes impossible. To address this problem, this revision adds the argument MAP_ALIGNED_SUPER to the first call to mmap() if the text section is larger than the smallest superpage size. To determine if the text section is larger than the smallest superpage size, rtld must always fetch the page size information. As a result, the private code for fetching the base page size in rtld's builtin malloc is redundant. Eliminate it. Requested by: kib Tested by: zbb (on arm) Reviewed by: kib (an earlier version) Discussed with: jhb
2014-04-11 16:55:25 +00:00
}
psa_filled:
pagesizes = psa;
}
npagesizes = size / sizeof(pagesizes[0]);
/* Discard any invalid entries at the end of the array. */
while (npagesizes > 0 && pagesizes[npagesizes - 1] == 0)
npagesizes--;
}
Revamp the code that calls shared libraries' init and fini functions. Formerly the init functions were called in the opposite of the order in which libraries were loaded, and libraries were loaded according to a breadth-first traversal of the dependency graph. That ordering came from SVR4.0, and it was easy to implement but not always sensible. Now we do a depth-first walk over the dependency graph and call the init functions in an order such that each shared object's needed objects are initialized before the shared object itself. At the same time we build a list of finalization (fini) functions in the opposite order, to guarantee correct C++ destructor ordering whenever possible. (It may not be possible if dlopen and dlclose are used in strange ways, but we come as close as one can come.) The need for this renovation has become apparent as more programs have started using multithreading. The multithreaded C library libc_r requires initialization, whereas the standard libc does not. Since virtually every other object depends on the C library, it is important that it get initialized first.
2000-07-26 04:24:40 +00:00
/*
* Add the init functions from a needed object list (and its recursive
* needed objects) to "list". This is not used directly; it is a helper
* function for initlist_add_objects(). The write lock must be held
* when this function is called.
*/
static void
initlist_add_neededs(Needed_Entry *needed, Objlist *list)
{
/* Recursively process the successor needed objects. */
if (needed->next != NULL)
initlist_add_neededs(needed->next, list);
/* Process the current needed object. */
if (needed->obj != NULL)
Fix initlist_add_object invocation parameters. The tail parameter should point to the last object for which dependencies should be processed. In most cases, this is the object itself.
2016-01-20 23:26:35 +00:00
initlist_add_objects(needed->obj, needed->obj, list);
Revamp the code that calls shared libraries' init and fini functions. Formerly the init functions were called in the opposite of the order in which libraries were loaded, and libraries were loaded according to a breadth-first traversal of the dependency graph. That ordering came from SVR4.0, and it was easy to implement but not always sensible. Now we do a depth-first walk over the dependency graph and call the init functions in an order such that each shared object's needed objects are initialized before the shared object itself. At the same time we build a list of finalization (fini) functions in the opposite order, to guarantee correct C++ destructor ordering whenever possible. (It may not be possible if dlopen and dlclose are used in strange ways, but we come as close as one can come.) The need for this renovation has become apparent as more programs have started using multithreading. The multithreaded C library libc_r requires initialization, whereas the standard libc does not. Since virtually every other object depends on the C library, it is important that it get initialized first.
2000-07-26 04:24:40 +00:00
}
/*
* Scan all of the DAGs rooted in the range of objects from "obj" to
* "tail" and add their init functions to "list". This recurses over
* the DAGs and ensure the proper init ordering such that each object's
* needed libraries are initialized before the object itself. At the
* same time, this function adds the objects to the global finalization
* list "list_fini" in the opposite order. The write lock must be
* held when this function is called.
*/
static void
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
initlist_add_objects(Obj_Entry *obj, Obj_Entry *tail, Objlist *list)
Revamp the code that calls shared libraries' init and fini functions. Formerly the init functions were called in the opposite of the order in which libraries were loaded, and libraries were loaded according to a breadth-first traversal of the dependency graph. That ordering came from SVR4.0, and it was easy to implement but not always sensible. Now we do a depth-first walk over the dependency graph and call the init functions in an order such that each shared object's needed objects are initialized before the shared object itself. At the same time we build a list of finalization (fini) functions in the opposite order, to guarantee correct C++ destructor ordering whenever possible. (It may not be possible if dlopen and dlclose are used in strange ways, but we come as close as one can come.) The need for this renovation has become apparent as more programs have started using multithreading. The multithreaded C library libc_r requires initialization, whereas the standard libc does not. Since virtually every other object depends on the C library, it is important that it get initialized first.
2000-07-26 04:24:40 +00:00
{
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
Obj_Entry *nobj;
Fix several problems with our ELF filters implementation. Do not relocate twice an object which happens to be needed by loaded binary (or dso) and some filtee opened due to symbol resolution when relocating need objects. Record the state of the relocation processing in Obj_Entry and short-circuit relocate_objects() if current object already processed. Do not call constructors for filtees loaded during the early relocation processing before image is initialized enough to run user-provided code. Filtees are loaded using dlopen_object(), which normally performs relocation and initialization. If filtee is lazy-loaded during the relocation of dso needed by the main object, dlopen_object() runs too earlier, when most runtime services are not yet ready. Postpone the constructors call to the time when main binary and depended libraries constructors are run, passing the new flag RTLD_LO_EARLY to dlopen_object(). Symbol lookups callers inform symlook_* functions about early stage of initialization with SYMLOOK_EARLY. Pass flags through all functions participating in object relocation. Use the opportunity and fix flags argument to find_symdef() in arch-specific reloc.c to use proper name SYMLOOK_IN_PLT instead of true, which happen to have the same numeric value. Reported and tested by: theraven Reviewed by: kan MFC after: 2 weeks
2012-03-20 13:20:49 +00:00
Allow order of initialization of loaded shared objects to be altered through their .init code. This might happen if init vector calls dlopen on its own and that dlopen causes some not yet initialized object to be initialized earlier as part of that dlopened DAG. Do not reset module reference counts to zero on final fini vector run when process is exiting. Just add an additional parameter to force fini vector invocation regardless of current reference count value if object was not destructed yet. This allows dlclose called from fini vector to proceed normally instead of failing with handle validation error. Reviewed by: kib Reported by: venki kaps
2009-06-20 14:16:41 +00:00
if (obj->init_scanned || obj->init_done)
Revamp the code that calls shared libraries' init and fini functions. Formerly the init functions were called in the opposite of the order in which libraries were loaded, and libraries were loaded according to a breadth-first traversal of the dependency graph. That ordering came from SVR4.0, and it was easy to implement but not always sensible. Now we do a depth-first walk over the dependency graph and call the init functions in an order such that each shared object's needed objects are initialized before the shared object itself. At the same time we build a list of finalization (fini) functions in the opposite order, to guarantee correct C++ destructor ordering whenever possible. (It may not be possible if dlopen and dlclose are used in strange ways, but we come as close as one can come.) The need for this renovation has become apparent as more programs have started using multithreading. The multithreaded C library libc_r requires initialization, whereas the standard libc does not. Since virtually every other object depends on the C library, it is important that it get initialized first.
2000-07-26 04:24:40 +00:00
return;
Allow order of initialization of loaded shared objects to be altered through their .init code. This might happen if init vector calls dlopen on its own and that dlopen causes some not yet initialized object to be initialized earlier as part of that dlopened DAG. Do not reset module reference counts to zero on final fini vector run when process is exiting. Just add an additional parameter to force fini vector invocation regardless of current reference count value if object was not destructed yet. This allows dlclose called from fini vector to proceed normally instead of failing with handle validation error. Reviewed by: kib Reported by: venki kaps
2009-06-20 14:16:41 +00:00
obj->init_scanned = true;
Revamp the code that calls shared libraries' init and fini functions. Formerly the init functions were called in the opposite of the order in which libraries were loaded, and libraries were loaded according to a breadth-first traversal of the dependency graph. That ordering came from SVR4.0, and it was easy to implement but not always sensible. Now we do a depth-first walk over the dependency graph and call the init functions in an order such that each shared object's needed objects are initialized before the shared object itself. At the same time we build a list of finalization (fini) functions in the opposite order, to guarantee correct C++ destructor ordering whenever possible. (It may not be possible if dlopen and dlclose are used in strange ways, but we come as close as one can come.) The need for this renovation has become apparent as more programs have started using multithreading. The multithreaded C library libc_r requires initialization, whereas the standard libc does not. Since virtually every other object depends on the C library, it is important that it get initialized first.
2000-07-26 04:24:40 +00:00
/* Recursively process the successor objects. */
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
nobj = globallist_next(obj);
Fix initlist_add_object invocation parameters. The tail parameter should point to the last object for which dependencies should be processed. In most cases, this is the object itself.
2016-01-20 23:26:35 +00:00
if (nobj != NULL && obj != tail)
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
initlist_add_objects(nobj, tail, list);
Revamp the code that calls shared libraries' init and fini functions. Formerly the init functions were called in the opposite of the order in which libraries were loaded, and libraries were loaded according to a breadth-first traversal of the dependency graph. That ordering came from SVR4.0, and it was easy to implement but not always sensible. Now we do a depth-first walk over the dependency graph and call the init functions in an order such that each shared object's needed objects are initialized before the shared object itself. At the same time we build a list of finalization (fini) functions in the opposite order, to guarantee correct C++ destructor ordering whenever possible. (It may not be possible if dlopen and dlclose are used in strange ways, but we come as close as one can come.) The need for this renovation has become apparent as more programs have started using multithreading. The multithreaded C library libc_r requires initialization, whereas the standard libc does not. Since virtually every other object depends on the C library, it is important that it get initialized first.
2000-07-26 04:24:40 +00:00
/* Recursively process the needed objects. */
if (obj->needed != NULL)
initlist_add_neededs(obj->needed, list);
Fix several problems with our ELF filters implementation. Do not relocate twice an object which happens to be needed by loaded binary (or dso) and some filtee opened due to symbol resolution when relocating need objects. Record the state of the relocation processing in Obj_Entry and short-circuit relocate_objects() if current object already processed. Do not call constructors for filtees loaded during the early relocation processing before image is initialized enough to run user-provided code. Filtees are loaded using dlopen_object(), which normally performs relocation and initialization. If filtee is lazy-loaded during the relocation of dso needed by the main object, dlopen_object() runs too earlier, when most runtime services are not yet ready. Postpone the constructors call to the time when main binary and depended libraries constructors are run, passing the new flag RTLD_LO_EARLY to dlopen_object(). Symbol lookups callers inform symlook_* functions about early stage of initialization with SYMLOOK_EARLY. Pass flags through all functions participating in object relocation. Use the opportunity and fix flags argument to find_symdef() in arch-specific reloc.c to use proper name SYMLOOK_IN_PLT instead of true, which happen to have the same numeric value. Reported and tested by: theraven Reviewed by: kan MFC after: 2 weeks
2012-03-20 13:20:49 +00:00
if (obj->needed_filtees != NULL)
initlist_add_neededs(obj->needed_filtees, list);
if (obj->needed_aux_filtees != NULL)
initlist_add_neededs(obj->needed_aux_filtees, list);
Revamp the code that calls shared libraries' init and fini functions. Formerly the init functions were called in the opposite of the order in which libraries were loaded, and libraries were loaded according to a breadth-first traversal of the dependency graph. That ordering came from SVR4.0, and it was easy to implement but not always sensible. Now we do a depth-first walk over the dependency graph and call the init functions in an order such that each shared object's needed objects are initialized before the shared object itself. At the same time we build a list of finalization (fini) functions in the opposite order, to guarantee correct C++ destructor ordering whenever possible. (It may not be possible if dlopen and dlclose are used in strange ways, but we come as close as one can come.) The need for this renovation has become apparent as more programs have started using multithreading. The multithreaded C library libc_r requires initialization, whereas the standard libc does not. Since virtually every other object depends on the C library, it is important that it get initialized first.
2000-07-26 04:24:40 +00:00
/* Add the object to the init list. */
Add support for preinit, init and fini arrays. Some ABIs, in particular on ARM, do require working init arrays. Traditional FreeBSD crt1 calls _init and _fini of the binary, instead of allowing runtime linker to arrange the calls. This was probably done to have the same crt code serve both statically and dynamically linked binaries. Since ABI mandates that first is called preinit array functions, then init, and then init array functions, the init have to be called from rtld now. To provide binary compatibility to old FreeBSD crt1, which calls _init itself, rtld only calls intializers and finalizers for main binary if binary has a note indicating that new crt was used for linking. Add parsing of ELF notes to rtld, and cache p_osrel value since we parsed it anyway. The patch is inspired by init_array support for DragonflyBSD, written by John Marino. Reviewed by: kan Tested by: andrew (arm, previous version), flo (sparc64, previous version) MFC after: 3 weeks
2012-03-11 20:03:09 +00:00
if (obj->preinit_array != (Elf_Addr)NULL || obj->init != (Elf_Addr)NULL ||
obj->init_array != (Elf_Addr)NULL)
Revamp the code that calls shared libraries' init and fini functions. Formerly the init functions were called in the opposite of the order in which libraries were loaded, and libraries were loaded according to a breadth-first traversal of the dependency graph. That ordering came from SVR4.0, and it was easy to implement but not always sensible. Now we do a depth-first walk over the dependency graph and call the init functions in an order such that each shared object's needed objects are initialized before the shared object itself. At the same time we build a list of finalization (fini) functions in the opposite order, to guarantee correct C++ destructor ordering whenever possible. (It may not be possible if dlopen and dlclose are used in strange ways, but we come as close as one can come.) The need for this renovation has become apparent as more programs have started using multithreading. The multithreaded C library libc_r requires initialization, whereas the standard libc does not. Since virtually every other object depends on the C library, it is important that it get initialized first.
2000-07-26 04:24:40 +00:00
objlist_push_tail(list, obj);
/* Add the object to the global fini list in the reverse order. */
Add support for preinit, init and fini arrays. Some ABIs, in particular on ARM, do require working init arrays. Traditional FreeBSD crt1 calls _init and _fini of the binary, instead of allowing runtime linker to arrange the calls. This was probably done to have the same crt code serve both statically and dynamically linked binaries. Since ABI mandates that first is called preinit array functions, then init, and then init array functions, the init have to be called from rtld now. To provide binary compatibility to old FreeBSD crt1, which calls _init itself, rtld only calls intializers and finalizers for main binary if binary has a note indicating that new crt was used for linking. Add parsing of ELF notes to rtld, and cache p_osrel value since we parsed it anyway. The patch is inspired by init_array support for DragonflyBSD, written by John Marino. Reviewed by: kan Tested by: andrew (arm, previous version), flo (sparc64, previous version) MFC after: 3 weeks
2012-03-11 20:03:09 +00:00
if ((obj->fini != (Elf_Addr)NULL || obj->fini_array != (Elf_Addr)NULL)
&& !obj->on_fini_list) {
Revamp the code that calls shared libraries' init and fini functions. Formerly the init functions were called in the opposite of the order in which libraries were loaded, and libraries were loaded according to a breadth-first traversal of the dependency graph. That ordering came from SVR4.0, and it was easy to implement but not always sensible. Now we do a depth-first walk over the dependency graph and call the init functions in an order such that each shared object's needed objects are initialized before the shared object itself. At the same time we build a list of finalization (fini) functions in the opposite order, to guarantee correct C++ destructor ordering whenever possible. (It may not be possible if dlopen and dlclose are used in strange ways, but we come as close as one can come.) The need for this renovation has become apparent as more programs have started using multithreading. The multithreaded C library libc_r requires initialization, whereas the standard libc does not. Since virtually every other object depends on the C library, it is important that it get initialized first.
2000-07-26 04:24:40 +00:00
objlist_push_head(&list_fini, obj);
Allow order of initialization of loaded shared objects to be altered through their .init code. This might happen if init vector calls dlopen on its own and that dlopen causes some not yet initialized object to be initialized earlier as part of that dlopened DAG. Do not reset module reference counts to zero on final fini vector run when process is exiting. Just add an additional parameter to force fini vector invocation regardless of current reference count value if object was not destructed yet. This allows dlclose called from fini vector to proceed normally instead of failing with handle validation error. Reviewed by: kib Reported by: venki kaps
2009-06-20 14:16:41 +00:00
obj->on_fini_list = true;
}
Revamp the code that calls shared libraries' init and fini functions. Formerly the init functions were called in the opposite of the order in which libraries were loaded, and libraries were loaded according to a breadth-first traversal of the dependency graph. That ordering came from SVR4.0, and it was easy to implement but not always sensible. Now we do a depth-first walk over the dependency graph and call the init functions in an order such that each shared object's needed objects are initialized before the shared object itself. At the same time we build a list of finalization (fini) functions in the opposite order, to guarantee correct C++ destructor ordering whenever possible. (It may not be possible if dlopen and dlclose are used in strange ways, but we come as close as one can come.) The need for this renovation has become apparent as more programs have started using multithreading. The multithreaded C library libc_r requires initialization, whereas the standard libc does not. Since virtually every other object depends on the C library, it is important that it get initialized first.
2000-07-26 04:24:40 +00:00
}
Add ia64 support. Various adjustments were made to existing targets to cope with a few interface changes required by the ia64. In particular, function pointers on ia64 need special treatment in rtld.
2001-10-15 18:48:42 +00:00
#ifndef FPTR_TARGET
#define FPTR_TARGET(f) ((Elf_Addr) (f))
#endif
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
static void
Avoid bind lock recursion. When dlclose(3) unloads an object with filtees, it recursively calls dlclose(3) on each filtee in free_needed_filtees(). Introduce dlclose_locked() helper, called from free_needed_filtees() instead of dlclose(), and pass the bind lockstate down to avoid recursing. Reported and tested by: jhibbits Sponsored by: The FreeBSD Foundation MFC after: 1 week
2017-03-09 21:05:47 +00:00
free_needed_filtees(Needed_Entry *n, RtldLockState *lockstate)
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
{
Needed_Entry *needed, *needed1;
for (needed = n; needed != NULL; needed = needed->next) {
if (needed->obj != NULL) {
Avoid bind lock recursion. When dlclose(3) unloads an object with filtees, it recursively calls dlclose(3) on each filtee in free_needed_filtees(). Introduce dlclose_locked() helper, called from free_needed_filtees() instead of dlclose(), and pass the bind lockstate down to avoid recursing. Reported and tested by: jhibbits Sponsored by: The FreeBSD Foundation MFC after: 1 week
2017-03-09 21:05:47 +00:00
dlclose_locked(needed->obj, lockstate);
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
needed->obj = NULL;
}
}
for (needed = n; needed != NULL; needed = needed1) {
needed1 = needed->next;
free(needed);
}
}
static void
Avoid bind lock recursion. When dlclose(3) unloads an object with filtees, it recursively calls dlclose(3) on each filtee in free_needed_filtees(). Introduce dlclose_locked() helper, called from free_needed_filtees() instead of dlclose(), and pass the bind lockstate down to avoid recursing. Reported and tested by: jhibbits Sponsored by: The FreeBSD Foundation MFC after: 1 week
2017-03-09 21:05:47 +00:00
unload_filtees(Obj_Entry *obj, RtldLockState *lockstate)
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
{
Avoid bind lock recursion. When dlclose(3) unloads an object with filtees, it recursively calls dlclose(3) on each filtee in free_needed_filtees(). Introduce dlclose_locked() helper, called from free_needed_filtees() instead of dlclose(), and pass the bind lockstate down to avoid recursing. Reported and tested by: jhibbits Sponsored by: The FreeBSD Foundation MFC after: 1 week
2017-03-09 21:05:47 +00:00
free_needed_filtees(obj->needed_filtees, lockstate);
obj->needed_filtees = NULL;
free_needed_filtees(obj->needed_aux_filtees, lockstate);
obj->needed_aux_filtees = NULL;
obj->filtees_loaded = false;
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
}
static void
Propagate the current state of rtld_bind_lock to dlopen_object() calls through the filter loading call chain. This fixes attempts to write-lock the already locked rtld_bind_lock when filter loading is initiated by relocation of dlopening dso. Reported and tested by: Taku YAMAMOTO <taku tackymt homeip net> MFC after: 1 week
2012-04-12 10:32:22 +00:00
load_filtee1(Obj_Entry *obj, Needed_Entry *needed, int flags,
RtldLockState *lockstate)
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
{
for (; needed != NULL; needed = needed->next) {
Implement fdlopen(3), an rtld interface to load shared object by file descriptor. Requested and tested by: des (previous version) Reviewed by: des, kan (previous version) MFC after: 2 weeks
2012-01-07 10:33:01 +00:00
needed->obj = dlopen_object(obj->strtab + needed->name, -1, obj,
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
flags, ((ld_loadfltr || obj->z_loadfltr) ? RTLD_NOW : RTLD_LAZY) |
Propagate the current state of rtld_bind_lock to dlopen_object() calls through the filter loading call chain. This fixes attempts to write-lock the already locked rtld_bind_lock when filter loading is initiated by relocation of dlopening dso. Reported and tested by: Taku YAMAMOTO <taku tackymt homeip net> MFC after: 1 week
2012-04-12 10:32:22 +00:00
RTLD_LOCAL, lockstate);
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
}
}
static void
load_filtees(Obj_Entry *obj, int flags, RtldLockState *lockstate)
{
lock_restart_for_upgrade(lockstate);
if (!obj->filtees_loaded) {
Propagate the current state of rtld_bind_lock to dlopen_object() calls through the filter loading call chain. This fixes attempts to write-lock the already locked rtld_bind_lock when filter loading is initiated by relocation of dlopening dso. Reported and tested by: Taku YAMAMOTO <taku tackymt homeip net> MFC after: 1 week
2012-04-12 10:32:22 +00:00
load_filtee1(obj, obj->needed_filtees, flags, lockstate);
load_filtee1(obj, obj->needed_aux_filtees, flags, lockstate);
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
obj->filtees_loaded = true;
}
}
static int
process_needed(Obj_Entry *obj, Needed_Entry *needed, int flags)
{
Obj_Entry *obj1;
for (; needed != NULL; needed = needed->next) {
Implement fdlopen(3), an rtld interface to load shared object by file descriptor. Requested and tested by: des (previous version) Reviewed by: des, kan (previous version) MFC after: 2 weeks
2012-01-07 10:33:01 +00:00
obj1 = needed->obj = load_object(obj->strtab + needed->name, -1, obj,
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
flags & ~RTLD_LO_NOLOAD);
if (obj1 == NULL && !ld_tracing && (flags & RTLD_LO_FILTEES) == 0)
return (-1);
}
return (0);
}
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
/*
* Given a shared object, traverse its list of needed objects, and load
* each of them. Returns 0 on success. Generates an error message and
* returns -1 on failure.
*/
static int
Implement rtld part of the support for -z nodlopen (see ld(1)). Reviewed by: kan MFC after: 3 weeks
2009-11-26 13:57:20 +00:00
load_needed_objects(Obj_Entry *first, int flags)
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
{
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
Obj_Entry *obj;
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
Remove all remaining uses of TAILQ_FOREACH_FROM() from rtld-elf. Sponsored by: The FreeBSD Foundation MFC after: 1 week
2016-08-12 18:29:11 +00:00
for (obj = first; obj != NULL; obj = TAILQ_NEXT(obj, next)) {
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
if (obj->marker)
continue;
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
if (process_needed(obj, obj->needed, flags) == -1)
return (-1);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
}
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
return (0);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
}
Make LD_PRELOAD work for ELF.
1998-09-22 02:09:56 +00:00
static int
load_preload_objects(void)
{
char *p = ld_preload;
Implement support for the interpose dso flag. Requested by: bf Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 1 week Approved by: re (glebius)
2013-10-07 08:19:30 +00:00
Obj_Entry *obj;
Allow files in LD_PRELOAD to be separated by white space, like Solaris and Linux.
2000-01-22 22:20:05 +00:00
static const char delim[] = " \t:;";
Make LD_PRELOAD work for ELF.
1998-09-22 02:09:56 +00:00
if (p == NULL)
Make NULL a (void*)0 whereever possible, and fix the warnings(-Werror) that this provokes. "Wherever possible" means "In the kernel OR NOT C++" (implying C). There are places where (void *) pointers are not valid, such as for function pointers, but in the special case of (void *)0, agreement settles on it being OK. Most of the fixes were NULL where an integer zero was needed; many of the fixes were NULL where ascii <nul> ('\0') was needed, and a few were just "other". Tested on: i386 sparc64
2004-03-05 08:10:19 +00:00
return 0;
Make LD_PRELOAD work for ELF.
1998-09-22 02:09:56 +00:00
Allow files in LD_PRELOAD to be separated by white space, like Solaris and Linux.
2000-01-22 22:20:05 +00:00
p += strspn(p, delim);
Make LD_PRELOAD work for ELF.
1998-09-22 02:09:56 +00:00
while (*p != '\0') {
Allow files in LD_PRELOAD to be separated by white space, like Solaris and Linux.
2000-01-22 22:20:05 +00:00
size_t len = strcspn(p, delim);
Make LD_PRELOAD work for ELF.
1998-09-22 02:09:56 +00:00
char savech;
savech = p[len];
p[len] = '\0';
Implement support for the interpose dso flag. Requested by: bf Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 1 week Approved by: re (glebius)
2013-10-07 08:19:30 +00:00
obj = load_object(p, -1, NULL, 0);
if (obj == NULL)
Make LD_PRELOAD work for ELF.
1998-09-22 02:09:56 +00:00
return -1; /* XXX - cleanup */
Implement support for the interpose dso flag. Requested by: bf Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 1 week Approved by: re (glebius)
2013-10-07 08:19:30 +00:00
obj->z_interpose = true;
Make LD_PRELOAD work for ELF.
1998-09-22 02:09:56 +00:00
p[len] = savech;
p += len;
Allow files in LD_PRELOAD to be separated by white space, like Solaris and Linux.
2000-01-22 22:20:05 +00:00
p += strspn(p, delim);
Make LD_PRELOAD work for ELF.
1998-09-22 02:09:56 +00:00
}
Add various utrace's for use with ktrace to the ELF runtime linker. To activate the traces, set the LD_UTRACE (or LD_32_UTRACE) environment variable. This also includes code in kdump(8) to parse the traces. Reviewed by: kan, jdp MFC after: 2 weeks
2007-01-09 17:50:05 +00:00
LD_UTRACE(UTRACE_PRELOAD_FINISHED, NULL, NULL, 0, 0, NULL);
Make LD_PRELOAD work for ELF.
1998-09-22 02:09:56 +00:00
return 0;
}
Implement fdlopen(3), an rtld interface to load shared object by file descriptor. Requested and tested by: des (previous version) Reviewed by: des, kan (previous version) MFC after: 2 weeks
2012-01-07 10:33:01 +00:00
static const char *
printable_path(const char *path)
{
return (path == NULL ? "<unknown>" : path);
}
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
/*
Implement fdlopen(3), an rtld interface to load shared object by file descriptor. Requested and tested by: des (previous version) Reviewed by: des, kan (previous version) MFC after: 2 weeks
2012-01-07 10:33:01 +00:00
* Load a shared object into memory, if it is not already loaded. The
* object may be specified by name or by user-supplied file descriptor
* fd_u. In the later case, the fd_u descriptor is not closed, but its
* duplicate is.
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
*
* Returns a pointer to the Obj_Entry for the object. Returns NULL
* on failure.
*/
static Obj_Entry *
Implement fdlopen(3), an rtld interface to load shared object by file descriptor. Requested and tested by: des (previous version) Reviewed by: des, kan (previous version) MFC after: 2 weeks
2012-01-07 10:33:01 +00:00
load_object(const char *name, int fd_u, const Obj_Entry *refobj, int flags)
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
{
Obj_Entry *obj;
Implement fdlopen(3), an rtld interface to load shared object by file descriptor. Requested and tested by: des (previous version) Reviewed by: des, kan (previous version) MFC after: 2 weeks
2012-01-07 10:33:01 +00:00
int fd;
When checking to see if a shared object is already loaded, look for a device/inode match if no pathname match is found.
1999-08-30 01:50:41 +00:00
struct stat sb;
Implement ELF symbol versioning using GNU semantics. This code aims to be compatible with symbol versioning support as implemented by GNU libc and documented by http://people.redhat.com/~drepper/symbol-versioning and LSB 3.0. Implement dlvsym() function to allow lookups for a specific version of a given symbol.
2005-12-18 19:43:33 +00:00
char *path;
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
Add the LD_LIBRARY_PATH_FDS environmental variable. This variable allows the loading of shared libraries via directory descriptors rather than via library paths. If LD_LIBRARY_PATH_FDS=3:4:12, the directories represented by file descriptors 3, 4 and 12 will searched for shared libraries before the normal path-based mechanisms are used. This allows us to execute unprivileged binaries from within a Capsicum sandbox even if they require shared libraries. Approved by: rwatson (mentor) Reviewed by: kib MFC after: 3 weeks Sponsored by: DARPA/AFRL
2014-06-20 17:08:32 +00:00
fd = -1;
Implement fdlopen(3), an rtld interface to load shared object by file descriptor. Requested and tested by: des (previous version) Reviewed by: des, kan (previous version) MFC after: 2 weeks
2012-01-07 10:33:01 +00:00
if (name != NULL) {
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
TAILQ_FOREACH(obj, &obj_list, next) {
rtld: Ensure that dlopen() cannot obtain a reference on a doomed object. rtld drops the bind lock to call fini functions in an object prior to unmapping it. The new "doomed" state flag prevents the acquisition of new references for an object while the lock is dropped. Reviewed by: kib MFC after: 2 weeks Sponsored by: Dell EMC Isilon
2016-12-22 17:41:32 +00:00
if (obj->marker || obj->doomed)
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
continue;
Implement fdlopen(3), an rtld interface to load shared object by file descriptor. Requested and tested by: des (previous version) Reviewed by: des, kan (previous version) MFC after: 2 weeks
2012-01-07 10:33:01 +00:00
if (object_match_name(obj, name))
return (obj);
}
Implement ELF symbol versioning using GNU semantics. This code aims to be compatible with symbol versioning support as implemented by GNU libc and documented by http://people.redhat.com/~drepper/symbol-versioning and LSB 3.0. Implement dlvsym() function to allow lookups for a specific version of a given symbol.
2005-12-18 19:43:33 +00:00
Add the LD_LIBRARY_PATH_FDS environmental variable. This variable allows the loading of shared libraries via directory descriptors rather than via library paths. If LD_LIBRARY_PATH_FDS=3:4:12, the directories represented by file descriptors 3, 4 and 12 will searched for shared libraries before the normal path-based mechanisms are used. This allows us to execute unprivileged binaries from within a Capsicum sandbox even if they require shared libraries. Approved by: rwatson (mentor) Reviewed by: kib MFC after: 3 weeks Sponsored by: DARPA/AFRL
2014-06-20 17:08:32 +00:00
path = find_library(name, refobj, &fd);
Implement fdlopen(3), an rtld interface to load shared object by file descriptor. Requested and tested by: des (previous version) Reviewed by: des, kan (previous version) MFC after: 2 weeks
2012-01-07 10:33:01 +00:00
if (path == NULL)
return (NULL);
} else
path = NULL;
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
Add the LD_LIBRARY_PATH_FDS environmental variable. This variable allows the loading of shared libraries via directory descriptors rather than via library paths. If LD_LIBRARY_PATH_FDS=3:4:12, the directories represented by file descriptors 3, 4 and 12 will searched for shared libraries before the normal path-based mechanisms are used. This allows us to execute unprivileged binaries from within a Capsicum sandbox even if they require shared libraries. Approved by: rwatson (mentor) Reviewed by: kib MFC after: 3 weeks Sponsored by: DARPA/AFRL
2014-06-20 17:08:32 +00:00
if (fd >= 0) {
/*
* search_library_pathfds() opens a fresh file descriptor for the
* library, so there is no need to dup().
*/
} else if (fd_u == -1) {
/*
* If we didn't find a match by pathname, or the name is not
* supplied, open the file and check again by device and inode.
* This avoids false mismatches caused by multiple links or ".."
* in pathnames.
*
* To avoid a race, we open the file and use fstat() rather than
* using stat().
*/
Support file verification in MAC. * Add VCREAT flag to indicate when a new file is being created * Add VVERIFY to indicate verification is required * Both VCREAT and VVERIFY are only passed on the MAC method vnode_check_open and are removed from the accmode after * Add O_VERIFY flag to rtld open of objects * Add 'v' flag to __sflags to set O_VERIFY flag. Submitted by: Steve Kiernan <stevek@juniper.net> Obtained from: Juniper Networks, Inc. GitHub Pull Request: https://github.com/freebsd/freebsd/pull/27 Relnotes: yes
2015-04-22 01:54:25 +00:00
if ((fd = open(path, O_RDONLY | O_CLOEXEC | O_VERIFY)) == -1) {
Implement fdlopen(3), an rtld interface to load shared object by file descriptor. Requested and tested by: des (previous version) Reviewed by: des, kan (previous version) MFC after: 2 weeks
2012-01-07 10:33:01 +00:00
_rtld_error("Cannot open \"%s\"", path);
free(path);
return (NULL);
}
} else {
rtld: Fix fd leak with parallel dlopen and fork/exec. Rtld did not set FD_CLOEXEC on its internal file descriptors; therefore, such a file descriptor may be passed to a process created by another thread running in parallel to dlopen() or fdlopen(). No other threads are expected to be running during parsing of the hints and libmap files but the file descriptors need not be passed to child processes so add O_CLOEXEC there as well. This change will break fdlopen() (as used by OpenPAM) on kernels without F_DUPFD_CLOEXEC (added in July). Note that running new userland on old kernels is not supported. Reviewed by: kib
2012-11-04 21:42:24 +00:00
fd = fcntl(fd_u, F_DUPFD_CLOEXEC, 0);
Implement fdlopen(3), an rtld interface to load shared object by file descriptor. Requested and tested by: des (previous version) Reviewed by: des, kan (previous version) MFC after: 2 weeks
2012-01-07 10:33:01 +00:00
if (fd == -1) {
_rtld_error("Cannot dup fd");
free(path);
return (NULL);
}
Implement ELF symbol versioning using GNU semantics. This code aims to be compatible with symbol versioning support as implemented by GNU libc and documented by http://people.redhat.com/~drepper/symbol-versioning and LSB 3.0. Implement dlvsym() function to allow lookups for a specific version of a given symbol.
2005-12-18 19:43:33 +00:00
}
if (fstat(fd, &sb) == -1) {
Implement fdlopen(3), an rtld interface to load shared object by file descriptor. Requested and tested by: des (previous version) Reviewed by: des, kan (previous version) MFC after: 2 weeks
2012-01-07 10:33:01 +00:00
_rtld_error("Cannot fstat \"%s\"", printable_path(path));
Implement ELF symbol versioning using GNU semantics. This code aims to be compatible with symbol versioning support as implemented by GNU libc and documented by http://people.redhat.com/~drepper/symbol-versioning and LSB 3.0. Implement dlvsym() function to allow lookups for a specific version of a given symbol.
2005-12-18 19:43:33 +00:00
close(fd);
free(path);
return NULL;
}
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
TAILQ_FOREACH(obj, &obj_list, next) {
rtld: Ensure that dlopen() cannot obtain a reference on a doomed object. rtld drops the bind lock to call fini functions in an object prior to unmapping it. The new "doomed" state flag prevents the acquisition of new references for an object while the lock is dropped. Reviewed by: kib MFC after: 2 weeks Sponsored by: Dell EMC Isilon
2016-12-22 17:41:32 +00:00
if (obj->marker || obj->doomed)
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
continue;
rtld: eliminate double call to close(2) that may occur in load_object The second close(2) call resulted in heisenbugs in some multi-threaded applications where e.g. dlopen(3) call in one thread could close a file descriptor for a file having been opened in other thread concurrently. My litmus test for this issue was an openoffice.org build. Reviewed by: jhb MFC after: 2 weeks
2011-03-25 18:23:10 +00:00
if (obj->ino == sb.st_ino && obj->dev == sb.st_dev)
Implement ELF symbol versioning using GNU semantics. This code aims to be compatible with symbol versioning support as implemented by GNU libc and documented by http://people.redhat.com/~drepper/symbol-versioning and LSB 3.0. Implement dlvsym() function to allow lookups for a specific version of a given symbol.
2005-12-18 19:43:33 +00:00
break;
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
}
Implement fdlopen(3), an rtld interface to load shared object by file descriptor. Requested and tested by: des (previous version) Reviewed by: des, kan (previous version) MFC after: 2 weeks
2012-01-07 10:33:01 +00:00
if (obj != NULL && name != NULL) {
Implement ELF symbol versioning using GNU semantics. This code aims to be compatible with symbol versioning support as implemented by GNU libc and documented by http://people.redhat.com/~drepper/symbol-versioning and LSB 3.0. Implement dlvsym() function to allow lookups for a specific version of a given symbol.
2005-12-18 19:43:33 +00:00
object_add_name(obj, name);
free(path);
close(fd);
return obj;
}
Plug possible memory leak. Found by: Coverity MFC after: 2 weeks
2010-06-03 00:25:43 +00:00
if (flags & RTLD_LO_NOLOAD) {
free(path);
Fix fd leak. Reported by: slonoman2011 yandex ru MFC after: 1 week
2011-11-18 09:55:47 +00:00
close(fd);
Implement RTLD_NOLOAD flag for dlopen(3). Requested and tested by: jkim Reviewed by: kan Approved by: re (kensmith)
2009-07-17 19:45:42 +00:00
return (NULL);
Plug possible memory leak. Found by: Coverity MFC after: 2 weeks
2010-06-03 00:25:43 +00:00
}
When checking to see if a shared object is already loaded, look for a device/inode match if no pathname match is found.
1999-08-30 01:50:41 +00:00
Implement ELF symbol versioning using GNU semantics. This code aims to be compatible with symbol versioning support as implemented by GNU libc and documented by http://people.redhat.com/~drepper/symbol-versioning and LSB 3.0. Implement dlvsym() function to allow lookups for a specific version of a given symbol.
2005-12-18 19:43:33 +00:00
/* First use of this object, so we must map it in */
Implement rtld part of the support for -z nodlopen (see ld(1)). Reviewed by: kan MFC after: 3 weeks
2009-11-26 13:57:20 +00:00
obj = do_load_object(fd, name, path, &sb, flags);
Implement ELF symbol versioning using GNU semantics. This code aims to be compatible with symbol versioning support as implemented by GNU libc and documented by http://people.redhat.com/~drepper/symbol-versioning and LSB 3.0. Implement dlvsym() function to allow lookups for a specific version of a given symbol.
2005-12-18 19:43:33 +00:00
if (obj == NULL)
free(path);
close(fd);
return obj;
}
static Obj_Entry *
Implement rtld part of the support for -z nodlopen (see ld(1)). Reviewed by: kan MFC after: 3 weeks
2009-11-26 13:57:20 +00:00
do_load_object(int fd, const char *name, char *path, struct stat *sbp,
int flags)
Implement ELF symbol versioning using GNU semantics. This code aims to be compatible with symbol versioning support as implemented by GNU libc and documented by http://people.redhat.com/~drepper/symbol-versioning and LSB 3.0. Implement dlvsym() function to allow lookups for a specific version of a given symbol.
2005-12-18 19:43:33 +00:00
{
Obj_Entry *obj;
struct statfs fs;
/*
* but first, make sure that environment variables haven't been
* used to circumvent the noexec flag on a filesystem.
*/
if (dangerous_ld_env) {
if (fstatfs(fd, &fs) != 0) {
Implement fdlopen(3), an rtld interface to load shared object by file descriptor. Requested and tested by: des (previous version) Reviewed by: des, kan (previous version) MFC after: 2 weeks
2012-01-07 10:33:01 +00:00
_rtld_error("Cannot fstatfs \"%s\"", printable_path(path));
return NULL;
If "dangerous" environment variables (LD_PRELOAD, LD_LIBMAP, LD_LIBMAP_DISABLE, LD_LIBRARY_PATH) are used, then make sure the libraries being loaded aren't on a noexec-mounted filesystem. This is a compromise position: I'm assuming that nobody will be silly enough to set the noexec mount flag on part of the default library path, in order to avoid adding extra overhead into the common case (where those environment variables aren't used). Discussed with: csjp, secteam MFC after: 1 week
2005-03-24 10:12:29 +00:00
}
Implement ELF symbol versioning using GNU semantics. This code aims to be compatible with symbol versioning support as implemented by GNU libc and documented by http://people.redhat.com/~drepper/symbol-versioning and LSB 3.0. Implement dlvsym() function to allow lookups for a specific version of a given symbol.
2005-12-18 19:43:33 +00:00
if (fs.f_flags & MNT_NOEXEC) {
_rtld_error("Cannot execute objects on %s\n", fs.f_mntonname);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
return NULL;
}
Implement ELF symbol versioning using GNU semantics. This code aims to be compatible with symbol versioning support as implemented by GNU libc and documented by http://people.redhat.com/~drepper/symbol-versioning and LSB 3.0. Implement dlvsym() function to allow lookups for a specific version of a given symbol.
2005-12-18 19:43:33 +00:00
}
Implement fdlopen(3), an rtld interface to load shared object by file descriptor. Requested and tested by: des (previous version) Reviewed by: des, kan (previous version) MFC after: 2 weeks
2012-01-07 10:33:01 +00:00
dbg("loading \"%s\"", printable_path(path));
obj = map_object(fd, printable_path(path), sbp);
Implement ELF symbol versioning using GNU semantics. This code aims to be compatible with symbol versioning support as implemented by GNU libc and documented by http://people.redhat.com/~drepper/symbol-versioning and LSB 3.0. Implement dlvsym() function to allow lookups for a specific version of a given symbol.
2005-12-18 19:43:33 +00:00
if (obj == NULL)
return NULL;
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
Implement fdlopen(3), an rtld interface to load shared object by file descriptor. Requested and tested by: des (previous version) Reviewed by: des, kan (previous version) MFC after: 2 weeks
2012-01-07 10:33:01 +00:00
/*
* If DT_SONAME is present in the object, digest_dynamic2 already
* added it to the object names.
*/
if (name != NULL)
object_add_name(obj, name);
Implement ELF symbol versioning using GNU semantics. This code aims to be compatible with symbol versioning support as implemented by GNU libc and documented by http://people.redhat.com/~drepper/symbol-versioning and LSB 3.0. Implement dlvsym() function to allow lookups for a specific version of a given symbol.
2005-12-18 19:43:33 +00:00
obj->path = path;
digest_dynamic(obj, 0);
Add GNU hash support for rtld. Based on dragonflybsd support for GNU hash by John Marino <draco marino st> Reviewed by: kan Tested by: bapt MFC after: 2 weeks
2012-04-30 13:31:10 +00:00
dbg("%s valid_hash_sysv %d valid_hash_gnu %d dynsymcount %d", obj->path,
obj->valid_hash_sysv, obj->valid_hash_gnu, obj->dynsymcount);
Allow to load not-openable dso when tracing. This fixes ldd on such dso or dso linked to non-openable object. Remove '\n' at the end of error message. End comments with dot. MFC after: 3 weeks (together with r199829)
2009-11-28 14:29:32 +00:00
if (obj->z_noopen && (flags & (RTLD_LO_DLOPEN | RTLD_LO_TRACE)) ==
RTLD_LO_DLOPEN) {
Implement rtld part of the support for -z nodlopen (see ld(1)). Reviewed by: kan MFC after: 3 weeks
2009-11-26 13:57:20 +00:00
dbg("refusing to load non-loadable \"%s\"", obj->path);
Allow to load not-openable dso when tracing. This fixes ldd on such dso or dso linked to non-openable object. Remove '\n' at the end of error message. End comments with dot. MFC after: 3 weeks (together with r199829)
2009-11-28 14:29:32 +00:00
_rtld_error("Cannot dlopen non-loadable %s", obj->path);
Implement rtld part of the support for -z nodlopen (see ld(1)). Reviewed by: kan MFC after: 3 weeks
2009-11-26 13:57:20 +00:00
munmap(obj->mapbase, obj->mapsize);
obj_free(obj);
return (NULL);
}
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
Add rtld private interface for dso to detect dynamic loading vs. static linking. Tested by: pho, antoine (exp-run) Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-01-03 18:09:53 +00:00
obj->dlopened = (flags & RTLD_LO_DLOPEN) != 0;
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
TAILQ_INSERT_TAIL(&obj_list, obj, next);
Implement ELF symbol versioning using GNU semantics. This code aims to be compatible with symbol versioning support as implemented by GNU libc and documented by http://people.redhat.com/~drepper/symbol-versioning and LSB 3.0. Implement dlvsym() function to allow lookups for a specific version of a given symbol.
2005-12-18 19:43:33 +00:00
obj_count++;
Implement dl_iterate_phdr function. Convert boolean flags in internal Obj_Entry structure into bitfields. Properly check for loaded segment alignment in map_object.
2007-04-03 18:31:20 +00:00
obj_loads++;
Implement ELF symbol versioning using GNU semantics. This code aims to be compatible with symbol versioning support as implemented by GNU libc and documented by http://people.redhat.com/~drepper/symbol-versioning and LSB 3.0. Implement dlvsym() function to allow lookups for a specific version of a given symbol.
2005-12-18 19:43:33 +00:00
linkmap_add(obj); /* for GDB & dlinfo() */
In rtld, read the initial stack access mode from AT_STACKPROT as set by kernel, and parse PT_GNU_STACK phdr from linked and loaded dsos. If the loaded dso requires executable stack, as specified by PF_X bit of p_flags of PT_GNU_STACK phdr, but current stack protection does not permit execution, the __pthread_map_stacks_exec symbol is looked up and called. It should be implemented in libc or threading library and change the protection mode of all thread stacks to be executable. Provide a private interface _rtld_get_stack_prot() to export the stack access mode as calculated by rtld. Reviewed by: kan
2011-01-08 17:11:49 +00:00
max_stack_flags |= obj->stack_flags;
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
Implement ELF symbol versioning using GNU semantics. This code aims to be compatible with symbol versioning support as implemented by GNU libc and documented by http://people.redhat.com/~drepper/symbol-versioning and LSB 3.0. Implement dlvsym() function to allow lookups for a specific version of a given symbol.
2005-12-18 19:43:33 +00:00
dbg(" %p .. %p: %s", obj->mapbase,
obj->mapbase + obj->mapsize - 1, obj->path);
if (obj->textrel)
dbg(" WARNING: %s has impure text", obj->path);
Add various utrace's for use with ktrace to the ELF runtime linker. To activate the traces, set the LD_UTRACE (or LD_32_UTRACE) environment variable. This also includes code in kdump(8) to parse the traces. Reviewed by: kan, jdp MFC after: 2 weeks
2007-01-09 17:50:05 +00:00
LD_UTRACE(UTRACE_LOAD_OBJECT, obj, obj->mapbase, obj->mapsize, 0,
obj->path);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
return obj;
}
static Obj_Entry *
obj_from_addr(const void *addr)
{
Obj_Entry *obj;
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
TAILQ_FOREACH(obj, &obj_list, next) {
if (obj->marker)
continue;
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
if (addr < (void *) obj->mapbase)
continue;
Do not depend on existence of _end symbol in obj_from_addr, use obj->mapbase and obj->mapsize instead. Prompted by: OpenOffice debugging session at last BSDCon.
2004-02-25 17:06:16 +00:00
if (addr < (void *) (obj->mapbase + obj->mapsize))
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
return obj;
}
return NULL;
}
Add support for preinit, init and fini arrays. Some ABIs, in particular on ARM, do require working init arrays. Traditional FreeBSD crt1 calls _init and _fini of the binary, instead of allowing runtime linker to arrange the calls. This was probably done to have the same crt code serve both statically and dynamically linked binaries. Since ABI mandates that first is called preinit array functions, then init, and then init array functions, the init have to be called from rtld now. To provide binary compatibility to old FreeBSD crt1, which calls _init itself, rtld only calls intializers and finalizers for main binary if binary has a note indicating that new crt was used for linking. Add parsing of ELF notes to rtld, and cache p_osrel value since we parsed it anyway. The patch is inspired by init_array support for DragonflyBSD, written by John Marino. Reviewed by: kan Tested by: andrew (arm, previous version), flo (sparc64, previous version) MFC after: 3 weeks
2012-03-11 20:03:09 +00:00
static void
preinit_main(void)
{
Elf_Addr *preinit_addr;
int index;
preinit_addr = (Elf_Addr *)obj_main->preinit_array;
Amend r232857, now dropping the casts entirely, as they were not necessary at all. Submitted by: stefanf
2012-03-12 11:22:23 +00:00
if (preinit_addr == NULL)
Add support for preinit, init and fini arrays. Some ABIs, in particular on ARM, do require working init arrays. Traditional FreeBSD crt1 calls _init and _fini of the binary, instead of allowing runtime linker to arrange the calls. This was probably done to have the same crt code serve both statically and dynamically linked binaries. Since ABI mandates that first is called preinit array functions, then init, and then init array functions, the init have to be called from rtld now. To provide binary compatibility to old FreeBSD crt1, which calls _init itself, rtld only calls intializers and finalizers for main binary if binary has a note indicating that new crt was used for linking. Add parsing of ELF notes to rtld, and cache p_osrel value since we parsed it anyway. The patch is inspired by init_array support for DragonflyBSD, written by John Marino. Reviewed by: kan Tested by: andrew (arm, previous version), flo (sparc64, previous version) MFC after: 3 weeks
2012-03-11 20:03:09 +00:00
return;
for (index = 0; index < obj_main->preinit_array_num; index++) {
if (preinit_addr[index] != 0 && preinit_addr[index] != 1) {
dbg("calling preinit function for %s at %p", obj_main->path,
(void *)preinit_addr[index]);
LD_UTRACE(UTRACE_INIT_CALL, obj_main, (void *)preinit_addr[index],
0, 0, obj_main->path);
call_init_pointer(obj_main, preinit_addr[index]);
}
}
}
Revamp the code that calls shared libraries' init and fini functions. Formerly the init functions were called in the opposite of the order in which libraries were loaded, and libraries were loaded according to a breadth-first traversal of the dependency graph. That ordering came from SVR4.0, and it was easy to implement but not always sensible. Now we do a depth-first walk over the dependency graph and call the init functions in an order such that each shared object's needed objects are initialized before the shared object itself. At the same time we build a list of finalization (fini) functions in the opposite order, to guarantee correct C++ destructor ordering whenever possible. (It may not be possible if dlopen and dlclose are used in strange ways, but we come as close as one can come.) The need for this renovation has become apparent as more programs have started using multithreading. The multithreaded C library libc_r requires initialization, whereas the standard libc does not. Since virtually every other object depends on the C library, it is important that it get initialized first.
2000-07-26 04:24:40 +00:00
/*
* Call the finalization functions for each of the objects in "list"
If dlclose() is called recursively from a _fini() function, the inner dlclose() call may unload the object of the outer call prematurely because objects are unreferenced before _fini() calls. Fix this by unreferencing objects after calling objlist_call_fini() in dlclose(). Therefore objlist_call_fini() now calls the fini function if the reference count of an object is 1. In addition we must restart the list_fini traversal after every _fini() call because another dlclose() call might have modified the reference counts. Add an XXX comment to objlist_call_fini() about possible race with dlopen(). PR: 133246, 149464 Reviewed by: kan, kib
2010-12-16 16:56:44 +00:00
* belonging to the DAG of "root" and referenced once. If NULL "root"
* is specified, every finalization function will be called regardless
* of the reference count and the list elements won't be freed. All of
* the objects are expected to have non-NULL fini functions.
Revamp the code that calls shared libraries' init and fini functions. Formerly the init functions were called in the opposite of the order in which libraries were loaded, and libraries were loaded according to a breadth-first traversal of the dependency graph. That ordering came from SVR4.0, and it was easy to implement but not always sensible. Now we do a depth-first walk over the dependency graph and call the init functions in an order such that each shared object's needed objects are initialized before the shared object itself. At the same time we build a list of finalization (fini) functions in the opposite order, to guarantee correct C++ destructor ordering whenever possible. (It may not be possible if dlopen and dlclose are used in strange ways, but we come as close as one can come.) The need for this renovation has become apparent as more programs have started using multithreading. The multithreaded C library libc_r requires initialization, whereas the standard libc does not. Since virtually every other object depends on the C library, it is important that it get initialized first.
2000-07-26 04:24:40 +00:00
*/
Revamp the symbol lookup algorithm to cope better with objects loaded separately by dlopen that have global symbols with identical names. Viewing each dlopened object as a DAG which is linked by its DT_NEEDED entries in the dynamic table, the search order is as follows: * If the referencing object was linked with -Bsymbolic, search it internally. * Search all dlopened DAGs containing the referencing object. * Search all objects loaded at program start up. * Search all objects which were dlopened() using the RTLD_GLOBAL flag (which is now supported too). The search terminates as soon as a strong definition is found. Lacking that, the first weak definition is used. These rules match those of Solaris, as best I could determine them from its vague manual pages and the results of experiments I performed. PR: misc/12438
1999-08-30 01:48:19 +00:00
static void
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
objlist_call_fini(Objlist *list, Obj_Entry *root, RtldLockState *lockstate)
Revamp the symbol lookup algorithm to cope better with objects loaded separately by dlopen that have global symbols with identical names. Viewing each dlopened object as a DAG which is linked by its DT_NEEDED entries in the dynamic table, the search order is as follows: * If the referencing object was linked with -Bsymbolic, search it internally. * Search all dlopened DAGs containing the referencing object. * Search all objects loaded at program start up. * Search all objects which were dlopened() using the RTLD_GLOBAL flag (which is now supported too). The search terminates as soon as a strong definition is found. Lacking that, the first weak definition is used. These rules match those of Solaris, as best I could determine them from its vague manual pages and the results of experiments I performed. PR: misc/12438
1999-08-30 01:48:19 +00:00
{
If dlclose() is called recursively from a _fini() function, the inner dlclose() call may unload the object of the outer call prematurely because objects are unreferenced before _fini() calls. Fix this by unreferencing objects after calling objlist_call_fini() in dlclose(). Therefore objlist_call_fini() now calls the fini function if the reference count of an object is 1. In addition we must restart the list_fini traversal after every _fini() call because another dlclose() call might have modified the reference counts. Add an XXX comment to objlist_call_fini() about possible race with dlopen(). PR: 133246, 149464 Reviewed by: kan, kib
2010-12-16 16:56:44 +00:00
Objlist_Entry *elm;
Fix a bug in which a program called dlclose from a destructor and got an assert failure in the dynamic linker.
2001-01-05 04:36:17 +00:00
char *saved_msg;
Add support for preinit, init and fini arrays. Some ABIs, in particular on ARM, do require working init arrays. Traditional FreeBSD crt1 calls _init and _fini of the binary, instead of allowing runtime linker to arrange the calls. This was probably done to have the same crt code serve both statically and dynamically linked binaries. Since ABI mandates that first is called preinit array functions, then init, and then init array functions, the init have to be called from rtld now. To provide binary compatibility to old FreeBSD crt1, which calls _init itself, rtld only calls intializers and finalizers for main binary if binary has a note indicating that new crt was used for linking. Add parsing of ELF notes to rtld, and cache p_osrel value since we parsed it anyway. The patch is inspired by init_array support for DragonflyBSD, written by John Marino. Reviewed by: kan Tested by: andrew (arm, previous version), flo (sparc64, previous version) MFC after: 3 weeks
2012-03-11 20:03:09 +00:00
Elf_Addr *fini_addr;
int index;
Revamp the symbol lookup algorithm to cope better with objects loaded separately by dlopen that have global symbols with identical names. Viewing each dlopened object as a DAG which is linked by its DT_NEEDED entries in the dynamic table, the search order is as follows: * If the referencing object was linked with -Bsymbolic, search it internally. * Search all dlopened DAGs containing the referencing object. * Search all objects loaded at program start up. * Search all objects which were dlopened() using the RTLD_GLOBAL flag (which is now supported too). The search terminates as soon as a strong definition is found. Lacking that, the first weak definition is used. These rules match those of Solaris, as best I could determine them from its vague manual pages and the results of experiments I performed. PR: misc/12438
1999-08-30 01:48:19 +00:00
If dlclose() is called recursively from a _fini() function, the inner dlclose() call may unload the object of the outer call prematurely because objects are unreferenced before _fini() calls. Fix this by unreferencing objects after calling objlist_call_fini() in dlclose(). Therefore objlist_call_fini() now calls the fini function if the reference count of an object is 1. In addition we must restart the list_fini traversal after every _fini() call because another dlclose() call might have modified the reference counts. Add an XXX comment to objlist_call_fini() about possible race with dlopen(). PR: 133246, 149464 Reviewed by: kan, kib
2010-12-16 16:56:44 +00:00
assert(root == NULL || root->refcount == 1);
rtld: Ensure that dlopen() cannot obtain a reference on a doomed object. rtld drops the bind lock to call fini functions in an object prior to unmapping it. The new "doomed" state flag prevents the acquisition of new references for an object while the lock is dropped. Reviewed by: kib MFC after: 2 weeks Sponsored by: Dell EMC Isilon
2016-12-22 17:41:32 +00:00
if (root != NULL)
root->doomed = true;
Fix a bug in which a program called dlclose from a destructor and got an assert failure in the dynamic linker.
2001-01-05 04:36:17 +00:00
/*
* Preserve the current error message since a fini function might
* call into the dynamic linker and overwrite it.
*/
saved_msg = errmsg_save();
If dlclose() is called recursively from a _fini() function, the inner dlclose() call may unload the object of the outer call prematurely because objects are unreferenced before _fini() calls. Fix this by unreferencing objects after calling objlist_call_fini() in dlclose(). Therefore objlist_call_fini() now calls the fini function if the reference count of an object is 1. In addition we must restart the list_fini traversal after every _fini() call because another dlclose() call might have modified the reference counts. Add an XXX comment to objlist_call_fini() about possible race with dlopen(). PR: 133246, 149464 Reviewed by: kan, kib
2010-12-16 16:56:44 +00:00
do {
STAILQ_FOREACH(elm, list, link) {
if (root != NULL && (elm->obj->refcount != 1 ||
objlist_find(&root->dagmembers, elm->obj) == NULL))
continue;
Allow order of initialization of loaded shared objects to be altered through their .init code. This might happen if init vector calls dlopen on its own and that dlopen causes some not yet initialized object to be initialized earlier as part of that dlopened DAG. Do not reset module reference counts to zero on final fini vector run when process is exiting. Just add an additional parameter to force fini vector invocation regardless of current reference count value if object was not destructed yet. This allows dlclose called from fini vector to proceed normally instead of failing with handle validation error. Reviewed by: kib Reported by: venki kaps
2009-06-20 14:16:41 +00:00
/* Remove object from fini list to prevent recursive invocation. */
STAILQ_REMOVE(list, elm, Struct_Objlist_Entry, link);
rtld: Ensure that dlopen() cannot obtain a reference on a doomed object. rtld drops the bind lock to call fini functions in an object prior to unmapping it. The new "doomed" state flag prevents the acquisition of new references for an object while the lock is dropped. Reviewed by: kib MFC after: 2 weeks Sponsored by: Dell EMC Isilon
2016-12-22 17:41:32 +00:00
/* Ensure that new references cannot be acquired. */
elm->obj->doomed = true;
rtld: Fix a race between dl_iterate_phdr() and dlclose(). Add a transient reference count to ensure that the phdr argument to the callback remains valid while the bind lock is dropped. Reviewed by: kib MFC after: 2 weeks Sponsored by: Dell EMC Isilon
2016-12-22 17:37:39 +00:00
hold_object(elm->obj);
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
lock_release(rtld_bind_lock, lockstate);
Add support for preinit, init and fini arrays. Some ABIs, in particular on ARM, do require working init arrays. Traditional FreeBSD crt1 calls _init and _fini of the binary, instead of allowing runtime linker to arrange the calls. This was probably done to have the same crt code serve both statically and dynamically linked binaries. Since ABI mandates that first is called preinit array functions, then init, and then init array functions, the init have to be called from rtld now. To provide binary compatibility to old FreeBSD crt1, which calls _init itself, rtld only calls intializers and finalizers for main binary if binary has a note indicating that new crt was used for linking. Add parsing of ELF notes to rtld, and cache p_osrel value since we parsed it anyway. The patch is inspired by init_array support for DragonflyBSD, written by John Marino. Reviewed by: kan Tested by: andrew (arm, previous version), flo (sparc64, previous version) MFC after: 3 weeks
2012-03-11 20:03:09 +00:00
/*
* It is legal to have both DT_FINI and DT_FINI_ARRAY defined.
* When this happens, DT_FINI_ARRAY is processed first.
*/
fini_addr = (Elf_Addr *)elm->obj->fini_array;
if (fini_addr != NULL && elm->obj->fini_array_num > 0) {
for (index = elm->obj->fini_array_num - 1; index >= 0;
index--) {
if (fini_addr[index] != 0 && fini_addr[index] != 1) {
dbg("calling fini function for %s at %p",
elm->obj->path, (void *)fini_addr[index]);
LD_UTRACE(UTRACE_FINI_CALL, elm->obj,
(void *)fini_addr[index], 0, 0, elm->obj->path);
call_initfini_pointer(elm->obj, fini_addr[index]);
}
}
}
if (elm->obj->fini != (Elf_Addr)NULL) {
dbg("calling fini function for %s at %p", elm->obj->path,
(void *)elm->obj->fini);
LD_UTRACE(UTRACE_FINI_CALL, elm->obj, (void *)elm->obj->fini,
0, 0, elm->obj->path);
call_initfini_pointer(elm->obj, elm->obj->fini);
}
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
wlock_acquire(rtld_bind_lock, lockstate);
rtld: Fix a race between dl_iterate_phdr() and dlclose(). Add a transient reference count to ensure that the phdr argument to the callback remains valid while the bind lock is dropped. Reviewed by: kib MFC after: 2 weeks Sponsored by: Dell EMC Isilon
2016-12-22 17:37:39 +00:00
unhold_object(elm->obj);
Allow order of initialization of loaded shared objects to be altered through their .init code. This might happen if init vector calls dlopen on its own and that dlopen causes some not yet initialized object to be initialized earlier as part of that dlopened DAG. Do not reset module reference counts to zero on final fini vector run when process is exiting. Just add an additional parameter to force fini vector invocation regardless of current reference count value if object was not destructed yet. This allows dlclose called from fini vector to proceed normally instead of failing with handle validation error. Reviewed by: kib Reported by: venki kaps
2009-06-20 14:16:41 +00:00
/* No need to free anything if process is going down. */
If dlclose() is called recursively from a _fini() function, the inner dlclose() call may unload the object of the outer call prematurely because objects are unreferenced before _fini() calls. Fix this by unreferencing objects after calling objlist_call_fini() in dlclose(). Therefore objlist_call_fini() now calls the fini function if the reference count of an object is 1. In addition we must restart the list_fini traversal after every _fini() call because another dlclose() call might have modified the reference counts. Add an XXX comment to objlist_call_fini() about possible race with dlopen(). PR: 133246, 149464 Reviewed by: kan, kib
2010-12-16 16:56:44 +00:00
if (root != NULL)
Allow order of initialization of loaded shared objects to be altered through their .init code. This might happen if init vector calls dlopen on its own and that dlopen causes some not yet initialized object to be initialized earlier as part of that dlopened DAG. Do not reset module reference counts to zero on final fini vector run when process is exiting. Just add an additional parameter to force fini vector invocation regardless of current reference count value if object was not destructed yet. This allows dlclose called from fini vector to proceed normally instead of failing with handle validation error. Reviewed by: kib Reported by: venki kaps
2009-06-20 14:16:41 +00:00
free(elm);
If dlclose() is called recursively from a _fini() function, the inner dlclose() call may unload the object of the outer call prematurely because objects are unreferenced before _fini() calls. Fix this by unreferencing objects after calling objlist_call_fini() in dlclose(). Therefore objlist_call_fini() now calls the fini function if the reference count of an object is 1. In addition we must restart the list_fini traversal after every _fini() call because another dlclose() call might have modified the reference counts. Add an XXX comment to objlist_call_fini() about possible race with dlopen(). PR: 133246, 149464 Reviewed by: kan, kib
2010-12-16 16:56:44 +00:00
/*
* We must restart the list traversal after every fini call
* because a dlclose() call from the fini function or from
* another thread might have modified the reference counts.
*/
break;
Revamp the code that calls shared libraries' init and fini functions. Formerly the init functions were called in the opposite of the order in which libraries were loaded, and libraries were loaded according to a breadth-first traversal of the dependency graph. That ordering came from SVR4.0, and it was easy to implement but not always sensible. Now we do a depth-first walk over the dependency graph and call the init functions in an order such that each shared object's needed objects are initialized before the shared object itself. At the same time we build a list of finalization (fini) functions in the opposite order, to guarantee correct C++ destructor ordering whenever possible. (It may not be possible if dlopen and dlclose are used in strange ways, but we come as close as one can come.) The need for this renovation has become apparent as more programs have started using multithreading. The multithreaded C library libc_r requires initialization, whereas the standard libc does not. Since virtually every other object depends on the C library, it is important that it get initialized first.
2000-07-26 04:24:40 +00:00
}
If dlclose() is called recursively from a _fini() function, the inner dlclose() call may unload the object of the outer call prematurely because objects are unreferenced before _fini() calls. Fix this by unreferencing objects after calling objlist_call_fini() in dlclose(). Therefore objlist_call_fini() now calls the fini function if the reference count of an object is 1. In addition we must restart the list_fini traversal after every _fini() call because another dlclose() call might have modified the reference counts. Add an XXX comment to objlist_call_fini() about possible race with dlopen(). PR: 133246, 149464 Reviewed by: kan, kib
2010-12-16 16:56:44 +00:00
} while (elm != NULL);
Fix a bug in which a program called dlclose from a destructor and got an assert failure in the dynamic linker.
2001-01-05 04:36:17 +00:00
errmsg_restore(saved_msg);
Revamp the code that calls shared libraries' init and fini functions. Formerly the init functions were called in the opposite of the order in which libraries were loaded, and libraries were loaded according to a breadth-first traversal of the dependency graph. That ordering came from SVR4.0, and it was easy to implement but not always sensible. Now we do a depth-first walk over the dependency graph and call the init functions in an order such that each shared object's needed objects are initialized before the shared object itself. At the same time we build a list of finalization (fini) functions in the opposite order, to guarantee correct C++ destructor ordering whenever possible. (It may not be possible if dlopen and dlclose are used in strange ways, but we come as close as one can come.) The need for this renovation has become apparent as more programs have started using multithreading. The multithreaded C library libc_r requires initialization, whereas the standard libc does not. Since virtually every other object depends on the C library, it is important that it get initialized first.
2000-07-26 04:24:40 +00:00
}
/*
* Call the initialization functions for each of the objects in
* "list". All of the objects are expected to have non-NULL init
* functions.
*/
static void
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
objlist_call_init(Objlist *list, RtldLockState *lockstate)
Revamp the code that calls shared libraries' init and fini functions. Formerly the init functions were called in the opposite of the order in which libraries were loaded, and libraries were loaded according to a breadth-first traversal of the dependency graph. That ordering came from SVR4.0, and it was easy to implement but not always sensible. Now we do a depth-first walk over the dependency graph and call the init functions in an order such that each shared object's needed objects are initialized before the shared object itself. At the same time we build a list of finalization (fini) functions in the opposite order, to guarantee correct C++ destructor ordering whenever possible. (It may not be possible if dlopen and dlclose are used in strange ways, but we come as close as one can come.) The need for this renovation has become apparent as more programs have started using multithreading. The multithreaded C library libc_r requires initialization, whereas the standard libc does not. Since virtually every other object depends on the C library, it is important that it get initialized first.
2000-07-26 04:24:40 +00:00
{
Objlist_Entry *elm;
Allow order of initialization of loaded shared objects to be altered through their .init code. This might happen if init vector calls dlopen on its own and that dlopen causes some not yet initialized object to be initialized earlier as part of that dlopened DAG. Do not reset module reference counts to zero on final fini vector run when process is exiting. Just add an additional parameter to force fini vector invocation regardless of current reference count value if object was not destructed yet. This allows dlclose called from fini vector to proceed normally instead of failing with handle validation error. Reviewed by: kib Reported by: venki kaps
2009-06-20 14:16:41 +00:00
Obj_Entry *obj;
Fix a bug in which a program called dlclose from a destructor and got an assert failure in the dynamic linker.
2001-01-05 04:36:17 +00:00
char *saved_msg;
Add support for preinit, init and fini arrays. Some ABIs, in particular on ARM, do require working init arrays. Traditional FreeBSD crt1 calls _init and _fini of the binary, instead of allowing runtime linker to arrange the calls. This was probably done to have the same crt code serve both statically and dynamically linked binaries. Since ABI mandates that first is called preinit array functions, then init, and then init array functions, the init have to be called from rtld now. To provide binary compatibility to old FreeBSD crt1, which calls _init itself, rtld only calls intializers and finalizers for main binary if binary has a note indicating that new crt was used for linking. Add parsing of ELF notes to rtld, and cache p_osrel value since we parsed it anyway. The patch is inspired by init_array support for DragonflyBSD, written by John Marino. Reviewed by: kan Tested by: andrew (arm, previous version), flo (sparc64, previous version) MFC after: 3 weeks
2012-03-11 20:03:09 +00:00
Elf_Addr *init_addr;
int index;
Revamp the code that calls shared libraries' init and fini functions. Formerly the init functions were called in the opposite of the order in which libraries were loaded, and libraries were loaded according to a breadth-first traversal of the dependency graph. That ordering came from SVR4.0, and it was easy to implement but not always sensible. Now we do a depth-first walk over the dependency graph and call the init functions in an order such that each shared object's needed objects are initialized before the shared object itself. At the same time we build a list of finalization (fini) functions in the opposite order, to guarantee correct C++ destructor ordering whenever possible. (It may not be possible if dlopen and dlclose are used in strange ways, but we come as close as one can come.) The need for this renovation has become apparent as more programs have started using multithreading. The multithreaded C library libc_r requires initialization, whereas the standard libc does not. Since virtually every other object depends on the C library, it is important that it get initialized first.
2000-07-26 04:24:40 +00:00
Allow order of initialization of loaded shared objects to be altered through their .init code. This might happen if init vector calls dlopen on its own and that dlopen causes some not yet initialized object to be initialized earlier as part of that dlopened DAG. Do not reset module reference counts to zero on final fini vector run when process is exiting. Just add an additional parameter to force fini vector invocation regardless of current reference count value if object was not destructed yet. This allows dlclose called from fini vector to proceed normally instead of failing with handle validation error. Reviewed by: kib Reported by: venki kaps
2009-06-20 14:16:41 +00:00
/*
* Clean init_scanned flag so that objects can be rechecked and
* possibly initialized earlier if any of vectors called below
* cause the change by using dlopen.
*/
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
TAILQ_FOREACH(obj, &obj_list, next) {
if (obj->marker)
continue;
Allow order of initialization of loaded shared objects to be altered through their .init code. This might happen if init vector calls dlopen on its own and that dlopen causes some not yet initialized object to be initialized earlier as part of that dlopened DAG. Do not reset module reference counts to zero on final fini vector run when process is exiting. Just add an additional parameter to force fini vector invocation regardless of current reference count value if object was not destructed yet. This allows dlclose called from fini vector to proceed normally instead of failing with handle validation error. Reviewed by: kib Reported by: venki kaps
2009-06-20 14:16:41 +00:00
obj->init_scanned = false;
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
}
Allow order of initialization of loaded shared objects to be altered through their .init code. This might happen if init vector calls dlopen on its own and that dlopen causes some not yet initialized object to be initialized earlier as part of that dlopened DAG. Do not reset module reference counts to zero on final fini vector run when process is exiting. Just add an additional parameter to force fini vector invocation regardless of current reference count value if object was not destructed yet. This allows dlclose called from fini vector to proceed normally instead of failing with handle validation error. Reviewed by: kib Reported by: venki kaps
2009-06-20 14:16:41 +00:00
Fix a bug in which a program called dlclose from a destructor and got an assert failure in the dynamic linker.
2001-01-05 04:36:17 +00:00
/*
* Preserve the current error message since an init function might
* call into the dynamic linker and overwrite it.
*/
saved_msg = errmsg_save();
Revamp the code that calls shared libraries' init and fini functions. Formerly the init functions were called in the opposite of the order in which libraries were loaded, and libraries were loaded according to a breadth-first traversal of the dependency graph. That ordering came from SVR4.0, and it was easy to implement but not always sensible. Now we do a depth-first walk over the dependency graph and call the init functions in an order such that each shared object's needed objects are initialized before the shared object itself. At the same time we build a list of finalization (fini) functions in the opposite order, to guarantee correct C++ destructor ordering whenever possible. (It may not be possible if dlopen and dlclose are used in strange ways, but we come as close as one can come.) The need for this renovation has become apparent as more programs have started using multithreading. The multithreaded C library libc_r requires initialization, whereas the standard libc does not. Since virtually every other object depends on the C library, it is important that it get initialized first.
2000-07-26 04:24:40 +00:00
STAILQ_FOREACH(elm, list, link) {
Allow order of initialization of loaded shared objects to be altered through their .init code. This might happen if init vector calls dlopen on its own and that dlopen causes some not yet initialized object to be initialized earlier as part of that dlopened DAG. Do not reset module reference counts to zero on final fini vector run when process is exiting. Just add an additional parameter to force fini vector invocation regardless of current reference count value if object was not destructed yet. This allows dlclose called from fini vector to proceed normally instead of failing with handle validation error. Reviewed by: kib Reported by: venki kaps
2009-06-20 14:16:41 +00:00
if (elm->obj->init_done) /* Initialized early. */
continue;
/*
* Race: other thread might try to use this object before current
Fix typo. MFC after: 3 days
2016-12-16 14:23:08 +00:00
* one completes the initialization. Not much can be done here
Allow order of initialization of loaded shared objects to be altered through their .init code. This might happen if init vector calls dlopen on its own and that dlopen causes some not yet initialized object to be initialized earlier as part of that dlopened DAG. Do not reset module reference counts to zero on final fini vector run when process is exiting. Just add an additional parameter to force fini vector invocation regardless of current reference count value if object was not destructed yet. This allows dlclose called from fini vector to proceed normally instead of failing with handle validation error. Reviewed by: kib Reported by: venki kaps
2009-06-20 14:16:41 +00:00
* without better locking.
*/
elm->obj->init_done = true;
rtld: Fix a race between dl_iterate_phdr() and dlclose(). Add a transient reference count to ensure that the phdr argument to the callback remains valid while the bind lock is dropped. Reviewed by: kib MFC after: 2 weeks Sponsored by: Dell EMC Isilon
2016-12-22 17:37:39 +00:00
hold_object(elm->obj);
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
lock_release(rtld_bind_lock, lockstate);
Add support for preinit, init and fini arrays. Some ABIs, in particular on ARM, do require working init arrays. Traditional FreeBSD crt1 calls _init and _fini of the binary, instead of allowing runtime linker to arrange the calls. This was probably done to have the same crt code serve both statically and dynamically linked binaries. Since ABI mandates that first is called preinit array functions, then init, and then init array functions, the init have to be called from rtld now. To provide binary compatibility to old FreeBSD crt1, which calls _init itself, rtld only calls intializers and finalizers for main binary if binary has a note indicating that new crt was used for linking. Add parsing of ELF notes to rtld, and cache p_osrel value since we parsed it anyway. The patch is inspired by init_array support for DragonflyBSD, written by John Marino. Reviewed by: kan Tested by: andrew (arm, previous version), flo (sparc64, previous version) MFC after: 3 weeks
2012-03-11 20:03:09 +00:00
/*
* It is legal to have both DT_INIT and DT_INIT_ARRAY defined.
* When this happens, DT_INIT is processed first.
*/
if (elm->obj->init != (Elf_Addr)NULL) {
dbg("calling init function for %s at %p", elm->obj->path,
(void *)elm->obj->init);
LD_UTRACE(UTRACE_INIT_CALL, elm->obj, (void *)elm->obj->init,
0, 0, elm->obj->path);
call_initfini_pointer(elm->obj, elm->obj->init);
}
init_addr = (Elf_Addr *)elm->obj->init_array;
Amend r232857, now dropping the casts entirely, as they were not necessary at all. Submitted by: stefanf
2012-03-12 11:22:23 +00:00
if (init_addr != NULL) {
Add support for preinit, init and fini arrays. Some ABIs, in particular on ARM, do require working init arrays. Traditional FreeBSD crt1 calls _init and _fini of the binary, instead of allowing runtime linker to arrange the calls. This was probably done to have the same crt code serve both statically and dynamically linked binaries. Since ABI mandates that first is called preinit array functions, then init, and then init array functions, the init have to be called from rtld now. To provide binary compatibility to old FreeBSD crt1, which calls _init itself, rtld only calls intializers and finalizers for main binary if binary has a note indicating that new crt was used for linking. Add parsing of ELF notes to rtld, and cache p_osrel value since we parsed it anyway. The patch is inspired by init_array support for DragonflyBSD, written by John Marino. Reviewed by: kan Tested by: andrew (arm, previous version), flo (sparc64, previous version) MFC after: 3 weeks
2012-03-11 20:03:09 +00:00
for (index = 0; index < elm->obj->init_array_num; index++) {
if (init_addr[index] != 0 && init_addr[index] != 1) {
dbg("calling init function for %s at %p", elm->obj->path,
(void *)init_addr[index]);
LD_UTRACE(UTRACE_INIT_CALL, elm->obj,
(void *)init_addr[index], 0, 0, elm->obj->path);
call_init_pointer(elm->obj, init_addr[index]);
}
}
}
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
wlock_acquire(rtld_bind_lock, lockstate);
rtld: Fix a race between dl_iterate_phdr() and dlclose(). Add a transient reference count to ensure that the phdr argument to the callback remains valid while the bind lock is dropped. Reviewed by: kib MFC after: 2 weeks Sponsored by: Dell EMC Isilon
2016-12-22 17:37:39 +00:00
unhold_object(elm->obj);
Revamp the code that calls shared libraries' init and fini functions. Formerly the init functions were called in the opposite of the order in which libraries were loaded, and libraries were loaded according to a breadth-first traversal of the dependency graph. That ordering came from SVR4.0, and it was easy to implement but not always sensible. Now we do a depth-first walk over the dependency graph and call the init functions in an order such that each shared object's needed objects are initialized before the shared object itself. At the same time we build a list of finalization (fini) functions in the opposite order, to guarantee correct C++ destructor ordering whenever possible. (It may not be possible if dlopen and dlclose are used in strange ways, but we come as close as one can come.) The need for this renovation has become apparent as more programs have started using multithreading. The multithreaded C library libc_r requires initialization, whereas the standard libc does not. Since virtually every other object depends on the C library, it is important that it get initialized first.
2000-07-26 04:24:40 +00:00
}
Fix a bug in which a program called dlclose from a destructor and got an assert failure in the dynamic linker.
2001-01-05 04:36:17 +00:00
errmsg_restore(saved_msg);
Revamp the code that calls shared libraries' init and fini functions. Formerly the init functions were called in the opposite of the order in which libraries were loaded, and libraries were loaded according to a breadth-first traversal of the dependency graph. That ordering came from SVR4.0, and it was easy to implement but not always sensible. Now we do a depth-first walk over the dependency graph and call the init functions in an order such that each shared object's needed objects are initialized before the shared object itself. At the same time we build a list of finalization (fini) functions in the opposite order, to guarantee correct C++ destructor ordering whenever possible. (It may not be possible if dlopen and dlclose are used in strange ways, but we come as close as one can come.) The need for this renovation has become apparent as more programs have started using multithreading. The multithreaded C library libc_r requires initialization, whereas the standard libc does not. Since virtually every other object depends on the C library, it is important that it get initialized first.
2000-07-26 04:24:40 +00:00
}
static void
objlist_clear(Objlist *list)
{
Objlist_Entry *elm;
while (!STAILQ_EMPTY(list)) {
elm = STAILQ_FIRST(list);
STAILQ_REMOVE_HEAD(list, link);
free(elm);
}
Revamp the symbol lookup algorithm to cope better with objects loaded separately by dlopen that have global symbols with identical names. Viewing each dlopened object as a DAG which is linked by its DT_NEEDED entries in the dynamic table, the search order is as follows: * If the referencing object was linked with -Bsymbolic, search it internally. * Search all dlopened DAGs containing the referencing object. * Search all objects loaded at program start up. * Search all objects which were dlopened() using the RTLD_GLOBAL flag (which is now supported too). The search terminates as soon as a strong definition is found. Lacking that, the first weak definition is used. These rules match those of Solaris, as best I could determine them from its vague manual pages and the results of experiments I performed. PR: misc/12438
1999-08-30 01:48:19 +00:00
}
static Objlist_Entry *
objlist_find(Objlist *list, const Obj_Entry *obj)
{
Objlist_Entry *elm;
STAILQ_FOREACH(elm, list, link)
if (elm->obj == obj)
return elm;
return NULL;
}
Revamp the code that calls shared libraries' init and fini functions. Formerly the init functions were called in the opposite of the order in which libraries were loaded, and libraries were loaded according to a breadth-first traversal of the dependency graph. That ordering came from SVR4.0, and it was easy to implement but not always sensible. Now we do a depth-first walk over the dependency graph and call the init functions in an order such that each shared object's needed objects are initialized before the shared object itself. At the same time we build a list of finalization (fini) functions in the opposite order, to guarantee correct C++ destructor ordering whenever possible. (It may not be possible if dlopen and dlclose are used in strange ways, but we come as close as one can come.) The need for this renovation has become apparent as more programs have started using multithreading. The multithreaded C library libc_r requires initialization, whereas the standard libc does not. Since virtually every other object depends on the C library, it is important that it get initialized first.
2000-07-26 04:24:40 +00:00
static void
objlist_init(Objlist *list)
{
STAILQ_INIT(list);
}
static void
objlist_push_head(Objlist *list, Obj_Entry *obj)
{
Objlist_Entry *elm;
elm = NEW(Objlist_Entry);
elm->obj = obj;
STAILQ_INSERT_HEAD(list, elm, link);
}
static void
objlist_push_tail(Objlist *list, Obj_Entry *obj)
{
Objlist_Entry *elm;
elm = NEW(Objlist_Entry);
elm->obj = obj;
STAILQ_INSERT_TAIL(list, elm, link);
}
Implement support for the interpose dso flag. Requested by: bf Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 1 week Approved by: re (glebius)
2013-10-07 08:19:30 +00:00
static void
objlist_put_after(Objlist *list, Obj_Entry *listobj, Obj_Entry *obj)
{
Objlist_Entry *elm, *listelm;
STAILQ_FOREACH(listelm, list, link) {
if (listelm->obj == listobj)
break;
}
elm = NEW(Objlist_Entry);
elm->obj = obj;
if (listelm != NULL)
STAILQ_INSERT_AFTER(list, listelm, elm, link);
else
STAILQ_INSERT_TAIL(list, elm, link);
}
Revamp the symbol lookup algorithm to cope better with objects loaded separately by dlopen that have global symbols with identical names. Viewing each dlopened object as a DAG which is linked by its DT_NEEDED entries in the dynamic table, the search order is as follows: * If the referencing object was linked with -Bsymbolic, search it internally. * Search all dlopened DAGs containing the referencing object. * Search all objects loaded at program start up. * Search all objects which were dlopened() using the RTLD_GLOBAL flag (which is now supported too). The search terminates as soon as a strong definition is found. Lacking that, the first weak definition is used. These rules match those of Solaris, as best I could determine them from its vague manual pages and the results of experiments I performed. PR: misc/12438
1999-08-30 01:48:19 +00:00
static void
objlist_remove(Objlist *list, Obj_Entry *obj)
{
Objlist_Entry *elm;
if ((elm = objlist_find(list, obj)) != NULL) {
Back out the previous change to the queue(3) interface. It was not discussed and should probably not happen. Requested by: msmith and others
2000-05-26 02:09:24 +00:00
STAILQ_REMOVE(list, elm, Struct_Objlist_Entry, link);
Revamp the symbol lookup algorithm to cope better with objects loaded separately by dlopen that have global symbols with identical names. Viewing each dlopened object as a DAG which is linked by its DT_NEEDED entries in the dynamic table, the search order is as follows: * If the referencing object was linked with -Bsymbolic, search it internally. * Search all dlopened DAGs containing the referencing object. * Search all objects loaded at program start up. * Search all objects which were dlopened() using the RTLD_GLOBAL flag (which is now supported too). The search terminates as soon as a strong definition is found. Lacking that, the first weak definition is used. These rules match those of Solaris, as best I could determine them from its vague manual pages and the results of experiments I performed. PR: misc/12438
1999-08-30 01:48:19 +00:00
free(elm);
}
}
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
/*
Ensure that for the object which is a dependency for some filtee, relocations are performed before the object's initializer is called. When dlopen()ing an object, relocate the whole DAG rooted in the object instead of only relocating the object itself and list of newly loaded dependencies. Reversed sequence currently can occur if the same object is a dependency for both filtee and filter, since filtees are loaded typically during the relocation processing, when some filter dependencies might be already loaded but not relocated yet. Reported and tested by: swills Reviewed by: kan MFC after: 1 week
2012-06-27 20:24:25 +00:00
* Relocate dag rooted in the specified object.
* Returns 0 on success, or -1 on failure.
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
*/
Ensure that for the object which is a dependency for some filtee, relocations are performed before the object's initializer is called. When dlopen()ing an object, relocate the whole DAG rooted in the object instead of only relocating the object itself and list of newly loaded dependencies. Reversed sequence currently can occur if the same object is a dependency for both filtee and filter, since filtees are loaded typically during the relocation processing, when some filter dependencies might be already loaded but not relocated yet. Reported and tested by: swills Reviewed by: kan MFC after: 1 week
2012-06-27 20:24:25 +00:00
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
static int
Ensure that for the object which is a dependency for some filtee, relocations are performed before the object's initializer is called. When dlopen()ing an object, relocate the whole DAG rooted in the object instead of only relocating the object itself and list of newly loaded dependencies. Reversed sequence currently can occur if the same object is a dependency for both filtee and filter, since filtees are loaded typically during the relocation processing, when some filter dependencies might be already loaded but not relocated yet. Reported and tested by: swills Reviewed by: kan MFC after: 1 week
2012-06-27 20:24:25 +00:00
relocate_object_dag(Obj_Entry *root, bool bind_now, Obj_Entry *rtldobj,
int flags, RtldLockState *lockstate)
{
Objlist_Entry *elm;
int error;
error = 0;
STAILQ_FOREACH(elm, &root->dagmembers, link) {
error = relocate_object(elm->obj, bind_now, rtldobj, flags,
lockstate);
if (error == -1)
break;
}
return (error);
}
Fix handling of DT_TEXTREL for an object with more than one read-only segment. According to gABI spec, presence of the tag indicates that dynamic linker must be prepared to handle relocations against any read-only segment, not only the segment which we, somewhat arbitrary, declared the text. For each read-only segment, add write permission before relocs are processed, and return to the mapping mode requested by the phdr, after relocs are done. Reported, tested, and reviewed by: emaste PR: 207631 Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2016-03-02 16:36:24 +00:00
/*
* Prepare for, or clean after, relocating an object marked with
* DT_TEXTREL or DF_TEXTREL. Before relocating, all read-only
* segments are remapped read-write. After relocations are done, the
* segment's permissions are returned back to the modes specified in
* the phdrs. If any relocation happened, or always for wired
* program, COW is triggered.
*/
static int
reloc_textrel_prot(Obj_Entry *obj, bool before)
{
const Elf_Phdr *ph;
void *base;
size_t l, sz;
int prot;
for (l = obj->phsize / sizeof(*ph), ph = obj->phdr; l > 0;
l--, ph++) {
if (ph->p_type != PT_LOAD || (ph->p_flags & PF_W) != 0)
continue;
base = obj->relocbase + trunc_page(ph->p_vaddr);
sz = round_page(ph->p_vaddr + ph->p_filesz) -
trunc_page(ph->p_vaddr);
prot = convert_prot(ph->p_flags) | (before ? PROT_WRITE : 0);
if (mprotect(base, sz, prot) == -1) {
_rtld_error("%s: Cannot write-%sable text segment: %s",
obj->path, before ? "en" : "dis",
rtld_strerror(errno));
return (-1);
}
}
return (0);
}
Ensure that for the object which is a dependency for some filtee, relocations are performed before the object's initializer is called. When dlopen()ing an object, relocate the whole DAG rooted in the object instead of only relocating the object itself and list of newly loaded dependencies. Reversed sequence currently can occur if the same object is a dependency for both filtee and filter, since filtees are loaded typically during the relocation processing, when some filter dependencies might be already loaded but not relocated yet. Reported and tested by: swills Reviewed by: kan MFC after: 1 week
2012-06-27 20:24:25 +00:00
/*
* Relocate single object.
* Returns 0 on success, or -1 on failure.
*/
static int
relocate_object(Obj_Entry *obj, bool bind_now, Obj_Entry *rtldobj,
Fix several problems with our ELF filters implementation. Do not relocate twice an object which happens to be needed by loaded binary (or dso) and some filtee opened due to symbol resolution when relocating need objects. Record the state of the relocation processing in Obj_Entry and short-circuit relocate_objects() if current object already processed. Do not call constructors for filtees loaded during the early relocation processing before image is initialized enough to run user-provided code. Filtees are loaded using dlopen_object(), which normally performs relocation and initialization. If filtee is lazy-loaded during the relocation of dso needed by the main object, dlopen_object() runs too earlier, when most runtime services are not yet ready. Postpone the constructors call to the time when main binary and depended libraries constructors are run, passing the new flag RTLD_LO_EARLY to dlopen_object(). Symbol lookups callers inform symlook_* functions about early stage of initialization with SYMLOOK_EARLY. Pass flags through all functions participating in object relocation. Use the opportunity and fix flags argument to find_symdef() in arch-specific reloc.c to use proper name SYMLOOK_IN_PLT instead of true, which happen to have the same numeric value. Reported and tested by: theraven Reviewed by: kan MFC after: 2 weeks
2012-03-20 13:20:49 +00:00
int flags, RtldLockState *lockstate)
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
{
Fix several problems with our ELF filters implementation. Do not relocate twice an object which happens to be needed by loaded binary (or dso) and some filtee opened due to symbol resolution when relocating need objects. Record the state of the relocation processing in Obj_Entry and short-circuit relocate_objects() if current object already processed. Do not call constructors for filtees loaded during the early relocation processing before image is initialized enough to run user-provided code. Filtees are loaded using dlopen_object(), which normally performs relocation and initialization. If filtee is lazy-loaded during the relocation of dso needed by the main object, dlopen_object() runs too earlier, when most runtime services are not yet ready. Postpone the constructors call to the time when main binary and depended libraries constructors are run, passing the new flag RTLD_LO_EARLY to dlopen_object(). Symbol lookups callers inform symlook_* functions about early stage of initialization with SYMLOOK_EARLY. Pass flags through all functions participating in object relocation. Use the opportunity and fix flags argument to find_symdef() in arch-specific reloc.c to use proper name SYMLOOK_IN_PLT instead of true, which happen to have the same numeric value. Reported and tested by: theraven Reviewed by: kan MFC after: 2 weeks
2012-03-20 13:20:49 +00:00
if (obj->relocated)
Ensure that for the object which is a dependency for some filtee, relocations are performed before the object's initializer is called. When dlopen()ing an object, relocate the whole DAG rooted in the object instead of only relocating the object itself and list of newly loaded dependencies. Reversed sequence currently can occur if the same object is a dependency for both filtee and filter, since filtees are loaded typically during the relocation processing, when some filter dependencies might be already loaded but not relocated yet. Reported and tested by: swills Reviewed by: kan MFC after: 1 week
2012-06-27 20:24:25 +00:00
return (0);
Fix several problems with our ELF filters implementation. Do not relocate twice an object which happens to be needed by loaded binary (or dso) and some filtee opened due to symbol resolution when relocating need objects. Record the state of the relocation processing in Obj_Entry and short-circuit relocate_objects() if current object already processed. Do not call constructors for filtees loaded during the early relocation processing before image is initialized enough to run user-provided code. Filtees are loaded using dlopen_object(), which normally performs relocation and initialization. If filtee is lazy-loaded during the relocation of dso needed by the main object, dlopen_object() runs too earlier, when most runtime services are not yet ready. Postpone the constructors call to the time when main binary and depended libraries constructors are run, passing the new flag RTLD_LO_EARLY to dlopen_object(). Symbol lookups callers inform symlook_* functions about early stage of initialization with SYMLOOK_EARLY. Pass flags through all functions participating in object relocation. Use the opportunity and fix flags argument to find_symdef() in arch-specific reloc.c to use proper name SYMLOOK_IN_PLT instead of true, which happen to have the same numeric value. Reported and tested by: theraven Reviewed by: kan MFC after: 2 weeks
2012-03-20 13:20:49 +00:00
obj->relocated = true;
Minor changes to make this work on sparc64. Approved by: jdp Tested on: alpha, i386, sparc64
2002-04-02 02:19:02 +00:00
if (obj != rtldobj)
Ensure that for the object which is a dependency for some filtee, relocations are performed before the object's initializer is called. When dlopen()ing an object, relocate the whole DAG rooted in the object instead of only relocating the object itself and list of newly loaded dependencies. Reversed sequence currently can occur if the same object is a dependency for both filtee and filter, since filtees are loaded typically during the relocation processing, when some filter dependencies might be already loaded but not relocated yet. Reported and tested by: swills Reviewed by: kan MFC after: 1 week
2012-06-27 20:24:25 +00:00
dbg("relocating \"%s\"", obj->path);
Fix several problems with our ELF filters implementation. Do not relocate twice an object which happens to be needed by loaded binary (or dso) and some filtee opened due to symbol resolution when relocating need objects. Record the state of the relocation processing in Obj_Entry and short-circuit relocate_objects() if current object already processed. Do not call constructors for filtees loaded during the early relocation processing before image is initialized enough to run user-provided code. Filtees are loaded using dlopen_object(), which normally performs relocation and initialization. If filtee is lazy-loaded during the relocation of dso needed by the main object, dlopen_object() runs too earlier, when most runtime services are not yet ready. Postpone the constructors call to the time when main binary and depended libraries constructors are run, passing the new flag RTLD_LO_EARLY to dlopen_object(). Symbol lookups callers inform symlook_* functions about early stage of initialization with SYMLOOK_EARLY. Pass flags through all functions participating in object relocation. Use the opportunity and fix flags argument to find_symdef() in arch-specific reloc.c to use proper name SYMLOOK_IN_PLT instead of true, which happen to have the same numeric value. Reported and tested by: theraven Reviewed by: kan MFC after: 2 weeks
2012-03-20 13:20:49 +00:00
Add GNU hash support for rtld. Based on dragonflybsd support for GNU hash by John Marino <draco marino st> Reviewed by: kan Tested by: bapt MFC after: 2 weeks
2012-04-30 13:31:10 +00:00
if (obj->symtab == NULL || obj->strtab == NULL ||
Ensure that for the object which is a dependency for some filtee, relocations are performed before the object's initializer is called. When dlopen()ing an object, relocate the whole DAG rooted in the object instead of only relocating the object itself and list of newly loaded dependencies. Reversed sequence currently can occur if the same object is a dependency for both filtee and filter, since filtees are loaded typically during the relocation processing, when some filter dependencies might be already loaded but not relocated yet. Reported and tested by: swills Reviewed by: kan MFC after: 1 week
2012-06-27 20:24:25 +00:00
!(obj->valid_hash_sysv || obj->valid_hash_gnu)) {
_rtld_error("%s: Shared object has no run-time symbol table",
obj->path);
return (-1);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
}
Fix handling of DT_TEXTREL for an object with more than one read-only segment. According to gABI spec, presence of the tag indicates that dynamic linker must be prepared to handle relocations against any read-only segment, not only the segment which we, somewhat arbitrary, declared the text. For each read-only segment, add write permission before relocs are processed, and return to the mapping mode requested by the phdr, after relocs are done. Reported, tested, and reviewed by: emaste PR: 207631 Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2016-03-02 16:36:24 +00:00
/* There are relocations to the write-protected text segment. */
if (obj->textrel && reloc_textrel_prot(obj, true) != 0)
return (-1);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
IFUNC symbol type shall be processed for non-PLT relocations, e.g. when a global variable is initialized with a pointer to ifunc. Add symbol type check and call resolver for STT_GNU_IFUNC symbol types when processing non-PLT relocations, but only after non-IFUNC relocations are done. The two-phase proceessing is required since resolvers may reference other symbols, which must be ready to use when resolver calls are done. Restructure reloc_non_plt() on x86 to call find_symdef() and handle IFUNC in single place. For non-x86 reloc_non_plt(), check for call for IFUNC relocation and do nothing, to avoid processing relocs twice. PR: 193048 Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2014-08-29 09:29:10 +00:00
/* Process the non-PLT non-IFUNC relocations. */
Fix several problems with our ELF filters implementation. Do not relocate twice an object which happens to be needed by loaded binary (or dso) and some filtee opened due to symbol resolution when relocating need objects. Record the state of the relocation processing in Obj_Entry and short-circuit relocate_objects() if current object already processed. Do not call constructors for filtees loaded during the early relocation processing before image is initialized enough to run user-provided code. Filtees are loaded using dlopen_object(), which normally performs relocation and initialization. If filtee is lazy-loaded during the relocation of dso needed by the main object, dlopen_object() runs too earlier, when most runtime services are not yet ready. Postpone the constructors call to the time when main binary and depended libraries constructors are run, passing the new flag RTLD_LO_EARLY to dlopen_object(). Symbol lookups callers inform symlook_* functions about early stage of initialization with SYMLOOK_EARLY. Pass flags through all functions participating in object relocation. Use the opportunity and fix flags argument to find_symdef() in arch-specific reloc.c to use proper name SYMLOOK_IN_PLT instead of true, which happen to have the same numeric value. Reported and tested by: theraven Reviewed by: kan MFC after: 2 weeks
2012-03-20 13:20:49 +00:00
if (reloc_non_plt(obj, rtldobj, flags, lockstate))
Ensure that for the object which is a dependency for some filtee, relocations are performed before the object's initializer is called. When dlopen()ing an object, relocate the whole DAG rooted in the object instead of only relocating the object itself and list of newly loaded dependencies. Reversed sequence currently can occur if the same object is a dependency for both filtee and filter, since filtees are loaded typically during the relocation processing, when some filter dependencies might be already loaded but not relocated yet. Reported and tested by: swills Reviewed by: kan MFC after: 1 week
2012-06-27 20:24:25 +00:00
return (-1);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
Fix handling of DT_TEXTREL for an object with more than one read-only segment. According to gABI spec, presence of the tag indicates that dynamic linker must be prepared to handle relocations against any read-only segment, not only the segment which we, somewhat arbitrary, declared the text. For each read-only segment, add write permission before relocs are processed, and return to the mapping mode requested by the phdr, after relocs are done. Reported, tested, and reviewed by: emaste PR: 207631 Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2016-03-02 16:36:24 +00:00
/* Re-protected the text segment. */
if (obj->textrel && reloc_textrel_prot(obj, false) != 0)
return (-1);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
Add support for STT_GNU_IFUNC and R_MACHINE_IRELATIVE GNU extensions to rtld on 386 and amd64. This adds runtime bits neccessary for the use of the dispatch functions from the dynamically-linked executables and shared libraries. To allow use of external references from the dispatch function, resolution of the R_MACHINE_IRESOLVE relocations in PLT is postponed until GOT entries for PLT are prepared, and normal resolution of the GOT entries is finished. Similar to how it is done by GNU, IRELATIVE relocations are resolved in advance, instead of normal lazy handling for PLT. Move the init_pltgot() call before the relocations for the object are processed. MFC after: 3 weeks
2011-12-12 11:03:14 +00:00
/* Set the special PLT or GOT entries. */
init_pltgot(obj);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
/* Process the PLT relocations. */
When a threads package registers locking methods with dllockinit(), figure out which shared object(s) contain the the locking methods and fully bind those objects as if they had been loaded with LD_BIND_NOW=1. The goal is to keep the locking methods from requiring any lazy binding. Otherwise infinite recursion occurs in _rtld_bind. This fixes the infinite recursion problem in the linuxthreads port.
2000-01-29 01:27:04 +00:00
if (reloc_plt(obj) == -1)
Ensure that for the object which is a dependency for some filtee, relocations are performed before the object's initializer is called. When dlopen()ing an object, relocate the whole DAG rooted in the object instead of only relocating the object itself and list of newly loaded dependencies. Reversed sequence currently can occur if the same object is a dependency for both filtee and filter, since filtees are loaded typically during the relocation processing, when some filter dependencies might be already loaded but not relocated yet. Reported and tested by: swills Reviewed by: kan MFC after: 1 week
2012-06-27 20:24:25 +00:00
return (-1);
When a threads package registers locking methods with dllockinit(), figure out which shared object(s) contain the the locking methods and fully bind those objects as if they had been loaded with LD_BIND_NOW=1. The goal is to keep the locking methods from requiring any lazy binding. Otherwise infinite recursion occurs in _rtld_bind. This fixes the infinite recursion problem in the linuxthreads port.
2000-01-29 01:27:04 +00:00
/* Relocate the jump slots if we are doing immediate binding. */
- Add support for DT_FLAGS. - Define various things from the most recent ELF spec.
2003-06-18 03:34:29 +00:00
if (obj->bind_now || bind_now)
Ensure that for the object which is a dependency for some filtee, relocations are performed before the object's initializer is called. When dlopen()ing an object, relocate the whole DAG rooted in the object instead of only relocating the object itself and list of newly loaded dependencies. Reversed sequence currently can occur if the same object is a dependency for both filtee and filter, since filtees are loaded typically during the relocation processing, when some filter dependencies might be already loaded but not relocated yet. Reported and tested by: swills Reviewed by: kan MFC after: 1 week
2012-06-27 20:24:25 +00:00
if (reloc_jmpslots(obj, flags, lockstate) == -1)
return (-1);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
IFUNC symbol type shall be processed for non-PLT relocations, e.g. when a global variable is initialized with a pointer to ifunc. Add symbol type check and call resolver for STT_GNU_IFUNC symbol types when processing non-PLT relocations, but only after non-IFUNC relocations are done. The two-phase proceessing is required since resolvers may reference other symbols, which must be ready to use when resolver calls are done. Restructure reloc_non_plt() on x86 to call find_symdef() and handle IFUNC in single place. For non-x86 reloc_non_plt(), check for call for IFUNC relocation and do nothing, to avoid processing relocs twice. PR: 193048 Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2014-08-29 09:29:10 +00:00
/*
* Process the non-PLT IFUNC relocations. The relocations are
* processed in two phases, because IFUNC resolvers may
* reference other symbols, which must be readily processed
* before resolvers are called.
*/
Optimize r270798, only do the second pass over non-plt relocations when the first pass found IFUNCs. Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2014-08-29 10:43:56 +00:00
if (obj->non_plt_gnu_ifunc &&
reloc_non_plt(obj, rtldobj, flags | SYMLOOK_IFUNC, lockstate))
IFUNC symbol type shall be processed for non-PLT relocations, e.g. when a global variable is initialized with a pointer to ifunc. Add symbol type check and call resolver for STT_GNU_IFUNC symbol types when processing non-PLT relocations, but only after non-IFUNC relocations are done. The two-phase proceessing is required since resolvers may reference other symbols, which must be ready to use when resolver calls are done. Restructure reloc_non_plt() on x86 to call find_symdef() and handle IFUNC in single place. For non-x86 reloc_non_plt(), check for call for IFUNC relocation and do nothing, to avoid processing relocs twice. PR: 193048 Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2014-08-29 09:29:10 +00:00
return (-1);
For the main binary, postpone enforcing relro read-only protection until copy relocations are done. Newer binutils and lld seems to output copy into relro-protected range. Reported by: Rafael Espц╜ndola via emaste Sponsored by: The FreeBSD Foundation MFC after: 1 week
2017-01-12 15:54:03 +00:00
if (!obj->mainprog && obj_enforce_relro(obj) == -1)
return (-1);
Add support for GNU RELRO. Submitted by: John Marino <draco marino st> MFC after: 2 weeks
2012-01-30 19:52:17 +00:00
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
/*
* Set up the magic number and version in the Obj_Entry. These
* were checked in the crt1.o from the original ElfKit, so we
* set them for backward compatibility.
*/
obj->magic = RTLD_MAGIC;
obj->version = RTLD_VERSION;
Ensure that for the object which is a dependency for some filtee, relocations are performed before the object's initializer is called. When dlopen()ing an object, relocate the whole DAG rooted in the object instead of only relocating the object itself and list of newly loaded dependencies. Reversed sequence currently can occur if the same object is a dependency for both filtee and filter, since filtees are loaded typically during the relocation processing, when some filter dependencies might be already loaded but not relocated yet. Reported and tested by: swills Reviewed by: kan MFC after: 1 week
2012-06-27 20:24:25 +00:00
return (0);
}
/*
* Relocate newly-loaded shared objects. The argument is a pointer to
* the Obj_Entry for the first such object. All objects from the first
* to the end of the list of objects are relocated. Returns 0 on success,
* or -1 on failure.
*/
static int
relocate_objects(Obj_Entry *first, bool bind_now, Obj_Entry *rtldobj,
int flags, RtldLockState *lockstate)
{
Obj_Entry *obj;
int error;
Remove all remaining uses of TAILQ_FOREACH_FROM() from rtld-elf. Sponsored by: The FreeBSD Foundation MFC after: 1 week
2016-08-12 18:29:11 +00:00
for (error = 0, obj = first; obj != NULL;
obj = TAILQ_NEXT(obj, next)) {
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
if (obj->marker)
continue;
Ensure that for the object which is a dependency for some filtee, relocations are performed before the object's initializer is called. When dlopen()ing an object, relocate the whole DAG rooted in the object instead of only relocating the object itself and list of newly loaded dependencies. Reversed sequence currently can occur if the same object is a dependency for both filtee and filter, since filtees are loaded typically during the relocation processing, when some filter dependencies might be already loaded but not relocated yet. Reported and tested by: swills Reviewed by: kan MFC after: 1 week
2012-06-27 20:24:25 +00:00
error = relocate_object(obj, bind_now, rtldobj, flags,
lockstate);
if (error == -1)
break;
}
return (error);
_rtld_bind() read-locks the bind lock, and possible plt resolution from the dispatcher would also acquire bind lock in read mode, which is the supported operation. plt is explicitely designed to allow safe multithreaded updates, so the shared lock do not cause problems. The error in r228435 is that it allows read lock acquisition after the write lock for the bind block. If we dlopened the shared object that contains IRELATIVE or jump slot which target is STT_GNU_IFUNC, then possible recursive plt resolve from the dispatcher would cause it. Postpone the resolution for irelative/ifunc right before initializers are called, and drop bind lock around calls to dispatcher. Use initlist to iterate over the objects instead of the ->next, due to drop of the bind lock in iteration. For i386/reloc.c:reloc_iresolve(), fix calculation of the dispatch function address for dso, by taking into account possible non-zero relocbase. MFC after: 3 weeks
2011-12-14 16:47:53 +00:00
}
/*
* The handling of R_MACHINE_IRELATIVE relocations and jumpslots
* referencing STT_GNU_IFUNC symbols is postponed till the other
* relocations are done. The indirect functions specified as
* ifunc are allowed to call other symbols, so we need to have
* objects relocated before asking for resolution from indirects.
*
* The R_MACHINE_IRELATIVE slots are resolved in greedy fashion,
* instead of the usual lazy handling of PLT slots. It is
* consistent with how GNU does it.
*/
static int
Fix several problems with our ELF filters implementation. Do not relocate twice an object which happens to be needed by loaded binary (or dso) and some filtee opened due to symbol resolution when relocating need objects. Record the state of the relocation processing in Obj_Entry and short-circuit relocate_objects() if current object already processed. Do not call constructors for filtees loaded during the early relocation processing before image is initialized enough to run user-provided code. Filtees are loaded using dlopen_object(), which normally performs relocation and initialization. If filtee is lazy-loaded during the relocation of dso needed by the main object, dlopen_object() runs too earlier, when most runtime services are not yet ready. Postpone the constructors call to the time when main binary and depended libraries constructors are run, passing the new flag RTLD_LO_EARLY to dlopen_object(). Symbol lookups callers inform symlook_* functions about early stage of initialization with SYMLOOK_EARLY. Pass flags through all functions participating in object relocation. Use the opportunity and fix flags argument to find_symdef() in arch-specific reloc.c to use proper name SYMLOOK_IN_PLT instead of true, which happen to have the same numeric value. Reported and tested by: theraven Reviewed by: kan MFC after: 2 weeks
2012-03-20 13:20:49 +00:00
resolve_object_ifunc(Obj_Entry *obj, bool bind_now, int flags,
RtldLockState *lockstate)
_rtld_bind() read-locks the bind lock, and possible plt resolution from the dispatcher would also acquire bind lock in read mode, which is the supported operation. plt is explicitely designed to allow safe multithreaded updates, so the shared lock do not cause problems. The error in r228435 is that it allows read lock acquisition after the write lock for the bind block. If we dlopened the shared object that contains IRELATIVE or jump slot which target is STT_GNU_IFUNC, then possible recursive plt resolve from the dispatcher would cause it. Postpone the resolution for irelative/ifunc right before initializers are called, and drop bind lock around calls to dispatcher. Use initlist to iterate over the objects instead of the ->next, due to drop of the bind lock in iteration. For i386/reloc.c:reloc_iresolve(), fix calculation of the dispatch function address for dso, by taking into account possible non-zero relocbase. MFC after: 3 weeks
2011-12-14 16:47:53 +00:00
{
Add support for STT_GNU_IFUNC and R_MACHINE_IRELATIVE GNU extensions to rtld on 386 and amd64. This adds runtime bits neccessary for the use of the dispatch functions from the dynamically-linked executables and shared libraries. To allow use of external references from the dispatch function, resolution of the R_MACHINE_IRESOLVE relocations in PLT is postponed until GOT entries for PLT are prepared, and normal resolution of the GOT entries is finished. Similar to how it is done by GNU, IRELATIVE relocations are resolved in advance, instead of normal lazy handling for PLT. Move the init_pltgot() call before the relocations for the object are processed. MFC after: 3 weeks
2011-12-12 11:03:14 +00:00
if (obj->irelative && reloc_iresolve(obj, lockstate) == -1)
_rtld_bind() read-locks the bind lock, and possible plt resolution from the dispatcher would also acquire bind lock in read mode, which is the supported operation. plt is explicitely designed to allow safe multithreaded updates, so the shared lock do not cause problems. The error in r228435 is that it allows read lock acquisition after the write lock for the bind block. If we dlopened the shared object that contains IRELATIVE or jump slot which target is STT_GNU_IFUNC, then possible recursive plt resolve from the dispatcher would cause it. Postpone the resolution for irelative/ifunc right before initializers are called, and drop bind lock around calls to dispatcher. Use initlist to iterate over the objects instead of the ->next, due to drop of the bind lock in iteration. For i386/reloc.c:reloc_iresolve(), fix calculation of the dispatch function address for dso, by taking into account possible non-zero relocbase. MFC after: 3 weeks
2011-12-14 16:47:53 +00:00
return (-1);
Add support for STT_GNU_IFUNC and R_MACHINE_IRELATIVE GNU extensions to rtld on 386 and amd64. This adds runtime bits neccessary for the use of the dispatch functions from the dynamically-linked executables and shared libraries. To allow use of external references from the dispatch function, resolution of the R_MACHINE_IRESOLVE relocations in PLT is postponed until GOT entries for PLT are prepared, and normal resolution of the GOT entries is finished. Similar to how it is done by GNU, IRELATIVE relocations are resolved in advance, instead of normal lazy handling for PLT. Move the init_pltgot() call before the relocations for the object are processed. MFC after: 3 weeks
2011-12-12 11:03:14 +00:00
if ((obj->bind_now || bind_now) && obj->gnu_ifunc &&
Fix several problems with our ELF filters implementation. Do not relocate twice an object which happens to be needed by loaded binary (or dso) and some filtee opened due to symbol resolution when relocating need objects. Record the state of the relocation processing in Obj_Entry and short-circuit relocate_objects() if current object already processed. Do not call constructors for filtees loaded during the early relocation processing before image is initialized enough to run user-provided code. Filtees are loaded using dlopen_object(), which normally performs relocation and initialization. If filtee is lazy-loaded during the relocation of dso needed by the main object, dlopen_object() runs too earlier, when most runtime services are not yet ready. Postpone the constructors call to the time when main binary and depended libraries constructors are run, passing the new flag RTLD_LO_EARLY to dlopen_object(). Symbol lookups callers inform symlook_* functions about early stage of initialization with SYMLOOK_EARLY. Pass flags through all functions participating in object relocation. Use the opportunity and fix flags argument to find_symdef() in arch-specific reloc.c to use proper name SYMLOOK_IN_PLT instead of true, which happen to have the same numeric value. Reported and tested by: theraven Reviewed by: kan MFC after: 2 weeks
2012-03-20 13:20:49 +00:00
reloc_gnu_ifunc(obj, flags, lockstate) == -1)
_rtld_bind() read-locks the bind lock, and possible plt resolution from the dispatcher would also acquire bind lock in read mode, which is the supported operation. plt is explicitely designed to allow safe multithreaded updates, so the shared lock do not cause problems. The error in r228435 is that it allows read lock acquisition after the write lock for the bind block. If we dlopened the shared object that contains IRELATIVE or jump slot which target is STT_GNU_IFUNC, then possible recursive plt resolve from the dispatcher would cause it. Postpone the resolution for irelative/ifunc right before initializers are called, and drop bind lock around calls to dispatcher. Use initlist to iterate over the objects instead of the ->next, due to drop of the bind lock in iteration. For i386/reloc.c:reloc_iresolve(), fix calculation of the dispatch function address for dso, by taking into account possible non-zero relocbase. MFC after: 3 weeks
2011-12-14 16:47:53 +00:00
return (-1);
return (0);
}
static int
Fix several problems with our ELF filters implementation. Do not relocate twice an object which happens to be needed by loaded binary (or dso) and some filtee opened due to symbol resolution when relocating need objects. Record the state of the relocation processing in Obj_Entry and short-circuit relocate_objects() if current object already processed. Do not call constructors for filtees loaded during the early relocation processing before image is initialized enough to run user-provided code. Filtees are loaded using dlopen_object(), which normally performs relocation and initialization. If filtee is lazy-loaded during the relocation of dso needed by the main object, dlopen_object() runs too earlier, when most runtime services are not yet ready. Postpone the constructors call to the time when main binary and depended libraries constructors are run, passing the new flag RTLD_LO_EARLY to dlopen_object(). Symbol lookups callers inform symlook_* functions about early stage of initialization with SYMLOOK_EARLY. Pass flags through all functions participating in object relocation. Use the opportunity and fix flags argument to find_symdef() in arch-specific reloc.c to use proper name SYMLOOK_IN_PLT instead of true, which happen to have the same numeric value. Reported and tested by: theraven Reviewed by: kan MFC after: 2 weeks
2012-03-20 13:20:49 +00:00
resolve_objects_ifunc(Obj_Entry *first, bool bind_now, int flags,
RtldLockState *lockstate)
_rtld_bind() read-locks the bind lock, and possible plt resolution from the dispatcher would also acquire bind lock in read mode, which is the supported operation. plt is explicitely designed to allow safe multithreaded updates, so the shared lock do not cause problems. The error in r228435 is that it allows read lock acquisition after the write lock for the bind block. If we dlopened the shared object that contains IRELATIVE or jump slot which target is STT_GNU_IFUNC, then possible recursive plt resolve from the dispatcher would cause it. Postpone the resolution for irelative/ifunc right before initializers are called, and drop bind lock around calls to dispatcher. Use initlist to iterate over the objects instead of the ->next, due to drop of the bind lock in iteration. For i386/reloc.c:reloc_iresolve(), fix calculation of the dispatch function address for dso, by taking into account possible non-zero relocbase. MFC after: 3 weeks
2011-12-14 16:47:53 +00:00
{
Obj_Entry *obj;
Remove all remaining uses of TAILQ_FOREACH_FROM() from rtld-elf. Sponsored by: The FreeBSD Foundation MFC after: 1 week
2016-08-12 18:29:11 +00:00
for (obj = first; obj != NULL; obj = TAILQ_NEXT(obj, next)) {
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
if (obj->marker)
continue;
Fix several problems with our ELF filters implementation. Do not relocate twice an object which happens to be needed by loaded binary (or dso) and some filtee opened due to symbol resolution when relocating need objects. Record the state of the relocation processing in Obj_Entry and short-circuit relocate_objects() if current object already processed. Do not call constructors for filtees loaded during the early relocation processing before image is initialized enough to run user-provided code. Filtees are loaded using dlopen_object(), which normally performs relocation and initialization. If filtee is lazy-loaded during the relocation of dso needed by the main object, dlopen_object() runs too earlier, when most runtime services are not yet ready. Postpone the constructors call to the time when main binary and depended libraries constructors are run, passing the new flag RTLD_LO_EARLY to dlopen_object(). Symbol lookups callers inform symlook_* functions about early stage of initialization with SYMLOOK_EARLY. Pass flags through all functions participating in object relocation. Use the opportunity and fix flags argument to find_symdef() in arch-specific reloc.c to use proper name SYMLOOK_IN_PLT instead of true, which happen to have the same numeric value. Reported and tested by: theraven Reviewed by: kan MFC after: 2 weeks
2012-03-20 13:20:49 +00:00
if (resolve_object_ifunc(obj, bind_now, flags, lockstate) == -1)
_rtld_bind() read-locks the bind lock, and possible plt resolution from the dispatcher would also acquire bind lock in read mode, which is the supported operation. plt is explicitely designed to allow safe multithreaded updates, so the shared lock do not cause problems. The error in r228435 is that it allows read lock acquisition after the write lock for the bind block. If we dlopened the shared object that contains IRELATIVE or jump slot which target is STT_GNU_IFUNC, then possible recursive plt resolve from the dispatcher would cause it. Postpone the resolution for irelative/ifunc right before initializers are called, and drop bind lock around calls to dispatcher. Use initlist to iterate over the objects instead of the ->next, due to drop of the bind lock in iteration. For i386/reloc.c:reloc_iresolve(), fix calculation of the dispatch function address for dso, by taking into account possible non-zero relocbase. MFC after: 3 weeks
2011-12-14 16:47:53 +00:00
return (-1);
}
return (0);
}
static int
Fix several problems with our ELF filters implementation. Do not relocate twice an object which happens to be needed by loaded binary (or dso) and some filtee opened due to symbol resolution when relocating need objects. Record the state of the relocation processing in Obj_Entry and short-circuit relocate_objects() if current object already processed. Do not call constructors for filtees loaded during the early relocation processing before image is initialized enough to run user-provided code. Filtees are loaded using dlopen_object(), which normally performs relocation and initialization. If filtee is lazy-loaded during the relocation of dso needed by the main object, dlopen_object() runs too earlier, when most runtime services are not yet ready. Postpone the constructors call to the time when main binary and depended libraries constructors are run, passing the new flag RTLD_LO_EARLY to dlopen_object(). Symbol lookups callers inform symlook_* functions about early stage of initialization with SYMLOOK_EARLY. Pass flags through all functions participating in object relocation. Use the opportunity and fix flags argument to find_symdef() in arch-specific reloc.c to use proper name SYMLOOK_IN_PLT instead of true, which happen to have the same numeric value. Reported and tested by: theraven Reviewed by: kan MFC after: 2 weeks
2012-03-20 13:20:49 +00:00
initlist_objects_ifunc(Objlist *list, bool bind_now, int flags,
RtldLockState *lockstate)
_rtld_bind() read-locks the bind lock, and possible plt resolution from the dispatcher would also acquire bind lock in read mode, which is the supported operation. plt is explicitely designed to allow safe multithreaded updates, so the shared lock do not cause problems. The error in r228435 is that it allows read lock acquisition after the write lock for the bind block. If we dlopened the shared object that contains IRELATIVE or jump slot which target is STT_GNU_IFUNC, then possible recursive plt resolve from the dispatcher would cause it. Postpone the resolution for irelative/ifunc right before initializers are called, and drop bind lock around calls to dispatcher. Use initlist to iterate over the objects instead of the ->next, due to drop of the bind lock in iteration. For i386/reloc.c:reloc_iresolve(), fix calculation of the dispatch function address for dso, by taking into account possible non-zero relocbase. MFC after: 3 weeks
2011-12-14 16:47:53 +00:00
{
Objlist_Entry *elm;
STAILQ_FOREACH(elm, list, link) {
Fix several problems with our ELF filters implementation. Do not relocate twice an object which happens to be needed by loaded binary (or dso) and some filtee opened due to symbol resolution when relocating need objects. Record the state of the relocation processing in Obj_Entry and short-circuit relocate_objects() if current object already processed. Do not call constructors for filtees loaded during the early relocation processing before image is initialized enough to run user-provided code. Filtees are loaded using dlopen_object(), which normally performs relocation and initialization. If filtee is lazy-loaded during the relocation of dso needed by the main object, dlopen_object() runs too earlier, when most runtime services are not yet ready. Postpone the constructors call to the time when main binary and depended libraries constructors are run, passing the new flag RTLD_LO_EARLY to dlopen_object(). Symbol lookups callers inform symlook_* functions about early stage of initialization with SYMLOOK_EARLY. Pass flags through all functions participating in object relocation. Use the opportunity and fix flags argument to find_symdef() in arch-specific reloc.c to use proper name SYMLOOK_IN_PLT instead of true, which happen to have the same numeric value. Reported and tested by: theraven Reviewed by: kan MFC after: 2 weeks
2012-03-20 13:20:49 +00:00
if (resolve_object_ifunc(elm->obj, bind_now, flags,
lockstate) == -1)
_rtld_bind() read-locks the bind lock, and possible plt resolution from the dispatcher would also acquire bind lock in read mode, which is the supported operation. plt is explicitely designed to allow safe multithreaded updates, so the shared lock do not cause problems. The error in r228435 is that it allows read lock acquisition after the write lock for the bind block. If we dlopened the shared object that contains IRELATIVE or jump slot which target is STT_GNU_IFUNC, then possible recursive plt resolve from the dispatcher would cause it. Postpone the resolution for irelative/ifunc right before initializers are called, and drop bind lock around calls to dispatcher. Use initlist to iterate over the objects instead of the ->next, due to drop of the bind lock in iteration. For i386/reloc.c:reloc_iresolve(), fix calculation of the dispatch function address for dso, by taking into account possible non-zero relocbase. MFC after: 3 weeks
2011-12-14 16:47:53 +00:00
return (-1);
}
return (0);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
}
/*
* Cleanup procedure. It will be called (by the atexit mechanism) just
* before the process exits.
*/
static void
rtld_exit(void)
{
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
RtldLockState lockstate;
Revamp the mechanism for enumerating and calling shared objects' init and fini functions. Now the code is very careful to hold no locks when calling these functions. Thus the dynamic linker cannot be re-entered with a lock already held. Remove the tolerance for recursive locking that I added in revision 1.2 of dllockinit.c. Recursive locking shouldn't happen any more. Mozilla and JDK users: I'd appreciate confirmation that things still work right (or at least the same) with these changes.
2000-01-09 21:13:48 +00:00
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
wlock_acquire(rtld_bind_lock, &lockstate);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
dbg("rtld_exit()");
If dlclose() is called recursively from a _fini() function, the inner dlclose() call may unload the object of the outer call prematurely because objects are unreferenced before _fini() calls. Fix this by unreferencing objects after calling objlist_call_fini() in dlclose(). Therefore objlist_call_fini() now calls the fini function if the reference count of an object is 1. In addition we must restart the list_fini traversal after every _fini() call because another dlclose() call might have modified the reference counts. Add an XXX comment to objlist_call_fini() about possible race with dlopen(). PR: 133246, 149464 Reviewed by: kan, kib
2010-12-16 16:56:44 +00:00
objlist_call_fini(&list_fini, NULL, &lockstate);
Revamp the code that calls shared libraries' init and fini functions. Formerly the init functions were called in the opposite of the order in which libraries were loaded, and libraries were loaded according to a breadth-first traversal of the dependency graph. That ordering came from SVR4.0, and it was easy to implement but not always sensible. Now we do a depth-first walk over the dependency graph and call the init functions in an order such that each shared object's needed objects are initialized before the shared object itself. At the same time we build a list of finalization (fini) functions in the opposite order, to guarantee correct C++ destructor ordering whenever possible. (It may not be possible if dlopen and dlclose are used in strange ways, but we come as close as one can come.) The need for this renovation has become apparent as more programs have started using multithreading. The multithreaded C library libc_r requires initialization, whereas the standard libc does not. Since virtually every other object depends on the C library, it is important that it get initialized first.
2000-07-26 04:24:40 +00:00
/* No need to remove the items from the list, since we are exiting. */
Use the environment variable LD_LIBMAP_DISABLE to disable libmap.conf(5) functionality.
2003-05-31 14:45:11 +00:00
if (!libmap_disable)
lm_fini();
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
lock_release(rtld_bind_lock, &lockstate);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
}
Make the directory mapping functionality, which was previously only available in 32-bit compatibility mode, unconditional. Overhaul the man page, which had evolved more by accretion than by design. Approved by: re (gjb) MFC after: 3 weeks
2013-09-21 21:03:52 +00:00
/*
* Iterate over a search path, translate each element, and invoke the
* callback on the result.
*/
Implement dlinfo() function. Introdice RTLD_SELF special handle and properly process it within dlsym() and dlinfo() functions. The intention is to improve our compatibility with Solaris and to make a Java port easier. Partially submitted by: phantom
2003-02-13 17:47:44 +00:00
static void *
path_enumerate(const char *path, path_enum_proc callback, void *arg)
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
{
Add initial support for compiling a special 32 bit version of ld-elf.so.1 on 64 bit systems. Most of this involves using alternate paths, environment variables and diagnostic messages. The build glue is seperate.
2004-03-21 01:21:26 +00:00
const char *trans;
Implement dlinfo() function. Introdice RTLD_SELF special handle and properly process it within dlsym() and dlinfo() functions. The intention is to improve our compatibility with Solaris and to make a Java port easier. Partially submitted by: phantom
2003-02-13 17:47:44 +00:00
if (path == NULL)
return (NULL);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
Implement dlinfo() function. Introdice RTLD_SELF special handle and properly process it within dlsym() and dlinfo() functions. The intention is to improve our compatibility with Solaris and to make a Java port easier. Partially submitted by: phantom
2003-02-13 17:47:44 +00:00
path += strspn(path, ":;");
while (*path != '\0') {
size_t len;
char *res;
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
Implement dlinfo() function. Introdice RTLD_SELF special handle and properly process it within dlsym() and dlinfo() functions. The intention is to improve our compatibility with Solaris and to make a Java port easier. Partially submitted by: phantom
2003-02-13 17:47:44 +00:00
len = strcspn(path, ":;");
Add initial support for compiling a special 32 bit version of ld-elf.so.1 on 64 bit systems. Most of this involves using alternate paths, environment variables and diagnostic messages. The build glue is seperate.
2004-03-21 01:21:26 +00:00
trans = lm_findn(NULL, path, len);
if (trans)
res = callback(trans, strlen(trans), arg);
else
Make the directory mapping functionality, which was previously only available in 32-bit compatibility mode, unconditional. Overhaul the man page, which had evolved more by accretion than by design. Approved by: re (gjb) MFC after: 3 weeks
2013-09-21 21:03:52 +00:00
res = callback(path, len, arg);
Implement dlinfo() function. Introdice RTLD_SELF special handle and properly process it within dlsym() and dlinfo() functions. The intention is to improve our compatibility with Solaris and to make a Java port easier. Partially submitted by: phantom
2003-02-13 17:47:44 +00:00
if (res != NULL)
return (res);
path += len;
path += strspn(path, ":;");
}
return (NULL);
}
struct try_library_args {
const char *name;
size_t namelen;
char *buffer;
size_t buflen;
};
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
Implement dlinfo() function. Introdice RTLD_SELF special handle and properly process it within dlsym() and dlinfo() functions. The intention is to improve our compatibility with Solaris and to make a Java port easier. Partially submitted by: phantom
2003-02-13 17:47:44 +00:00
static void *
try_library_path(const char *dir, size_t dirlen, void *param)
{
struct try_library_args *arg;
arg = param;
if (*dir == '/' || trust) {
char *pathname;
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
Implement dlinfo() function. Introdice RTLD_SELF special handle and properly process it within dlsym() and dlinfo() functions. The intention is to improve our compatibility with Solaris and to make a Java port easier. Partially submitted by: phantom
2003-02-13 17:47:44 +00:00
if (dirlen + 1 + arg->namelen + 1 > arg->buflen)
return (NULL);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
Implement dlinfo() function. Introdice RTLD_SELF special handle and properly process it within dlsym() and dlinfo() functions. The intention is to improve our compatibility with Solaris and to make a Java port easier. Partially submitted by: phantom
2003-02-13 17:47:44 +00:00
pathname = arg->buffer;
strncpy(pathname, dir, dirlen);
pathname[dirlen] = '/';
strcpy(pathname + dirlen + 1, arg->name);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
Implement dlinfo() function. Introdice RTLD_SELF special handle and properly process it within dlsym() and dlinfo() functions. The intention is to improve our compatibility with Solaris and to make a Java port easier. Partially submitted by: phantom
2003-02-13 17:47:44 +00:00
dbg(" Trying \"%s\"", pathname);
if (access(pathname, F_OK) == 0) { /* We found it */
pathname = xmalloc(dirlen + 1 + arg->namelen + 1);
strcpy(pathname, arg->buffer);
return (pathname);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
}
}
Implement dlinfo() function. Introdice RTLD_SELF special handle and properly process it within dlsym() and dlinfo() functions. The intention is to improve our compatibility with Solaris and to make a Java port easier. Partially submitted by: phantom
2003-02-13 17:47:44 +00:00
return (NULL);
}
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
Implement dlinfo() function. Introdice RTLD_SELF special handle and properly process it within dlsym() and dlinfo() functions. The intention is to improve our compatibility with Solaris and to make a Java port easier. Partially submitted by: phantom
2003-02-13 17:47:44 +00:00
static char *
search_library_path(const char *name, const char *path)
{
char *p;
struct try_library_args arg;
if (path == NULL)
return NULL;
arg.name = name;
arg.namelen = strlen(name);
arg.buffer = xmalloc(PATH_MAX);
arg.buflen = PATH_MAX;
p = path_enumerate(path, try_library_path, &arg);
free(arg.buffer);
return (p);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
}
Add the LD_LIBRARY_PATH_FDS environmental variable. This variable allows the loading of shared libraries via directory descriptors rather than via library paths. If LD_LIBRARY_PATH_FDS=3:4:12, the directories represented by file descriptors 3, 4 and 12 will searched for shared libraries before the normal path-based mechanisms are used. This allows us to execute unprivileged binaries from within a Capsicum sandbox even if they require shared libraries. Approved by: rwatson (mentor) Reviewed by: kib MFC after: 3 weeks Sponsored by: DARPA/AFRL
2014-06-20 17:08:32 +00:00
/*
* Finds the library with the given name using the directory descriptors
* listed in the LD_LIBRARY_PATH_FDS environment variable.
*
* Returns a freshly-opened close-on-exec file descriptor for the library,
* or -1 if the library cannot be found.
*/
static char *
search_library_pathfds(const char *name, const char *path, int *fdp)
{
char *envcopy, *fdstr, *found, *last_token;
size_t len;
int dirfd, fd;
Remove stray newline.
2014-08-27 00:48:09 +00:00
dbg("%s('%s', '%s', fdp)", __func__, name, path);
Add the LD_LIBRARY_PATH_FDS environmental variable. This variable allows the loading of shared libraries via directory descriptors rather than via library paths. If LD_LIBRARY_PATH_FDS=3:4:12, the directories represented by file descriptors 3, 4 and 12 will searched for shared libraries before the normal path-based mechanisms are used. This allows us to execute unprivileged binaries from within a Capsicum sandbox even if they require shared libraries. Approved by: rwatson (mentor) Reviewed by: kib MFC after: 3 weeks Sponsored by: DARPA/AFRL
2014-06-20 17:08:32 +00:00
/* Don't load from user-specified libdirs into setuid binaries. */
if (!trust)
return (NULL);
/* We can't do anything if LD_LIBRARY_PATH_FDS isn't set. */
if (path == NULL)
return (NULL);
/* LD_LIBRARY_PATH_FDS only works with relative paths. */
if (name[0] == '/') {
dbg("Absolute path (%s) passed to %s", name, __func__);
return (NULL);
}
/*
* Use strtok_r() to walk the FD:FD:FD list. This requires a local
* copy of the path, as strtok_r rewrites separator tokens
* with '\0'.
*/
found = NULL;
envcopy = xstrdup(path);
for (fdstr = strtok_r(envcopy, ":", &last_token); fdstr != NULL;
fdstr = strtok_r(NULL, ":", &last_token)) {
Rename rtld's parse_libdir to parse_integer. This is a more accurate name, as the integer doesn't have to be a library directory descriptor. It is also a prerequisite for more argument parsing coming in the near future (e.g., parsing explicit binary descriptors). Reviewed by: kib MFC after: 2 weeks Sponsored by: NSERC
2017-05-16 13:27:44 +00:00
dirfd = parse_integer(fdstr);
if (dirfd < 0) {
_rtld_error("failed to parse directory FD: '%s'",
fdstr);
Add the LD_LIBRARY_PATH_FDS environmental variable. This variable allows the loading of shared libraries via directory descriptors rather than via library paths. If LD_LIBRARY_PATH_FDS=3:4:12, the directories represented by file descriptors 3, 4 and 12 will searched for shared libraries before the normal path-based mechanisms are used. This allows us to execute unprivileged binaries from within a Capsicum sandbox even if they require shared libraries. Approved by: rwatson (mentor) Reviewed by: kib MFC after: 3 weeks Sponsored by: DARPA/AFRL
2014-06-20 17:08:32 +00:00
break;
Rename rtld's parse_libdir to parse_integer. This is a more accurate name, as the integer doesn't have to be a library directory descriptor. It is also a prerequisite for more argument parsing coming in the near future (e.g., parsing explicit binary descriptors). Reviewed by: kib MFC after: 2 weeks Sponsored by: NSERC
2017-05-16 13:27:44 +00:00
}
Support file verification in MAC. * Add VCREAT flag to indicate when a new file is being created * Add VVERIFY to indicate verification is required * Both VCREAT and VVERIFY are only passed on the MAC method vnode_check_open and are removed from the accmode after * Add O_VERIFY flag to rtld open of objects * Add 'v' flag to __sflags to set O_VERIFY flag. Submitted by: Steve Kiernan <stevek@juniper.net> Obtained from: Juniper Networks, Inc. GitHub Pull Request: https://github.com/freebsd/freebsd/pull/27 Relnotes: yes
2015-04-22 01:54:25 +00:00
fd = __sys_openat(dirfd, name, O_RDONLY | O_CLOEXEC | O_VERIFY);
Add the LD_LIBRARY_PATH_FDS environmental variable. This variable allows the loading of shared libraries via directory descriptors rather than via library paths. If LD_LIBRARY_PATH_FDS=3:4:12, the directories represented by file descriptors 3, 4 and 12 will searched for shared libraries before the normal path-based mechanisms are used. This allows us to execute unprivileged binaries from within a Capsicum sandbox even if they require shared libraries. Approved by: rwatson (mentor) Reviewed by: kib MFC after: 3 weeks Sponsored by: DARPA/AFRL
2014-06-20 17:08:32 +00:00
if (fd >= 0) {
*fdp = fd;
len = strlen(fdstr) + strlen(name) + 3;
found = xmalloc(len);
if (rtld_snprintf(found, len, "#%d/%s", dirfd, name) < 0) {
_rtld_error("error generating '%d/%s'",
dirfd, name);
Make die available as rtld_die for use by MD relocation code Reviewed by: kib Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D2213
2015-04-02 21:35:36 +00:00
rtld_die();
Add the LD_LIBRARY_PATH_FDS environmental variable. This variable allows the loading of shared libraries via directory descriptors rather than via library paths. If LD_LIBRARY_PATH_FDS=3:4:12, the directories represented by file descriptors 3, 4 and 12 will searched for shared libraries before the normal path-based mechanisms are used. This allows us to execute unprivileged binaries from within a Capsicum sandbox even if they require shared libraries. Approved by: rwatson (mentor) Reviewed by: kib MFC after: 3 weeks Sponsored by: DARPA/AFRL
2014-06-20 17:08:32 +00:00
}
dbg("open('%s') => %d", found, fd);
break;
}
}
free(envcopy);
return (found);
}
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
int
dlclose(void *handle)
Avoid bind lock recursion. When dlclose(3) unloads an object with filtees, it recursively calls dlclose(3) on each filtee in free_needed_filtees(). Introduce dlclose_locked() helper, called from free_needed_filtees() instead of dlclose(), and pass the bind lockstate down to avoid recursing. Reported and tested by: jhibbits Sponsored by: The FreeBSD Foundation MFC after: 1 week
2017-03-09 21:05:47 +00:00
{
RtldLockState lockstate;
int error;
wlock_acquire(rtld_bind_lock, &lockstate);
error = dlclose_locked(handle, &lockstate);
lock_release(rtld_bind_lock, &lockstate);
return (error);
}
static int
dlclose_locked(void *handle, RtldLockState *lockstate)
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
{
Add a new function dllockinit() for registering thread locking functions to be used by the dynamic linker. This can be called by threads packages at start-up time. I will add the call to libc_r soon. Also add a default locking method that is used up until dllockinit() is called. The default method works by blocking SIGVTALRM, SIGPROF, and SIGALRM in critical sections. It is based on the observation that most user-space threads packages implement thread preemption with one of these signals (usually SIGVTALRM). The dynamic linker has never been reentrant, but it became less reentrant in revision 1.34 of "src/libexec/rtld-elf/rtld.c". Starting with that revision, multiple threads each doing lazy binding could interfere with each other. The usual symptom was that a symbol was falsely reported as undefined at start-up time. It was rare but not unseen. This commit fixes it.
1999-12-27 04:44:04 +00:00
Obj_Entry *root;
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
Add a new function dllockinit() for registering thread locking functions to be used by the dynamic linker. This can be called by threads packages at start-up time. I will add the call to libc_r soon. Also add a default locking method that is used up until dllockinit() is called. The default method works by blocking SIGVTALRM, SIGPROF, and SIGALRM in critical sections. It is based on the observation that most user-space threads packages implement thread preemption with one of these signals (usually SIGVTALRM). The dynamic linker has never been reentrant, but it became less reentrant in revision 1.34 of "src/libexec/rtld-elf/rtld.c". Starting with that revision, multiple threads each doing lazy binding could interfere with each other. The usual symptom was that a symbol was falsely reported as undefined at start-up time. It was rare but not unseen. This commit fixes it.
1999-12-27 04:44:04 +00:00
root = dlcheck(handle);
Avoid bind lock recursion. When dlclose(3) unloads an object with filtees, it recursively calls dlclose(3) on each filtee in free_needed_filtees(). Introduce dlclose_locked() helper, called from free_needed_filtees() instead of dlclose(), and pass the bind lockstate down to avoid recursing. Reported and tested by: jhibbits Sponsored by: The FreeBSD Foundation MFC after: 1 week
2017-03-09 21:05:47 +00:00
if (root == NULL)
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
return -1;
Add various utrace's for use with ktrace to the ELF runtime linker. To activate the traces, set the LD_UTRACE (or LD_32_UTRACE) environment variable. This also includes code in kdump(8) to parse the traces. Reviewed by: kan, jdp MFC after: 2 weeks
2007-01-09 17:50:05 +00:00
LD_UTRACE(UTRACE_DLCLOSE_START, handle, NULL, 0, root->dl_refcount,
root->path);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
Revamp the mechanism for enumerating and calling shared objects' init and fini functions. Now the code is very careful to hold no locks when calling these functions. Thus the dynamic linker cannot be re-entered with a lock already held. Remove the tolerance for recursive locking that I added in revision 1.2 of dllockinit.c. Recursive locking shouldn't happen any more. Mozilla and JDK users: I'd appreciate confirmation that things still work right (or at least the same) with these changes.
2000-01-09 21:13:48 +00:00
/* Unreference the object and its dependencies. */
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
root->dl_refcount--;
Rethink the way we count module references. Simply following DT_NEEDED links is not flexible enough for cases where dynamically loaded modules form a dependency cycle. This should fix an infinite recursion problem encountered by Yahoo. Approved by: re (jhb)
2003-05-08 01:31:36 +00:00
If dlclose() is called recursively from a _fini() function, the inner dlclose() call may unload the object of the outer call prematurely because objects are unreferenced before _fini() calls. Fix this by unreferencing objects after calling objlist_call_fini() in dlclose(). Therefore objlist_call_fini() now calls the fini function if the reference count of an object is 1. In addition we must restart the list_fini traversal after every _fini() call because another dlclose() call might have modified the reference counts. Add an XXX comment to objlist_call_fini() about possible race with dlopen(). PR: 133246, 149464 Reviewed by: kan, kib
2010-12-16 16:56:44 +00:00
if (root->refcount == 1) {
Revamp the mechanism for enumerating and calling shared objects' init and fini functions. Now the code is very careful to hold no locks when calling these functions. Thus the dynamic linker cannot be re-entered with a lock already held. Remove the tolerance for recursive locking that I added in revision 1.2 of dllockinit.c. Recursive locking shouldn't happen any more. Mozilla and JDK users: I'd appreciate confirmation that things still work right (or at least the same) with these changes.
2000-01-09 21:13:48 +00:00
/*
If dlclose() is called recursively from a _fini() function, the inner dlclose() call may unload the object of the outer call prematurely because objects are unreferenced before _fini() calls. Fix this by unreferencing objects after calling objlist_call_fini() in dlclose(). Therefore objlist_call_fini() now calls the fini function if the reference count of an object is 1. In addition we must restart the list_fini traversal after every _fini() call because another dlclose() call might have modified the reference counts. Add an XXX comment to objlist_call_fini() about possible race with dlopen(). PR: 133246, 149464 Reviewed by: kan, kib
2010-12-16 16:56:44 +00:00
* The object will be no longer referenced, so we must unload it.
Make sure internal rtld malloc routines are not called from unlocked contexts as rtld's malloc is not thread safe and is only supposed to be called with exclusive bind lock already held. The originating PR submitted a patch on top of different pre-requisite workaroud for unsafe dlopen calls, and the patch was midief slighlty to apply to stock sources for the purpose of this commit. Running rtld malloc from unlocked contexts is a bug on its own. PR: 126950 Submited by: Oleg Dolgov
2008-09-03 01:05:32 +00:00
* First, call the fini functions.
Revamp the mechanism for enumerating and calling shared objects' init and fini functions. Now the code is very careful to hold no locks when calling these functions. Thus the dynamic linker cannot be re-entered with a lock already held. Remove the tolerance for recursive locking that I added in revision 1.2 of dllockinit.c. Recursive locking shouldn't happen any more. Mozilla and JDK users: I'd appreciate confirmation that things still work right (or at least the same) with these changes.
2000-01-09 21:13:48 +00:00
*/
Avoid bind lock recursion. When dlclose(3) unloads an object with filtees, it recursively calls dlclose(3) on each filtee in free_needed_filtees(). Introduce dlclose_locked() helper, called from free_needed_filtees() instead of dlclose(), and pass the bind lockstate down to avoid recursing. Reported and tested by: jhibbits Sponsored by: The FreeBSD Foundation MFC after: 1 week
2017-03-09 21:05:47 +00:00
objlist_call_fini(&list_fini, root, lockstate);
If dlclose() is called recursively from a _fini() function, the inner dlclose() call may unload the object of the outer call prematurely because objects are unreferenced before _fini() calls. Fix this by unreferencing objects after calling objlist_call_fini() in dlclose(). Therefore objlist_call_fini() now calls the fini function if the reference count of an object is 1. In addition we must restart the list_fini traversal after every _fini() call because another dlclose() call might have modified the reference counts. Add an XXX comment to objlist_call_fini() about possible race with dlopen(). PR: 133246, 149464 Reviewed by: kan, kib
2010-12-16 16:56:44 +00:00
unref_dag(root);
Revamp the mechanism for enumerating and calling shared objects' init and fini functions. Now the code is very careful to hold no locks when calling these functions. Thus the dynamic linker cannot be re-entered with a lock already held. Remove the tolerance for recursive locking that I added in revision 1.2 of dllockinit.c. Recursive locking shouldn't happen any more. Mozilla and JDK users: I'd appreciate confirmation that things still work right (or at least the same) with these changes.
2000-01-09 21:13:48 +00:00
/* Finish cleaning up the newly-unreferenced objects. */
Pass two pointer parameters to the r_debug_state() hook function, thus allowing a debugger or other trace tool to easily grab the addresses of the needed structures off the stack. This change is transparent to gdb, which locates the link_map list and transfers it to debugger memory for comparison purposes. A sample program will be committed showing how this can be used. Reviewed by: John Polstra <jdp@FreeBSD.org>
2000-08-26 05:13:29 +00:00
GDB_STATE(RT_DELETE,&root->linkmap);
Avoid bind lock recursion. When dlclose(3) unloads an object with filtees, it recursively calls dlclose(3) on each filtee in free_needed_filtees(). Introduce dlclose_locked() helper, called from free_needed_filtees() instead of dlclose(), and pass the bind lockstate down to avoid recursing. Reported and tested by: jhibbits Sponsored by: The FreeBSD Foundation MFC after: 1 week
2017-03-09 21:05:47 +00:00
unload_object(root, lockstate);
Pass two pointer parameters to the r_debug_state() hook function, thus allowing a debugger or other trace tool to easily grab the addresses of the needed structures off the stack. This change is transparent to gdb, which locates the link_map list and transfers it to debugger memory for comparison purposes. A sample program will be committed showing how this can be used. Reviewed by: John Polstra <jdp@FreeBSD.org>
2000-08-26 05:13:29 +00:00
GDB_STATE(RT_CONSISTENT,NULL);
If dlclose() is called recursively from a _fini() function, the inner dlclose() call may unload the object of the outer call prematurely because objects are unreferenced before _fini() calls. Fix this by unreferencing objects after calling objlist_call_fini() in dlclose(). Therefore objlist_call_fini() now calls the fini function if the reference count of an object is 1. In addition we must restart the list_fini traversal after every _fini() call because another dlclose() call might have modified the reference counts. Add an XXX comment to objlist_call_fini() about possible race with dlopen(). PR: 133246, 149464 Reviewed by: kan, kib
2010-12-16 16:56:44 +00:00
} else
unref_dag(root);
Add various utrace's for use with ktrace to the ELF runtime linker. To activate the traces, set the LD_UTRACE (or LD_32_UTRACE) environment variable. This also includes code in kdump(8) to parse the traces. Reviewed by: kan, jdp MFC after: 2 weeks
2007-01-09 17:50:05 +00:00
LD_UTRACE(UTRACE_DLCLOSE_STOP, handle, NULL, 0, 0, NULL);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
return 0;
}
- Remove const'ness from dlerror(3) prototype, for consistency with POSIX. Approved by: cognet MFC after: 1 week
2010-03-24 15:59:51 +00:00
char *
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
dlerror(void)
{
char *msg = error_message;
error_message = NULL;
return msg;
}
Solve the dynamic linker's problems with multithreaded programs once and for all (I hope). Packages such as wine, JDK, and linuxthreads should no longer have any problems with re-entering the dynamic linker. This commit replaces the locking used in the dynamic linker with a new spinlock-based reader/writer lock implementation. Brian Fundakowski Feldman <green> argued for this from the very beginning, but it took me a long time to come around to his point of view. Spinlocks are the only kinds of locks that work with all thread packages. But on uniprocessor systems they can be inefficient, because while a contender for the lock is spinning the holder of the lock cannot make any progress toward releasing it. To alleviate this disadvantage I have borrowed a trick from Sleepycat's Berkeley DB implementation. When spinning for a lock, the requester does a nanosleep() call for 1 usec. each time around the loop. This will generally yield the CPU to other threads, allowing the lock holder to finish its business and release the lock. I chose 1 usec. as the minimum sleep which would with reasonable certainty not be rounded down to 0. The formerly machine-independent file "lockdflt.c" has been moved into the architecture-specific subdirectories by repository copy. It now contains the machine-dependent spinlocking code. For the spinlocks I used the very nifty "simple, non-scalable reader-preference lock" which I found at <http://www.cs.rochester.edu/u/scott/synchronization/pseudocode/rw.html> on all CPUs except the 80386 (the specific CPU model, not the architecture). The 80386 CPU doesn't support the necessary "cmpxchg" instruction, so on that CPU a simple exclusive test-and-set lock is used instead. 80386 CPUs are detected at initialization time by trying to execute "cmpxchg" and catching the resulting SIGILL signal. To reduce contention for the locks, I have revamped a couple of key data structures, permitting all common operations to be done under non-exclusive (reader) locking. The only operations that require exclusive locking now are the rare intrusive operations such as dlopen() and dlclose(). The dllockinit() interface is now deprecated. It still exists, but only as a do-nothing stub. I plan to remove it as soon as is reasonably possible. (From the very beginning it was clearly labeled as experimental and subject to change.) As far as I know, only the linuxthreads port uses dllockinit(). This interface turned out to have several problems. As one example, when the dynamic linker called a client-supplied locking function, that function sometimes needed lazy binding, causing re-entry into the dynamic linker and a big looping mess. And in any case, it turned out to be too burdensome to require threads packages to register themselves with the dynamic linker.
2000-07-08 04:10:38 +00:00
/*
* This function is deprecated and has no effect.
*/
Add a new function dllockinit() for registering thread locking functions to be used by the dynamic linker. This can be called by threads packages at start-up time. I will add the call to libc_r soon. Also add a default locking method that is used up until dllockinit() is called. The default method works by blocking SIGVTALRM, SIGPROF, and SIGALRM in critical sections. It is based on the observation that most user-space threads packages implement thread preemption with one of these signals (usually SIGVTALRM). The dynamic linker has never been reentrant, but it became less reentrant in revision 1.34 of "src/libexec/rtld-elf/rtld.c". Starting with that revision, multiple threads each doing lazy binding could interfere with each other. The usual symptom was that a symbol was falsely reported as undefined at start-up time. It was rare but not unseen. This commit fixes it.
1999-12-27 04:44:04 +00:00
void
dllockinit(void *context,
void *(*lock_create)(void *context),
void (*rlock_acquire)(void *lock),
void (*wlock_acquire)(void *lock),
void (*lock_release)(void *lock),
void (*lock_destroy)(void *lock),
void (*context_destroy)(void *context))
{
Solve the dynamic linker's problems with multithreaded programs once and for all (I hope). Packages such as wine, JDK, and linuxthreads should no longer have any problems with re-entering the dynamic linker. This commit replaces the locking used in the dynamic linker with a new spinlock-based reader/writer lock implementation. Brian Fundakowski Feldman <green> argued for this from the very beginning, but it took me a long time to come around to his point of view. Spinlocks are the only kinds of locks that work with all thread packages. But on uniprocessor systems they can be inefficient, because while a contender for the lock is spinning the holder of the lock cannot make any progress toward releasing it. To alleviate this disadvantage I have borrowed a trick from Sleepycat's Berkeley DB implementation. When spinning for a lock, the requester does a nanosleep() call for 1 usec. each time around the loop. This will generally yield the CPU to other threads, allowing the lock holder to finish its business and release the lock. I chose 1 usec. as the minimum sleep which would with reasonable certainty not be rounded down to 0. The formerly machine-independent file "lockdflt.c" has been moved into the architecture-specific subdirectories by repository copy. It now contains the machine-dependent spinlocking code. For the spinlocks I used the very nifty "simple, non-scalable reader-preference lock" which I found at <http://www.cs.rochester.edu/u/scott/synchronization/pseudocode/rw.html> on all CPUs except the 80386 (the specific CPU model, not the architecture). The 80386 CPU doesn't support the necessary "cmpxchg" instruction, so on that CPU a simple exclusive test-and-set lock is used instead. 80386 CPUs are detected at initialization time by trying to execute "cmpxchg" and catching the resulting SIGILL signal. To reduce contention for the locks, I have revamped a couple of key data structures, permitting all common operations to be done under non-exclusive (reader) locking. The only operations that require exclusive locking now are the rare intrusive operations such as dlopen() and dlclose(). The dllockinit() interface is now deprecated. It still exists, but only as a do-nothing stub. I plan to remove it as soon as is reasonably possible. (From the very beginning it was clearly labeled as experimental and subject to change.) As far as I know, only the linuxthreads port uses dllockinit(). This interface turned out to have several problems. As one example, when the dynamic linker called a client-supplied locking function, that function sometimes needed lazy binding, causing re-entry into the dynamic linker and a big looping mess. And in any case, it turned out to be too burdensome to require threads packages to register themselves with the dynamic linker.
2000-07-08 04:10:38 +00:00
static void *cur_context;
static void (*cur_context_destroy)(void *);
/* Just destroy the context from the previous call, if necessary. */
if (cur_context_destroy != NULL)
cur_context_destroy(cur_context);
cur_context = context;
cur_context_destroy = context_destroy;
When a threads package registers locking methods with dllockinit(), figure out which shared object(s) contain the the locking methods and fully bind those objects as if they had been loaded with LD_BIND_NOW=1. The goal is to keep the locking methods from requiring any lazy binding. Otherwise infinite recursion occurs in _rtld_bind. This fixes the infinite recursion problem in the linuxthreads port.
2000-01-29 01:27:04 +00:00
}
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
void *
dlopen(const char *name, int mode)
Implement fdlopen(3), an rtld interface to load shared object by file descriptor. Requested and tested by: des (previous version) Reviewed by: des, kan (previous version) MFC after: 2 weeks
2012-01-07 10:33:01 +00:00
{
return (rtld_dlopen(name, -1, mode));
}
void *
fdlopen(int fd, int mode)
{
return (rtld_dlopen(NULL, fd, mode));
}
static void *
rtld_dlopen(const char *name, int fd, int mode)
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
{
get_program_var_addr() must prefer the strong symbol to the weak one. Search global objects, together with main object and dependencies, for the requested symbol. Move the common code from symlook_default() into new helper symlook_global(), and use it both in symlook_global() and get_program_var_addr(). Supply lock state to get_program_var_addr(). Reviewed by: kan Tested by: Mykola Dzham <i levsha me>
2011-01-10 16:09:35 +00:00
RtldLockState lockstate;
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
int lo_flags;
Allow ldd(1) be used on shared libraries in addition to executables.
2002-02-04 10:33:48 +00:00
Add various utrace's for use with ktrace to the ELF runtime linker. To activate the traces, set the LD_UTRACE (or LD_32_UTRACE) environment variable. This also includes code in kdump(8) to parse the traces. Reviewed by: kan, jdp MFC after: 2 weeks
2007-01-09 17:50:05 +00:00
LD_UTRACE(UTRACE_DLOPEN_START, NULL, NULL, 0, mode, name);
Allow ldd(1) be used on shared libraries in addition to executables.
2002-02-04 10:33:48 +00:00
ld_tracing = (mode & RTLD_TRACE) == 0 ? NULL : "1";
get_program_var_addr() must prefer the strong symbol to the weak one. Search global objects, together with main object and dependencies, for the requested symbol. Move the common code from symlook_default() into new helper symlook_global(), and use it both in symlook_global() and get_program_var_addr(). Supply lock state to get_program_var_addr(). Reviewed by: kan Tested by: Mykola Dzham <i levsha me>
2011-01-10 16:09:35 +00:00
if (ld_tracing != NULL) {
rlock_acquire(rtld_bind_lock, &lockstate);
Use sigsetjmp/siglongjmp with disabled signal mask access for lock upgrade in rtld. There is no need to care about the mask, which causes a lot of unneeded sigprocmask(2) calls during each symbol lookup.
2011-02-09 09:20:27 +00:00
if (sigsetjmp(lockstate.env, 0) != 0)
get_program_var_addr() must prefer the strong symbol to the weak one. Search global objects, together with main object and dependencies, for the requested symbol. Move the common code from symlook_default() into new helper symlook_global(), and use it both in symlook_global() and get_program_var_addr(). Supply lock state to get_program_var_addr(). Reviewed by: kan Tested by: Mykola Dzham <i levsha me>
2011-01-10 16:09:35 +00:00
lock_upgrade(rtld_bind_lock, &lockstate);
environ = (char **)*get_program_var_addr("environ", &lockstate);
lock_release(rtld_bind_lock, &lockstate);
}
Implement rtld part of the support for -z nodlopen (see ld(1)). Reviewed by: kan MFC after: 3 weeks
2009-11-26 13:57:20 +00:00
lo_flags = RTLD_LO_DLOPEN;
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
if (mode & RTLD_NODELETE)
lo_flags |= RTLD_LO_NODELETE;
Implement rtld part of the support for -z nodlopen (see ld(1)). Reviewed by: kan MFC after: 3 weeks
2009-11-26 13:57:20 +00:00
if (mode & RTLD_NOLOAD)
lo_flags |= RTLD_LO_NOLOAD;
Allow to load not-openable dso when tracing. This fixes ldd on such dso or dso linked to non-openable object. Remove '\n' at the end of error message. End comments with dot. MFC after: 3 weeks (together with r199829)
2009-11-28 14:29:32 +00:00
if (ld_tracing != NULL)
lo_flags |= RTLD_LO_TRACE;
Revamp the mechanism for enumerating and calling shared objects' init and fini functions. Now the code is very careful to hold no locks when calling these functions. Thus the dynamic linker cannot be re-entered with a lock already held. Remove the tolerance for recursive locking that I added in revision 1.2 of dllockinit.c. Recursive locking shouldn't happen any more. Mozilla and JDK users: I'd appreciate confirmation that things still work right (or at least the same) with these changes.
2000-01-09 21:13:48 +00:00
Implement fdlopen(3), an rtld interface to load shared object by file descriptor. Requested and tested by: des (previous version) Reviewed by: des, kan (previous version) MFC after: 2 weeks
2012-01-07 10:33:01 +00:00
return (dlopen_object(name, fd, obj_main, lo_flags,
Propagate the current state of rtld_bind_lock to dlopen_object() calls through the filter loading call chain. This fixes attempts to write-lock the already locked rtld_bind_lock when filter loading is initiated by relocation of dlopening dso. Reported and tested by: Taku YAMAMOTO <taku tackymt homeip net> MFC after: 1 week
2012-04-12 10:32:22 +00:00
mode & (RTLD_MODEMASK | RTLD_GLOBAL), NULL));
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
}
_rtld_bind() read-locks the bind lock, and possible plt resolution from the dispatcher would also acquire bind lock in read mode, which is the supported operation. plt is explicitely designed to allow safe multithreaded updates, so the shared lock do not cause problems. The error in r228435 is that it allows read lock acquisition after the write lock for the bind block. If we dlopened the shared object that contains IRELATIVE or jump slot which target is STT_GNU_IFUNC, then possible recursive plt resolve from the dispatcher would cause it. Postpone the resolution for irelative/ifunc right before initializers are called, and drop bind lock around calls to dispatcher. Use initlist to iterate over the objects instead of the ->next, due to drop of the bind lock in iteration. For i386/reloc.c:reloc_iresolve(), fix calculation of the dispatch function address for dso, by taking into account possible non-zero relocbase. MFC after: 3 weeks
2011-12-14 16:47:53 +00:00
static void
Avoid bind lock recursion. When dlclose(3) unloads an object with filtees, it recursively calls dlclose(3) on each filtee in free_needed_filtees(). Introduce dlclose_locked() helper, called from free_needed_filtees() instead of dlclose(), and pass the bind lockstate down to avoid recursing. Reported and tested by: jhibbits Sponsored by: The FreeBSD Foundation MFC after: 1 week
2017-03-09 21:05:47 +00:00
dlopen_cleanup(Obj_Entry *obj, RtldLockState *lockstate)
_rtld_bind() read-locks the bind lock, and possible plt resolution from the dispatcher would also acquire bind lock in read mode, which is the supported operation. plt is explicitely designed to allow safe multithreaded updates, so the shared lock do not cause problems. The error in r228435 is that it allows read lock acquisition after the write lock for the bind block. If we dlopened the shared object that contains IRELATIVE or jump slot which target is STT_GNU_IFUNC, then possible recursive plt resolve from the dispatcher would cause it. Postpone the resolution for irelative/ifunc right before initializers are called, and drop bind lock around calls to dispatcher. Use initlist to iterate over the objects instead of the ->next, due to drop of the bind lock in iteration. For i386/reloc.c:reloc_iresolve(), fix calculation of the dispatch function address for dso, by taking into account possible non-zero relocbase. MFC after: 3 weeks
2011-12-14 16:47:53 +00:00
{
obj->dl_refcount--;
unref_dag(obj);
if (obj->refcount == 0)
Avoid bind lock recursion. When dlclose(3) unloads an object with filtees, it recursively calls dlclose(3) on each filtee in free_needed_filtees(). Introduce dlclose_locked() helper, called from free_needed_filtees() instead of dlclose(), and pass the bind lockstate down to avoid recursing. Reported and tested by: jhibbits Sponsored by: The FreeBSD Foundation MFC after: 1 week
2017-03-09 21:05:47 +00:00
unload_object(obj, lockstate);
_rtld_bind() read-locks the bind lock, and possible plt resolution from the dispatcher would also acquire bind lock in read mode, which is the supported operation. plt is explicitely designed to allow safe multithreaded updates, so the shared lock do not cause problems. The error in r228435 is that it allows read lock acquisition after the write lock for the bind block. If we dlopened the shared object that contains IRELATIVE or jump slot which target is STT_GNU_IFUNC, then possible recursive plt resolve from the dispatcher would cause it. Postpone the resolution for irelative/ifunc right before initializers are called, and drop bind lock around calls to dispatcher. Use initlist to iterate over the objects instead of the ->next, due to drop of the bind lock in iteration. For i386/reloc.c:reloc_iresolve(), fix calculation of the dispatch function address for dso, by taking into account possible non-zero relocbase. MFC after: 3 weeks
2011-12-14 16:47:53 +00:00
}
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
static Obj_Entry *
Implement fdlopen(3), an rtld interface to load shared object by file descriptor. Requested and tested by: des (previous version) Reviewed by: des, kan (previous version) MFC after: 2 weeks
2012-01-07 10:33:01 +00:00
dlopen_object(const char *name, int fd, Obj_Entry *refobj, int lo_flags,
Propagate the current state of rtld_bind_lock to dlopen_object() calls through the filter loading call chain. This fixes attempts to write-lock the already locked rtld_bind_lock when filter loading is initiated by relocation of dlopening dso. Reported and tested by: Taku YAMAMOTO <taku tackymt homeip net> MFC after: 1 week
2012-04-12 10:32:22 +00:00
int mode, RtldLockState *lockstate)
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
{
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
Obj_Entry *old_obj_tail;
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
Obj_Entry *obj;
Objlist initlist;
Propagate the current state of rtld_bind_lock to dlopen_object() calls through the filter loading call chain. This fixes attempts to write-lock the already locked rtld_bind_lock when filter loading is initiated by relocation of dlopening dso. Reported and tested by: Taku YAMAMOTO <taku tackymt homeip net> MFC after: 1 week
2012-04-12 10:32:22 +00:00
RtldLockState mlockstate;
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
int result;
Revamp the code that calls shared libraries' init and fini functions. Formerly the init functions were called in the opposite of the order in which libraries were loaded, and libraries were loaded according to a breadth-first traversal of the dependency graph. That ordering came from SVR4.0, and it was easy to implement but not always sensible. Now we do a depth-first walk over the dependency graph and call the init functions in an order such that each shared object's needed objects are initialized before the shared object itself. At the same time we build a list of finalization (fini) functions in the opposite order, to guarantee correct C++ destructor ordering whenever possible. (It may not be possible if dlopen and dlclose are used in strange ways, but we come as close as one can come.) The need for this renovation has become apparent as more programs have started using multithreading. The multithreaded C library libc_r requires initialization, whereas the standard libc does not. Since virtually every other object depends on the C library, it is important that it get initialized first.
2000-07-26 04:24:40 +00:00
objlist_init(&initlist);
Add GDB support. The method and some of the code came from NetBSD's elf runtime linker.
1998-04-30 07:48:02 +00:00
Propagate the current state of rtld_bind_lock to dlopen_object() calls through the filter loading call chain. This fixes attempts to write-lock the already locked rtld_bind_lock when filter loading is initiated by relocation of dlopening dso. Reported and tested by: Taku YAMAMOTO <taku tackymt homeip net> MFC after: 1 week
2012-04-12 10:32:22 +00:00
if (lockstate == NULL && !(lo_flags & RTLD_LO_EARLY)) {
wlock_acquire(rtld_bind_lock, &mlockstate);
lockstate = &mlockstate;
}
Pass two pointer parameters to the r_debug_state() hook function, thus allowing a debugger or other trace tool to easily grab the addresses of the needed structures off the stack. This change is transparent to gdb, which locates the link_map list and transfers it to debugger memory for comparison purposes. A sample program will be committed showing how this can be used. Reviewed by: John Polstra <jdp@FreeBSD.org>
2000-08-26 05:13:29 +00:00
GDB_STATE(RT_ADD,NULL);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
old_obj_tail = globallist_curr(TAILQ_LAST(&obj_list, obj_entry_q));
Add a new function dllockinit() for registering thread locking functions to be used by the dynamic linker. This can be called by threads packages at start-up time. I will add the call to libc_r soon. Also add a default locking method that is used up until dllockinit() is called. The default method works by blocking SIGVTALRM, SIGPROF, and SIGALRM in critical sections. It is based on the observation that most user-space threads packages implement thread preemption with one of these signals (usually SIGVTALRM). The dynamic linker has never been reentrant, but it became less reentrant in revision 1.34 of "src/libexec/rtld-elf/rtld.c". Starting with that revision, multiple threads each doing lazy binding could interfere with each other. The usual symptom was that a symbol was falsely reported as undefined at start-up time. It was rare but not unseen. This commit fixes it.
1999-12-27 04:44:04 +00:00
obj = NULL;
Implement fdlopen(3), an rtld interface to load shared object by file descriptor. Requested and tested by: des (previous version) Reviewed by: des, kan (previous version) MFC after: 2 weeks
2012-01-07 10:33:01 +00:00
if (name == NULL && fd == -1) {
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
obj = obj_main;
Fix a reference counting problem when using dlopen(NULL, ...). PR: bin/12129
1999-06-25 04:50:06 +00:00
obj->refcount++;
} else {
Implement fdlopen(3), an rtld interface to load shared object by file descriptor. Requested and tested by: des (previous version) Reviewed by: des, kan (previous version) MFC after: 2 weeks
2012-01-07 10:33:01 +00:00
obj = load_object(name, fd, refobj, lo_flags);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
}
Add GDB support. The method and some of the code came from NetBSD's elf runtime linker.
1998-04-30 07:48:02 +00:00
if (obj) {
obj->dl_refcount++;
Revamp the symbol lookup algorithm to cope better with objects loaded separately by dlopen that have global symbols with identical names. Viewing each dlopened object as a DAG which is linked by its DT_NEEDED entries in the dynamic table, the search order is as follows: * If the referencing object was linked with -Bsymbolic, search it internally. * Search all dlopened DAGs containing the referencing object. * Search all objects loaded at program start up. * Search all objects which were dlopened() using the RTLD_GLOBAL flag (which is now supported too). The search terminates as soon as a strong definition is found. Lacking that, the first weak definition is used. These rules match those of Solaris, as best I could determine them from its vague manual pages and the results of experiments I performed. PR: misc/12438
1999-08-30 01:48:19 +00:00
if (mode & RTLD_GLOBAL && objlist_find(&list_global, obj) == NULL)
Revamp the code that calls shared libraries' init and fini functions. Formerly the init functions were called in the opposite of the order in which libraries were loaded, and libraries were loaded according to a breadth-first traversal of the dependency graph. That ordering came from SVR4.0, and it was easy to implement but not always sensible. Now we do a depth-first walk over the dependency graph and call the init functions in an order such that each shared object's needed objects are initialized before the shared object itself. At the same time we build a list of finalization (fini) functions in the opposite order, to guarantee correct C++ destructor ordering whenever possible. (It may not be possible if dlopen and dlclose are used in strange ways, but we come as close as one can come.) The need for this renovation has become apparent as more programs have started using multithreading. The multithreaded C library libc_r requires initialization, whereas the standard libc does not. Since virtually every other object depends on the C library, it is important that it get initialized first.
2000-07-26 04:24:40 +00:00
objlist_push_tail(&list_global, obj);
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
if (globallist_next(old_obj_tail) != NULL) {
/* We loaded something new. */
assert(globallist_next(old_obj_tail) == obj);
Fix several problems with our ELF filters implementation. Do not relocate twice an object which happens to be needed by loaded binary (or dso) and some filtee opened due to symbol resolution when relocating need objects. Record the state of the relocation processing in Obj_Entry and short-circuit relocate_objects() if current object already processed. Do not call constructors for filtees loaded during the early relocation processing before image is initialized enough to run user-provided code. Filtees are loaded using dlopen_object(), which normally performs relocation and initialization. If filtee is lazy-loaded during the relocation of dso needed by the main object, dlopen_object() runs too earlier, when most runtime services are not yet ready. Postpone the constructors call to the time when main binary and depended libraries constructors are run, passing the new flag RTLD_LO_EARLY to dlopen_object(). Symbol lookups callers inform symlook_* functions about early stage of initialization with SYMLOOK_EARLY. Pass flags through all functions participating in object relocation. Use the opportunity and fix flags argument to find_symdef() in arch-specific reloc.c to use proper name SYMLOOK_IN_PLT instead of true, which happen to have the same numeric value. Reported and tested by: theraven Reviewed by: kan MFC after: 2 weeks
2012-03-20 13:20:49 +00:00
result = load_needed_objects(obj,
lo_flags & (RTLD_LO_DLOPEN | RTLD_LO_EARLY));
Initialize object dagmembers list before checking version dependencies.
2005-12-22 16:42:38 +00:00
init_dag(obj);
In r214728, if dlopen() is called for the object that has been already loaded as a dependency and marked -z nodlopen, object' DAG is already initialized by load_needed_objects(). Due to this, the init_dag() call from dlopen() does not increment refcount for the object [1]. Change init_dag() to not increment DAG refcount. Require explicit calls to ref_dag() to increment, and assert that ref_dag() and unref_dag() are called for root that has dag initialized. To fix the noted issue, unconditionally call both init_dag() and ref_dag() in dlopen() for the case when the object was already loaded, making it similar to the case of newly loaded object. Noted by: jh [1] Reviewed by: jh, kan MFC after: 6 days
2010-11-04 09:29:00 +00:00
ref_dag(obj);
Implement ELF symbol versioning using GNU semantics. This code aims to be compatible with symbol versioning support as implemented by GNU libc and documented by http://people.redhat.com/~drepper/symbol-versioning and LSB 3.0. Implement dlvsym() function to allow lookups for a specific version of a given symbol.
2005-12-18 19:43:33 +00:00
if (result != -1)
result = rtld_verify_versions(&obj->dagmembers);
Fix a problem with RTLD_TRACE flag to dlopen(3), which sometimes can return even if there was no error occured (when trying to dlopen(3) object that already linked into executable which does dlopen(3) call). This is more proper fix for `ldd /usr/lib/libc.so' problem, because the new behaviour conforms to documentation. Remove workaround from ldd.c (rev.1.32). PR: 35099 Submitted by: Nathan Hawkins <utsl@quic.net> MFC after: 1 week
2002-10-19 10:18:29 +00:00
if (result != -1 && ld_tracing)
goto trace;
Ensure that for the object which is a dependency for some filtee, relocations are performed before the object's initializer is called. When dlopen()ing an object, relocate the whole DAG rooted in the object instead of only relocating the object itself and list of newly loaded dependencies. Reversed sequence currently can occur if the same object is a dependency for both filtee and filter, since filtees are loaded typically during the relocation processing, when some filter dependencies might be already loaded but not relocated yet. Reported and tested by: swills Reviewed by: kan MFC after: 1 week
2012-06-27 20:24:25 +00:00
if (result == -1 || relocate_object_dag(obj,
(mode & RTLD_MODEMASK) == RTLD_NOW, &obj_rtld,
Fix several problems with our ELF filters implementation. Do not relocate twice an object which happens to be needed by loaded binary (or dso) and some filtee opened due to symbol resolution when relocating need objects. Record the state of the relocation processing in Obj_Entry and short-circuit relocate_objects() if current object already processed. Do not call constructors for filtees loaded during the early relocation processing before image is initialized enough to run user-provided code. Filtees are loaded using dlopen_object(), which normally performs relocation and initialization. If filtee is lazy-loaded during the relocation of dso needed by the main object, dlopen_object() runs too earlier, when most runtime services are not yet ready. Postpone the constructors call to the time when main binary and depended libraries constructors are run, passing the new flag RTLD_LO_EARLY to dlopen_object(). Symbol lookups callers inform symlook_* functions about early stage of initialization with SYMLOOK_EARLY. Pass flags through all functions participating in object relocation. Use the opportunity and fix flags argument to find_symdef() in arch-specific reloc.c to use proper name SYMLOOK_IN_PLT instead of true, which happen to have the same numeric value. Reported and tested by: theraven Reviewed by: kan MFC after: 2 weeks
2012-03-20 13:20:49 +00:00
(lo_flags & RTLD_LO_EARLY) ? SYMLOOK_EARLY : 0,
Ensure that for the object which is a dependency for some filtee, relocations are performed before the object's initializer is called. When dlopen()ing an object, relocate the whole DAG rooted in the object instead of only relocating the object itself and list of newly loaded dependencies. Reversed sequence currently can occur if the same object is a dependency for both filtee and filter, since filtees are loaded typically during the relocation processing, when some filter dependencies might be already loaded but not relocated yet. Reported and tested by: swills Reviewed by: kan MFC after: 1 week
2012-06-27 20:24:25 +00:00
lockstate) == -1) {
Avoid bind lock recursion. When dlclose(3) unloads an object with filtees, it recursively calls dlclose(3) on each filtee in free_needed_filtees(). Introduce dlclose_locked() helper, called from free_needed_filtees() instead of dlclose(), and pass the bind lockstate down to avoid recursing. Reported and tested by: jhibbits Sponsored by: The FreeBSD Foundation MFC after: 1 week
2017-03-09 21:05:47 +00:00
dlopen_cleanup(obj, lockstate);
Add GDB support. The method and some of the code came from NetBSD's elf runtime linker.
1998-04-30 07:48:02 +00:00
obj = NULL;
Fix several problems with our ELF filters implementation. Do not relocate twice an object which happens to be needed by loaded binary (or dso) and some filtee opened due to symbol resolution when relocating need objects. Record the state of the relocation processing in Obj_Entry and short-circuit relocate_objects() if current object already processed. Do not call constructors for filtees loaded during the early relocation processing before image is initialized enough to run user-provided code. Filtees are loaded using dlopen_object(), which normally performs relocation and initialization. If filtee is lazy-loaded during the relocation of dso needed by the main object, dlopen_object() runs too earlier, when most runtime services are not yet ready. Postpone the constructors call to the time when main binary and depended libraries constructors are run, passing the new flag RTLD_LO_EARLY to dlopen_object(). Symbol lookups callers inform symlook_* functions about early stage of initialization with SYMLOOK_EARLY. Pass flags through all functions participating in object relocation. Use the opportunity and fix flags argument to find_symdef() in arch-specific reloc.c to use proper name SYMLOOK_IN_PLT instead of true, which happen to have the same numeric value. Reported and tested by: theraven Reviewed by: kan MFC after: 2 weeks
2012-03-20 13:20:49 +00:00
} else if (lo_flags & RTLD_LO_EARLY) {
/*
* Do not call the init functions for early loaded
* filtees. The image is still not initialized enough
* for them to work.
*
* Our object is found by the global object list and
* will be ordered among all init calls done right
* before transferring control to main.
*/
Revamp the mechanism for enumerating and calling shared objects' init and fini functions. Now the code is very careful to hold no locks when calling these functions. Thus the dynamic linker cannot be re-entered with a lock already held. Remove the tolerance for recursive locking that I added in revision 1.2 of dllockinit.c. Recursive locking shouldn't happen any more. Mozilla and JDK users: I'd appreciate confirmation that things still work right (or at least the same) with these changes.
2000-01-09 21:13:48 +00:00
} else {
Revamp the code that calls shared libraries' init and fini functions. Formerly the init functions were called in the opposite of the order in which libraries were loaded, and libraries were loaded according to a breadth-first traversal of the dependency graph. That ordering came from SVR4.0, and it was easy to implement but not always sensible. Now we do a depth-first walk over the dependency graph and call the init functions in an order such that each shared object's needed objects are initialized before the shared object itself. At the same time we build a list of finalization (fini) functions in the opposite order, to guarantee correct C++ destructor ordering whenever possible. (It may not be possible if dlopen and dlclose are used in strange ways, but we come as close as one can come.) The need for this renovation has become apparent as more programs have started using multithreading. The multithreaded C library libc_r requires initialization, whereas the standard libc does not. Since virtually every other object depends on the C library, it is important that it get initialized first.
2000-07-26 04:24:40 +00:00
/* Make list of init functions to call. */
Fix initlist_add_object invocation parameters. The tail parameter should point to the last object for which dependencies should be processed. In most cases, this is the object itself.
2016-01-20 23:26:35 +00:00
initlist_add_objects(obj, obj, &initlist);
Revamp the mechanism for enumerating and calling shared objects' init and fini functions. Now the code is very careful to hold no locks when calling these functions. Thus the dynamic linker cannot be re-entered with a lock already held. Remove the tolerance for recursive locking that I added in revision 1.2 of dllockinit.c. Recursive locking shouldn't happen any more. Mozilla and JDK users: I'd appreciate confirmation that things still work right (or at least the same) with these changes.
2000-01-09 21:13:48 +00:00
}
Pospone the DF_1_NODELETE processing until object DAG is fully loaded. Trying to up the reference from the load loop risks missing dependencies that have not been loaded yet. MFC afer: 1 week Reported by: nox Reviewd by: kib
2012-08-14 13:28:30 +00:00
/*
Implement support -z global linker option. It marks the shared object as always participating in the global symbols namespace, regardless of the way the object was brought into the process address space. Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-04-15 08:16:34 +00:00
* Process all no_delete or global objects here, given
* them own DAGs to prevent their dependencies from being
* unloaded. This has to be done after we have loaded all
* of the dependencies, so that we do not miss any.
Pospone the DF_1_NODELETE processing until object DAG is fully loaded. Trying to up the reference from the load loop risks missing dependencies that have not been loaded yet. MFC afer: 1 week Reported by: nox Reviewd by: kib
2012-08-14 13:28:30 +00:00
*/
Do not call process_nodelete with NULL object pointer. The place where the function is called can be reached if object loading and relocation fails too, in which case obj pointer will be NULL. Do not call process_nodelete then, or crash will follow. Pointy hat to: kan
2012-08-20 23:43:03 +00:00
if (obj != NULL)
Implement support -z global linker option. It marks the shared object as always participating in the global symbols namespace, regardless of the way the object was brought into the process address space. Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-04-15 08:16:34 +00:00
process_z(obj);
Rethink the way we count module references. Simply following DT_NEEDED links is not flexible enough for cases where dynamically loaded modules form a dependency cycle. This should fix an infinite recursion problem encountered by Yahoo. Approved by: re (jhb)
2003-05-08 01:31:36 +00:00
} else {
If dlopen() is called for the dso that has been already loaded as a dependency, then the dso never has its DAG initialized. Empty DAG makes ref_dag() call in dlopen() a nop, and the dso refcount is off by one. Initialize the DAG on the first dlopen() call, using a boolean flag to prevent double initialization. From the PR (edited): Assume we have a library liba.so, containing a function a(), and a library libb.so, containing function b(). liba.so needs functionality from libb.so, so liba.so links in libb.so. An application doesn't know about the relation between these libraries, but needs to call a() and b(). It dlopen()s liba.so and obtains a pointer to a(), then it dlopen()s libb.so and obtains a pointer to b(). As soon as the application doesn't need a() anymore, it dlclose()s liba.so. Expected result: the pointer to b() is still valid and can be called Actual result: the pointer to b() has become invalid, even though the application did not dlclose() the handle to libb.so. On calling b(), the application crashes with a segmentation fault. PR: misc/151861 Based on patch by: jh Reviewed by: kan Tested by: Arjan van Leeuwen <freebsd-maintainer opera com> MFC after: 1 week
2010-11-03 09:23:08 +00:00
/*
* Bump the reference counts for objects on this DAG. If
* this is the first dlopen() call for the object that was
* already loaded as a dependency, initialize the dag
* starting at it.
*/
In r214728, if dlopen() is called for the object that has been already loaded as a dependency and marked -z nodlopen, object' DAG is already initialized by load_needed_objects(). Due to this, the init_dag() call from dlopen() does not increment refcount for the object [1]. Change init_dag() to not increment DAG refcount. Require explicit calls to ref_dag() to increment, and assert that ref_dag() and unref_dag() are called for root that has dag initialized. To fix the noted issue, unconditionally call both init_dag() and ref_dag() in dlopen() for the case when the object was already loaded, making it similar to the case of newly loaded object. Noted by: jh [1] Reviewed by: jh, kan MFC after: 6 days
2010-11-04 09:29:00 +00:00
init_dag(obj);
ref_dag(obj);
Rethink the way we count module references. Simply following DT_NEEDED links is not flexible enough for cases where dynamically loaded modules form a dependency cycle. This should fix an infinite recursion problem encountered by Yahoo. Approved by: re (jhb)
2003-05-08 01:31:36 +00:00
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
if ((lo_flags & RTLD_LO_TRACE) != 0)
Rethink the way we count module references. Simply following DT_NEEDED links is not flexible enough for cases where dynamically loaded modules form a dependency cycle. This should fix an infinite recursion problem encountered by Yahoo. Approved by: re (jhb)
2003-05-08 01:31:36 +00:00
goto trace;
}
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
if (obj != NULL && ((lo_flags & RTLD_LO_NODELETE) != 0 ||
obj->z_nodelete) && !obj->ref_nodel) {
Implement support for RTLD_NODELETE flag for dlopen() and -z nodelete static linker option. Do it by incrementing reference count on the loaded object and its dependencies. Reviewed by: davidxu, kan
2009-03-30 08:47:28 +00:00
dbg("obj %s nodelete", obj->path);
ref_dag(obj);
obj->z_nodelete = obj->ref_nodel = true;
}
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
}
Add GDB support. The method and some of the code came from NetBSD's elf runtime linker.
1998-04-30 07:48:02 +00:00
Add various utrace's for use with ktrace to the ELF runtime linker. To activate the traces, set the LD_UTRACE (or LD_32_UTRACE) environment variable. This also includes code in kdump(8) to parse the traces. Reviewed by: kan, jdp MFC after: 2 weeks
2007-01-09 17:50:05 +00:00
LD_UTRACE(UTRACE_DLOPEN_STOP, obj, NULL, 0, obj ? obj->dl_refcount : 0,
name);
Pass two pointer parameters to the r_debug_state() hook function, thus allowing a debugger or other trace tool to easily grab the addresses of the needed structures off the stack. This change is transparent to gdb, which locates the link_map list and transfers it to debugger memory for comparison purposes. A sample program will be committed showing how this can be used. Reviewed by: John Polstra <jdp@FreeBSD.org>
2000-08-26 05:13:29 +00:00
GDB_STATE(RT_CONSISTENT,obj ? &obj->linkmap : NULL);
Revamp the mechanism for enumerating and calling shared objects' init and fini functions. Now the code is very careful to hold no locks when calling these functions. Thus the dynamic linker cannot be re-entered with a lock already held. Remove the tolerance for recursive locking that I added in revision 1.2 of dllockinit.c. Recursive locking shouldn't happen any more. Mozilla and JDK users: I'd appreciate confirmation that things still work right (or at least the same) with these changes.
2000-01-09 21:13:48 +00:00
Do not try to adjust stacks if dlopen_object is called too early. This is a follow-up to r233231, which fixed similar issue with object initialization code. Reviewed by: kib MFC after: 1 week (with 233231)
2012-04-02 11:50:14 +00:00
if (!(lo_flags & RTLD_LO_EARLY)) {
Propagate the current state of rtld_bind_lock to dlopen_object() calls through the filter loading call chain. This fixes attempts to write-lock the already locked rtld_bind_lock when filter loading is initiated by relocation of dlopening dso. Reported and tested by: Taku YAMAMOTO <taku tackymt homeip net> MFC after: 1 week
2012-04-12 10:32:22 +00:00
map_stacks_exec(lockstate);
Do not try to adjust stacks if dlopen_object is called too early. This is a follow-up to r233231, which fixed similar issue with object initialization code. Reviewed by: kib MFC after: 1 week (with 233231)
2012-04-02 11:50:14 +00:00
}
In rtld, read the initial stack access mode from AT_STACKPROT as set by kernel, and parse PT_GNU_STACK phdr from linked and loaded dsos. If the loaded dso requires executable stack, as specified by PF_X bit of p_flags of PT_GNU_STACK phdr, but current stack protection does not permit execution, the __pthread_map_stacks_exec symbol is looked up and called. It should be implemented in libc or threading library and change the protection mode of all thread stacks to be executable. Provide a private interface _rtld_get_stack_prot() to export the stack access mode as calculated by rtld. Reviewed by: kan
2011-01-08 17:11:49 +00:00
_rtld_bind() read-locks the bind lock, and possible plt resolution from the dispatcher would also acquire bind lock in read mode, which is the supported operation. plt is explicitely designed to allow safe multithreaded updates, so the shared lock do not cause problems. The error in r228435 is that it allows read lock acquisition after the write lock for the bind block. If we dlopened the shared object that contains IRELATIVE or jump slot which target is STT_GNU_IFUNC, then possible recursive plt resolve from the dispatcher would cause it. Postpone the resolution for irelative/ifunc right before initializers are called, and drop bind lock around calls to dispatcher. Use initlist to iterate over the objects instead of the ->next, due to drop of the bind lock in iteration. For i386/reloc.c:reloc_iresolve(), fix calculation of the dispatch function address for dso, by taking into account possible non-zero relocbase. MFC after: 3 weeks
2011-12-14 16:47:53 +00:00
if (initlist_objects_ifunc(&initlist, (mode & RTLD_MODEMASK) == RTLD_NOW,
Fix several problems with our ELF filters implementation. Do not relocate twice an object which happens to be needed by loaded binary (or dso) and some filtee opened due to symbol resolution when relocating need objects. Record the state of the relocation processing in Obj_Entry and short-circuit relocate_objects() if current object already processed. Do not call constructors for filtees loaded during the early relocation processing before image is initialized enough to run user-provided code. Filtees are loaded using dlopen_object(), which normally performs relocation and initialization. If filtee is lazy-loaded during the relocation of dso needed by the main object, dlopen_object() runs too earlier, when most runtime services are not yet ready. Postpone the constructors call to the time when main binary and depended libraries constructors are run, passing the new flag RTLD_LO_EARLY to dlopen_object(). Symbol lookups callers inform symlook_* functions about early stage of initialization with SYMLOOK_EARLY. Pass flags through all functions participating in object relocation. Use the opportunity and fix flags argument to find_symdef() in arch-specific reloc.c to use proper name SYMLOOK_IN_PLT instead of true, which happen to have the same numeric value. Reported and tested by: theraven Reviewed by: kan MFC after: 2 weeks
2012-03-20 13:20:49 +00:00
(lo_flags & RTLD_LO_EARLY) ? SYMLOOK_EARLY : 0,
Propagate the current state of rtld_bind_lock to dlopen_object() calls through the filter loading call chain. This fixes attempts to write-lock the already locked rtld_bind_lock when filter loading is initiated by relocation of dlopening dso. Reported and tested by: Taku YAMAMOTO <taku tackymt homeip net> MFC after: 1 week
2012-04-12 10:32:22 +00:00
lockstate) == -1) {
_rtld_bind() read-locks the bind lock, and possible plt resolution from the dispatcher would also acquire bind lock in read mode, which is the supported operation. plt is explicitely designed to allow safe multithreaded updates, so the shared lock do not cause problems. The error in r228435 is that it allows read lock acquisition after the write lock for the bind block. If we dlopened the shared object that contains IRELATIVE or jump slot which target is STT_GNU_IFUNC, then possible recursive plt resolve from the dispatcher would cause it. Postpone the resolution for irelative/ifunc right before initializers are called, and drop bind lock around calls to dispatcher. Use initlist to iterate over the objects instead of the ->next, due to drop of the bind lock in iteration. For i386/reloc.c:reloc_iresolve(), fix calculation of the dispatch function address for dso, by taking into account possible non-zero relocbase. MFC after: 3 weeks
2011-12-14 16:47:53 +00:00
objlist_clear(&initlist);
Avoid bind lock recursion. When dlclose(3) unloads an object with filtees, it recursively calls dlclose(3) on each filtee in free_needed_filtees(). Introduce dlclose_locked() helper, called from free_needed_filtees() instead of dlclose(), and pass the bind lockstate down to avoid recursing. Reported and tested by: jhibbits Sponsored by: The FreeBSD Foundation MFC after: 1 week
2017-03-09 21:05:47 +00:00
dlopen_cleanup(obj, lockstate);
Propagate the current state of rtld_bind_lock to dlopen_object() calls through the filter loading call chain. This fixes attempts to write-lock the already locked rtld_bind_lock when filter loading is initiated by relocation of dlopening dso. Reported and tested by: Taku YAMAMOTO <taku tackymt homeip net> MFC after: 1 week
2012-04-12 10:32:22 +00:00
if (lockstate == &mlockstate)
lock_release(rtld_bind_lock, lockstate);
_rtld_bind() read-locks the bind lock, and possible plt resolution from the dispatcher would also acquire bind lock in read mode, which is the supported operation. plt is explicitely designed to allow safe multithreaded updates, so the shared lock do not cause problems. The error in r228435 is that it allows read lock acquisition after the write lock for the bind block. If we dlopened the shared object that contains IRELATIVE or jump slot which target is STT_GNU_IFUNC, then possible recursive plt resolve from the dispatcher would cause it. Postpone the resolution for irelative/ifunc right before initializers are called, and drop bind lock around calls to dispatcher. Use initlist to iterate over the objects instead of the ->next, due to drop of the bind lock in iteration. For i386/reloc.c:reloc_iresolve(), fix calculation of the dispatch function address for dso, by taking into account possible non-zero relocbase. MFC after: 3 weeks
2011-12-14 16:47:53 +00:00
return (NULL);
}
Fix several problems with our ELF filters implementation. Do not relocate twice an object which happens to be needed by loaded binary (or dso) and some filtee opened due to symbol resolution when relocating need objects. Record the state of the relocation processing in Obj_Entry and short-circuit relocate_objects() if current object already processed. Do not call constructors for filtees loaded during the early relocation processing before image is initialized enough to run user-provided code. Filtees are loaded using dlopen_object(), which normally performs relocation and initialization. If filtee is lazy-loaded during the relocation of dso needed by the main object, dlopen_object() runs too earlier, when most runtime services are not yet ready. Postpone the constructors call to the time when main binary and depended libraries constructors are run, passing the new flag RTLD_LO_EARLY to dlopen_object(). Symbol lookups callers inform symlook_* functions about early stage of initialization with SYMLOOK_EARLY. Pass flags through all functions participating in object relocation. Use the opportunity and fix flags argument to find_symdef() in arch-specific reloc.c to use proper name SYMLOOK_IN_PLT instead of true, which happen to have the same numeric value. Reported and tested by: theraven Reviewed by: kan MFC after: 2 weeks
2012-03-20 13:20:49 +00:00
if (!(lo_flags & RTLD_LO_EARLY)) {
/* Call the init functions. */
Propagate the current state of rtld_bind_lock to dlopen_object() calls through the filter loading call chain. This fixes attempts to write-lock the already locked rtld_bind_lock when filter loading is initiated by relocation of dlopening dso. Reported and tested by: Taku YAMAMOTO <taku tackymt homeip net> MFC after: 1 week
2012-04-12 10:32:22 +00:00
objlist_call_init(&initlist, lockstate);
Fix several problems with our ELF filters implementation. Do not relocate twice an object which happens to be needed by loaded binary (or dso) and some filtee opened due to symbol resolution when relocating need objects. Record the state of the relocation processing in Obj_Entry and short-circuit relocate_objects() if current object already processed. Do not call constructors for filtees loaded during the early relocation processing before image is initialized enough to run user-provided code. Filtees are loaded using dlopen_object(), which normally performs relocation and initialization. If filtee is lazy-loaded during the relocation of dso needed by the main object, dlopen_object() runs too earlier, when most runtime services are not yet ready. Postpone the constructors call to the time when main binary and depended libraries constructors are run, passing the new flag RTLD_LO_EARLY to dlopen_object(). Symbol lookups callers inform symlook_* functions about early stage of initialization with SYMLOOK_EARLY. Pass flags through all functions participating in object relocation. Use the opportunity and fix flags argument to find_symdef() in arch-specific reloc.c to use proper name SYMLOOK_IN_PLT instead of true, which happen to have the same numeric value. Reported and tested by: theraven Reviewed by: kan MFC after: 2 weeks
2012-03-20 13:20:49 +00:00
}
Revamp the code that calls shared libraries' init and fini functions. Formerly the init functions were called in the opposite of the order in which libraries were loaded, and libraries were loaded according to a breadth-first traversal of the dependency graph. That ordering came from SVR4.0, and it was easy to implement but not always sensible. Now we do a depth-first walk over the dependency graph and call the init functions in an order such that each shared object's needed objects are initialized before the shared object itself. At the same time we build a list of finalization (fini) functions in the opposite order, to guarantee correct C++ destructor ordering whenever possible. (It may not be possible if dlopen and dlclose are used in strange ways, but we come as close as one can come.) The need for this renovation has become apparent as more programs have started using multithreading. The multithreaded C library libc_r requires initialization, whereas the standard libc does not. Since virtually every other object depends on the C library, it is important that it get initialized first.
2000-07-26 04:24:40 +00:00
objlist_clear(&initlist);
Propagate the current state of rtld_bind_lock to dlopen_object() calls through the filter loading call chain. This fixes attempts to write-lock the already locked rtld_bind_lock when filter loading is initiated by relocation of dlopening dso. Reported and tested by: Taku YAMAMOTO <taku tackymt homeip net> MFC after: 1 week
2012-04-12 10:32:22 +00:00
if (lockstate == &mlockstate)
lock_release(rtld_bind_lock, lockstate);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
return obj;
Fix a problem with RTLD_TRACE flag to dlopen(3), which sometimes can return even if there was no error occured (when trying to dlopen(3) object that already linked into executable which does dlopen(3) call). This is more proper fix for `ldd /usr/lib/libc.so' problem, because the new behaviour conforms to documentation. Remove workaround from ldd.c (rev.1.32). PR: 35099 Submitted by: Nathan Hawkins <utsl@quic.net> MFC after: 1 week
2002-10-19 10:18:29 +00:00
trace:
trace_loaded_objects(obj);
Propagate the current state of rtld_bind_lock to dlopen_object() calls through the filter loading call chain. This fixes attempts to write-lock the already locked rtld_bind_lock when filter loading is initiated by relocation of dlopening dso. Reported and tested by: Taku YAMAMOTO <taku tackymt homeip net> MFC after: 1 week
2012-04-12 10:32:22 +00:00
if (lockstate == &mlockstate)
lock_release(rtld_bind_lock, lockstate);
Fix a problem with RTLD_TRACE flag to dlopen(3), which sometimes can return even if there was no error occured (when trying to dlopen(3) object that already linked into executable which does dlopen(3) call). This is more proper fix for `ldd /usr/lib/libc.so' problem, because the new behaviour conforms to documentation. Remove workaround from ldd.c (rev.1.32). PR: 35099 Submitted by: Nathan Hawkins <utsl@quic.net> MFC after: 1 week
2002-10-19 10:18:29 +00:00
exit(0);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
}
Implement ELF symbol versioning using GNU semantics. This code aims to be compatible with symbol versioning support as implemented by GNU libc and documented by http://people.redhat.com/~drepper/symbol-versioning and LSB 3.0. Implement dlvsym() function to allow lookups for a specific version of a given symbol.
2005-12-18 19:43:33 +00:00
static void *
do_dlsym(void *handle, const char *name, void *retaddr, const Ver_Entry *ve,
int flags)
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
{
Clean up white spaces and fix style(9).
2006-09-19 16:48:08 +00:00
DoneList donelist;
const Obj_Entry *obj, *defobj;
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
const Elf_Sym *def;
SymLook req;
RtldLockState lockstate;
Import DragonFly BSD commit From: Sascha Wildner <saw@online.de> Date: Fri, 2 Mar 2012 09:15:56 +0000 (+0100) Subject: rtld: Add a special case in do_dlsym() for TLS stored symbols. X-Git-Url: http://gitweb.dragonflybsd.org/dragonfly.git/commitdiff_plain/1388aaafe730c85693936aaf9bc6d83fc2d981be?hp=bca4412595a8979ab9f1bf36068c806ce88a667a rtld: Add a special case in do_dlsym() for TLS stored symbols. Submitted-by: Markus Pfeiffer <markus.pfeiffer@morphism.de> Discussed with: kan MFC after: 1 week
2012-03-29 10:32:34 +00:00
tls_index ti;
Allow tracing dlfunc() / dlsym() events. MFC after: 1 week
2015-01-25 12:11:50 +00:00
void *sym;
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
int res;
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
Handle dlsym(NULL, ...) properly, by searching in the caller's shared object. Note, this searches _only_ that object, and not its needed objects, in accordance with the documentation. Also fix dlopen(NULL, ...) so that the executable's needed objects are searched as well as the executable itself.
1998-09-02 01:09:34 +00:00
def = NULL;
Revamp the symbol lookup algorithm to cope better with objects loaded separately by dlopen that have global symbols with identical names. Viewing each dlopened object as a DAG which is linked by its DT_NEEDED entries in the dynamic table, the search order is as follows: * If the referencing object was linked with -Bsymbolic, search it internally. * Search all dlopened DAGs containing the referencing object. * Search all objects loaded at program start up. * Search all objects which were dlopened() using the RTLD_GLOBAL flag (which is now supported too). The search terminates as soon as a strong definition is found. Lacking that, the first weak definition is used. These rules match those of Solaris, as best I could determine them from its vague manual pages and the results of experiments I performed. PR: misc/12438
1999-08-30 01:48:19 +00:00
defobj = NULL;
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
symlook_init(&req, name);
req.ventry = ve;
req.flags = flags | SYMLOOK_IN_PLT;
req.lockstate = &lockstate;
Allow tracing dlfunc() / dlsym() events. MFC after: 1 week
2015-01-25 12:11:50 +00:00
LD_UTRACE(UTRACE_DLSYM_START, handle, NULL, 0, 0, name);
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
rlock_acquire(rtld_bind_lock, &lockstate);
Use sigsetjmp/siglongjmp with disabled signal mask access for lock upgrade in rtld. There is no need to care about the mask, which causes a lot of unneeded sigprocmask(2) calls during each symbol lookup.
2011-02-09 09:20:27 +00:00
if (sigsetjmp(lockstate.env, 0) != 0)
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
lock_upgrade(rtld_bind_lock, &lockstate);
Implement dlinfo() function. Introdice RTLD_SELF special handle and properly process it within dlsym() and dlinfo() functions. The intention is to improve our compatibility with Solaris and to make a Java port easier. Partially submitted by: phantom
2003-02-13 17:47:44 +00:00
if (handle == NULL || handle == RTLD_NEXT ||
handle == RTLD_DEFAULT || handle == RTLD_SELF) {
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
if ((obj = obj_from_addr(retaddr)) == NULL) {
_rtld_error("Cannot determine caller's shared object");
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
lock_release(rtld_bind_lock, &lockstate);
Allow tracing dlfunc() / dlsym() events. MFC after: 1 week
2015-01-25 12:11:50 +00:00
LD_UTRACE(UTRACE_DLSYM_STOP, handle, NULL, 0, 0, name);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
return NULL;
}
Revamp the symbol lookup algorithm to cope better with objects loaded separately by dlopen that have global symbols with identical names. Viewing each dlopened object as a DAG which is linked by its DT_NEEDED entries in the dynamic table, the search order is as follows: * If the referencing object was linked with -Bsymbolic, search it internally. * Search all dlopened DAGs containing the referencing object. * Search all objects loaded at program start up. * Search all objects which were dlopened() using the RTLD_GLOBAL flag (which is now supported too). The search terminates as soon as a strong definition is found. Lacking that, the first weak definition is used. These rules match those of Solaris, as best I could determine them from its vague manual pages and the results of experiments I performed. PR: misc/12438
1999-08-30 01:48:19 +00:00
if (handle == NULL) { /* Just the caller's shared object. */
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
res = symlook_obj(&req, obj);
if (res == 0) {
def = req.sym_out;
defobj = req.defobj_out;
}
Implement dlinfo() function. Introdice RTLD_SELF special handle and properly process it within dlsym() and dlinfo() functions. The intention is to improve our compatibility with Solaris and to make a Java port easier. Partially submitted by: phantom
2003-02-13 17:47:44 +00:00
} else if (handle == RTLD_NEXT || /* Objects after caller's */
handle == RTLD_SELF) { /* ... caller included */
if (handle == RTLD_NEXT)
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
obj = globallist_next(obj);
Fix dlsym(RTLD_NEXT) handling to only return the next library in last library cases. The root of the problem here is that TAILQ_FOREACH_FROM will default to the head of the list if passed NULL, which will be the case if there are no libraries loaded after this one. Thus all libraries, including the current, were iterated in that case rather than none. This was broken in r294373. Reviewed by: markj (earlier version), cem, kib, ngie MFC after: 1 week Sponsored by: EMC / Isilon Storage Division Differential Revision: https://reviews.freebsd.org/D7216
2016-07-15 19:07:00 +00:00
for (; obj != NULL; obj = TAILQ_NEXT(obj, next)) {
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
if (obj->marker)
continue;
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
res = symlook_obj(&req, obj);
if (res == 0) {
if (def == NULL ||
ELF_ST_BIND(req.sym_out->st_info) != STB_WEAK) {
def = req.sym_out;
defobj = req.defobj_out;
Allow strong symbols to override weak ones for lookups done through dlsym with RTLD_NEXT/RTLD_SELF handles. Allow symbols from ld-elf.so to be located this way too. Based on report and original patch from sobomax@.
2008-10-10 00:16:32 +00:00
if (ELF_ST_BIND(def->st_info) != STB_WEAK)
break;
}
}
}
/*
* Search the dynamic linker itself, and possibly resolve the
* symbol from there. This is how the application links to
Remove exports table. Export control by the version script is enough. Reviewed by: kan MFC after: 3 weeks
2010-08-24 13:01:14 +00:00
* dynamic linker services such as dlopen.
Allow strong symbols to override weak ones for lookups done through dlsym with RTLD_NEXT/RTLD_SELF handles. Allow symbols from ld-elf.so to be located this way too. Based on report and original patch from sobomax@.
2008-10-10 00:16:32 +00:00
*/
if (def == NULL || ELF_ST_BIND(def->st_info) == STB_WEAK) {
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
res = symlook_obj(&req, &obj_rtld);
if (res == 0) {
def = req.sym_out;
defobj = req.defobj_out;
Revamp the symbol lookup algorithm to cope better with objects loaded separately by dlopen that have global symbols with identical names. Viewing each dlopened object as a DAG which is linked by its DT_NEEDED entries in the dynamic table, the search order is as follows: * If the referencing object was linked with -Bsymbolic, search it internally. * Search all dlopened DAGs containing the referencing object. * Search all objects loaded at program start up. * Search all objects which were dlopened() using the RTLD_GLOBAL flag (which is now supported too). The search terminates as soon as a strong definition is found. Lacking that, the first weak definition is used. These rules match those of Solaris, as best I could determine them from its vague manual pages and the results of experiments I performed. PR: misc/12438
1999-08-30 01:48:19 +00:00
}
}
Add support for dlsym(RTLD_DEFAULT, ...).
2000-09-19 04:27:16 +00:00
} else {
assert(handle == RTLD_DEFAULT);
Fix an apparent cop-and-paste mistake in previous commit. This makes dlsym(RTLD_DEFAULT) work properly again.
2010-12-27 00:30:29 +00:00
res = symlook_default(&req, obj);
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
if (res == 0) {
defobj = req.defobj_out;
def = req.sym_out;
}
Handle dlsym(NULL, ...) properly, by searching in the caller's shared object. Note, this searches _only_ that object, and not its needed objects, in accordance with the documentation. Also fix dlopen(NULL, ...) so that the executable's needed objects are searched as well as the executable itself.
1998-09-02 01:09:34 +00:00
}
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
} else {
Add a new function dllockinit() for registering thread locking functions to be used by the dynamic linker. This can be called by threads packages at start-up time. I will add the call to libc_r soon. Also add a default locking method that is used up until dllockinit() is called. The default method works by blocking SIGVTALRM, SIGPROF, and SIGALRM in critical sections. It is based on the observation that most user-space threads packages implement thread preemption with one of these signals (usually SIGVTALRM). The dynamic linker has never been reentrant, but it became less reentrant in revision 1.34 of "src/libexec/rtld-elf/rtld.c". Starting with that revision, multiple threads each doing lazy binding could interfere with each other. The usual symptom was that a symbol was falsely reported as undefined at start-up time. It was rare but not unseen. This commit fixes it.
1999-12-27 04:44:04 +00:00
if ((obj = dlcheck(handle)) == NULL) {
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
lock_release(rtld_bind_lock, &lockstate);
Allow tracing dlfunc() / dlsym() events. MFC after: 1 week
2015-01-25 12:11:50 +00:00
LD_UTRACE(UTRACE_DLSYM_STOP, handle, NULL, 0, 0, name);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
return NULL;
Add a new function dllockinit() for registering thread locking functions to be used by the dynamic linker. This can be called by threads packages at start-up time. I will add the call to libc_r soon. Also add a default locking method that is used up until dllockinit() is called. The default method works by blocking SIGVTALRM, SIGPROF, and SIGALRM in critical sections. It is based on the observation that most user-space threads packages implement thread preemption with one of these signals (usually SIGVTALRM). The dynamic linker has never been reentrant, but it became less reentrant in revision 1.34 of "src/libexec/rtld-elf/rtld.c". Starting with that revision, multiple threads each doing lazy binding could interfere with each other. The usual symptom was that a symbol was falsely reported as undefined at start-up time. It was rare but not unseen. This commit fixes it.
1999-12-27 04:44:04 +00:00
}
Handle dlsym(NULL, ...) properly, by searching in the caller's shared object. Note, this searches _only_ that object, and not its needed objects, in accordance with the documentation. Also fix dlopen(NULL, ...) so that the executable's needed objects are searched as well as the executable itself.
1998-09-02 01:09:34 +00:00
When looking up the symbol by dlsym, look it not only in the object given as dso handle, but also in the implicit dependencies of that dso. Also, const-ify the read-only parameter objlist of symlook_list. Reported by: "Simon 'corecode' Schubert" <corecode at fs ei tum de> Approved by: kan (mentor) X-MFC-After: 6.2
2006-09-08 14:59:54 +00:00
donelist_init(&donelist);
Handle dlsym(NULL, ...) properly, by searching in the caller's shared object. Note, this searches _only_ that object, and not its needed objects, in accordance with the documentation. Also fix dlopen(NULL, ...) so that the executable's needed objects are searched as well as the executable itself.
1998-09-02 01:09:34 +00:00
if (obj->mainprog) {
Eliminate the use of symlook_needed function in favor of DAGS. Place elements on DAG lists in breadth-first order. This allows us to walk pre-built list in all cases where breadth-first dependency chain enumeration is required. Fix dlsym on special handle obtained by dlopen(NULL, ...) to do what comment claims it does. Take advantage of recently added symlook_global function to iterate over main objects and global DAGs lists properly in search of a symbol. Since rtld itself provides part of the global namespace, search rtld_obj too. Remove recursion from init_dag and symlook_needed functions. Use symlook_needed for ELF filtee processing only and change lookup order used in the function to match the order used by Solaris runtime linker under same circumstances. While there, fix weak symbol handling in the loop so that we return the first weak symbol definition if no strong one was found, instead of the last one. Reviewed by: kib MFC after: 1 month
2011-01-28 23:44:57 +00:00
/* Handle obtained by dlopen(NULL, ...) implies global scope. */
res = symlook_global(&req, &donelist);
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
if (res == 0) {
def = req.sym_out;
defobj = req.defobj_out;
Eliminate the use of symlook_needed function in favor of DAGS. Place elements on DAG lists in breadth-first order. This allows us to walk pre-built list in all cases where breadth-first dependency chain enumeration is required. Fix dlsym on special handle obtained by dlopen(NULL, ...) to do what comment claims it does. Take advantage of recently added symlook_global function to iterate over main objects and global DAGs lists properly in search of a symbol. Since rtld itself provides part of the global namespace, search rtld_obj too. Remove recursion from init_dag and symlook_needed functions. Use symlook_needed for ELF filtee processing only and change lookup order used in the function to match the order used by Solaris runtime linker under same circumstances. While there, fix weak symbol handling in the loop so that we return the first weak symbol definition if no strong one was found, instead of the last one. Reviewed by: kib MFC after: 1 month
2011-01-28 23:44:57 +00:00
}
/*
* 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.
*/
if (def == NULL || ELF_ST_BIND(def->st_info) == STB_WEAK) {
res = symlook_obj(&req, &obj_rtld);
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
if (res == 0) {
def = req.sym_out;
defobj = req.defobj_out;
}
}
Eliminate the use of symlook_needed function in favor of DAGS. Place elements on DAG lists in breadth-first order. This allows us to walk pre-built list in all cases where breadth-first dependency chain enumeration is required. Fix dlsym on special handle obtained by dlopen(NULL, ...) to do what comment claims it does. Take advantage of recently added symlook_global function to iterate over main objects and global DAGs lists properly in search of a symbol. Since rtld itself provides part of the global namespace, search rtld_obj too. Remove recursion from init_dag and symlook_needed functions. Use symlook_needed for ELF filtee processing only and change lookup order used in the function to match the order used by Solaris runtime linker under same circumstances. While there, fix weak symbol handling in the loop so that we return the first weak symbol definition if no strong one was found, instead of the last one. Reviewed by: kib MFC after: 1 month
2011-01-28 23:44:57 +00:00
}
else {
When looking up the symbol by dlsym, look it not only in the object given as dso handle, but also in the implicit dependencies of that dso. Also, const-ify the read-only parameter objlist of symlook_list. Reported by: "Simon 'corecode' Schubert" <corecode at fs ei tum de> Approved by: kan (mentor) X-MFC-After: 6.2
2006-09-08 14:59:54 +00:00
/* Search the whole DAG rooted at the given object. */
Eliminate the use of symlook_needed function in favor of DAGS. Place elements on DAG lists in breadth-first order. This allows us to walk pre-built list in all cases where breadth-first dependency chain enumeration is required. Fix dlsym on special handle obtained by dlopen(NULL, ...) to do what comment claims it does. Take advantage of recently added symlook_global function to iterate over main objects and global DAGs lists properly in search of a symbol. Since rtld itself provides part of the global namespace, search rtld_obj too. Remove recursion from init_dag and symlook_needed functions. Use symlook_needed for ELF filtee processing only and change lookup order used in the function to match the order used by Solaris runtime linker under same circumstances. While there, fix weak symbol handling in the loop so that we return the first weak symbol definition if no strong one was found, instead of the last one. Reviewed by: kib MFC after: 1 month
2011-01-28 23:44:57 +00:00
res = symlook_list(&req, &obj->dagmembers, &donelist);
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
if (res == 0) {
def = req.sym_out;
defobj = req.defobj_out;
}
Handle dlsym(NULL, ...) properly, by searching in the caller's shared object. Note, this searches _only_ that object, and not its needed objects, in accordance with the documentation. Also fix dlopen(NULL, ...) so that the executable's needed objects are searched as well as the executable itself.
1998-09-02 01:09:34 +00:00
}
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
}
Add a new function dllockinit() for registering thread locking functions to be used by the dynamic linker. This can be called by threads packages at start-up time. I will add the call to libc_r soon. Also add a default locking method that is used up until dllockinit() is called. The default method works by blocking SIGVTALRM, SIGPROF, and SIGALRM in critical sections. It is based on the observation that most user-space threads packages implement thread preemption with one of these signals (usually SIGVTALRM). The dynamic linker has never been reentrant, but it became less reentrant in revision 1.34 of "src/libexec/rtld-elf/rtld.c". Starting with that revision, multiple threads each doing lazy binding could interfere with each other. The usual symptom was that a symbol was falsely reported as undefined at start-up time. It was rare but not unseen. This commit fixes it.
1999-12-27 04:44:04 +00:00
if (def != NULL) {
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
lock_release(rtld_bind_lock, &lockstate);
Add ia64 support. Various adjustments were made to existing targets to cope with a few interface changes required by the ia64. In particular, function pointers on ia64 need special treatment in rtld.
2001-10-15 18:48:42 +00:00
/*
* The value required by the caller is derived from the value
Remove ia64. This includes: o All directories named *ia64* o All files named *ia64* o All ia64-specific code guarded by __ia64__ o All ia64-specific makefile logic o Mention of ia64 in comments and documentation This excludes: o Everything under contrib/ o Everything under crypto/ o sys/xen/interface o sys/sys/elf_common.h Discussed at: BSDcan
2014-07-07 00:27:09 +00:00
* of the symbol. this is simply the relocated value of the
* symbol.
Add ia64 support. Various adjustments were made to existing targets to cope with a few interface changes required by the ia64. In particular, function pointers on ia64 need special treatment in rtld.
2001-10-15 18:48:42 +00:00
*/
if (ELF_ST_TYPE(def->st_info) == STT_FUNC)
Allow tracing dlfunc() / dlsym() events. MFC after: 1 week
2015-01-25 12:11:50 +00:00
sym = make_function_pointer(def, defobj);
Add support for STT_GNU_IFUNC and R_MACHINE_IRELATIVE GNU extensions to rtld on 386 and amd64. This adds runtime bits neccessary for the use of the dispatch functions from the dynamically-linked executables and shared libraries. To allow use of external references from the dispatch function, resolution of the R_MACHINE_IRESOLVE relocations in PLT is postponed until GOT entries for PLT are prepared, and normal resolution of the GOT entries is finished. Similar to how it is done by GNU, IRELATIVE relocations are resolved in advance, instead of normal lazy handling for PLT. Move the init_pltgot() call before the relocations for the object are processed. MFC after: 3 weeks
2011-12-12 11:03:14 +00:00
else if (ELF_ST_TYPE(def->st_info) == STT_GNU_IFUNC)
Allow tracing dlfunc() / dlsym() events. MFC after: 1 week
2015-01-25 12:11:50 +00:00
sym = rtld_resolve_ifunc(defobj, def);
Import DragonFly BSD commit From: Sascha Wildner <saw@online.de> Date: Fri, 2 Mar 2012 09:15:56 +0000 (+0100) Subject: rtld: Add a special case in do_dlsym() for TLS stored symbols. X-Git-Url: http://gitweb.dragonflybsd.org/dragonfly.git/commitdiff_plain/1388aaafe730c85693936aaf9bc6d83fc2d981be?hp=bca4412595a8979ab9f1bf36068c806ce88a667a rtld: Add a special case in do_dlsym() for TLS stored symbols. Submitted-by: Markus Pfeiffer <markus.pfeiffer@morphism.de> Discussed with: kan MFC after: 1 week
2012-03-29 10:32:34 +00:00
else if (ELF_ST_TYPE(def->st_info) == STT_TLS) {
ti.ti_module = defobj->tlsindex;
ti.ti_offset = def->st_value;
Allow tracing dlfunc() / dlsym() events. MFC after: 1 week
2015-01-25 12:11:50 +00:00
sym = __tls_get_addr(&ti);
Import DragonFly BSD commit From: Sascha Wildner <saw@online.de> Date: Fri, 2 Mar 2012 09:15:56 +0000 (+0100) Subject: rtld: Add a special case in do_dlsym() for TLS stored symbols. X-Git-Url: http://gitweb.dragonflybsd.org/dragonfly.git/commitdiff_plain/1388aaafe730c85693936aaf9bc6d83fc2d981be?hp=bca4412595a8979ab9f1bf36068c806ce88a667a rtld: Add a special case in do_dlsym() for TLS stored symbols. Submitted-by: Markus Pfeiffer <markus.pfeiffer@morphism.de> Discussed with: kan MFC after: 1 week
2012-03-29 10:32:34 +00:00
} else
Allow tracing dlfunc() / dlsym() events. MFC after: 1 week
2015-01-25 12:11:50 +00:00
sym = defobj->relocbase + def->st_value;
LD_UTRACE(UTRACE_DLSYM_STOP, handle, sym, 0, 0, name);
return (sym);
Add a new function dllockinit() for registering thread locking functions to be used by the dynamic linker. This can be called by threads packages at start-up time. I will add the call to libc_r soon. Also add a default locking method that is used up until dllockinit() is called. The default method works by blocking SIGVTALRM, SIGPROF, and SIGALRM in critical sections. It is based on the observation that most user-space threads packages implement thread preemption with one of these signals (usually SIGVTALRM). The dynamic linker has never been reentrant, but it became less reentrant in revision 1.34 of "src/libexec/rtld-elf/rtld.c". Starting with that revision, multiple threads each doing lazy binding could interfere with each other. The usual symptom was that a symbol was falsely reported as undefined at start-up time. It was rare but not unseen. This commit fixes it.
1999-12-27 04:44:04 +00:00
}
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
When reporting undefined symbol, note the version, if specified. Use the standard syntax of name@version, I do not expect a confusion due to unlikely possibility of the name containing the '@' character. Requested by: emaste Sponsored by: The FreeBSD Foundation MFC after: 1 week
2017-07-04 20:19:36 +00:00
_rtld_error("Undefined symbol \"%s%s%s\"", name, ve != NULL ? "@" : "",
ve != NULL ? ve->name : "");
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
lock_release(rtld_bind_lock, &lockstate);
Allow tracing dlfunc() / dlsym() events. MFC after: 1 week
2015-01-25 12:11:50 +00:00
LD_UTRACE(UTRACE_DLSYM_STOP, handle, NULL, 0, 0, name);
Import the ELF dynamic linker. This is the ElfKit version with quite a few enhancements and bug fixes. There are still some known deficiencies, but it should be adequate to get us started with ELF. Submitted by: John Polstra <jdp@polstra.com>
1998-03-07 19:24:35 +00:00
return NULL;
}
Implement ELF symbol versioning using GNU semantics. This code aims to be compatible with symbol versioning support as implemented by GNU libc and documented by http://people.redhat.com/~drepper/symbol-versioning and LSB 3.0. Implement dlvsym() function to allow lookups for a specific version of a given symbol.
2005-12-18 19:43:33 +00:00
void *
dlsym(void *handle, const char *name)
{
return do_dlsym(handle, name, __builtin_return_address(0), NULL,
SYMLOOK_DLSYM);
}
Allow the NULL, RTLD_SELF and RTLD_NEXT handles to work with dlfunc(3). dlfunc() called dlsym() to do the work, and dlsym() determines the dso that originating the call by the return address. Due to this, dlfunc() operated as if the caller is always the libc. To fix this, move the dlfunc() to rtld, where it can call the internal implementation of dlsym, and still correctly fetch return address. Provide usual weak stub for the symbol from libc for static binaries. dlfunc is put to FBSD_1.0 symver namespace in the ld.so export to override dlfunc@FBSD_1.0 weak symbol, exported by libc. Reported, analyzed and tested by: Tijl Coosemans <tijl ulyssis org> PR: standards/133339 Reviewed by: kan
2009-04-03 19:17:23 +00:00
dlfunc_t
dlfunc(void *handle, const char *name)
{
union {
void *d;
dlfunc_t f;
} rv;
rv.d = do_dlsym(handle, name, __builtin_return_address(0), NULL,
SYMLOOK_DLSYM);
return (rv.f);
}
Implement ELF symbol versioning using GNU semantics. This code aims to be compatible with symbol versioning support as implemented by GNU libc and documented by http://people.redhat.com/~drepper/symbol-versioning and LSB 3.0. Implement dlvsym() function to allow lookups for a specific version of a given symbol.
2005-12-18 19:43:33 +00:00
void *
dlvsym(void *handle, const char *name, const char *version)
{
Ver_Entry ventry;
ventry.name = version;
ventry.file = NULL;
ventry.hash = elf_hash(version);
ventry.flags= 0;
return do_dlsym(handle, name, __builtin_return_address(0), &ventry,
SYMLOOK_DLSYM);
}
Introduce implementation-private rtld interface _rtld_addr_phdr, which fills struct dl_phdr_info for the shared object that contains the specified address, if any. Idea and reviewed by: kan MFC after: 3 weeks
2010-08-23 15:27:03 +00:00
int
_rtld_addr_phdr(const void *addr, struct dl_phdr_info *phdr_info)
{
const Obj_Entry *obj;
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
RtldLockState lockstate;
Introduce implementation-private rtld interface _rtld_addr_phdr, which fills struct dl_phdr_info for the shared object that contains the specified address, if any. Idea and reviewed by: kan MFC after: 3 weeks
2010-08-23 15:27:03 +00:00
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
rlock_acquire(rtld_bind_lock, &lockstate);
Introduce implementation-private rtld interface _rtld_addr_phdr, which fills struct dl_phdr_info for the shared object that contains the specified address, if any. Idea and reviewed by: kan MFC after: 3 weeks
2010-08-23 15:27:03 +00:00
obj = obj_from_addr(addr);
if (obj == NULL) {
_rtld_error("No shared object contains address");
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
lock_release(rtld_bind_lock, &lockstate);
Introduce implementation-private rtld interface _rtld_addr_phdr, which fills struct dl_phdr_info for the shared object that contains the specified address, if any. Idea and reviewed by: kan MFC after: 3 weeks
2010-08-23 15:27:03 +00:00
return (0);
}
rtld_fill_dl_phdr_info(obj, phdr_info);
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
lock_release(rtld_bind_lock, &lockstate);
Introduce implementation-private rtld interface _rtld_addr_phdr, which fills struct dl_phdr_info for the shared object that contains the specified address, if any. Idea and reviewed by: kan MFC after: 3 weeks
2010-08-23 15:27:03 +00:00
return (1);
}
- Added dladdr(3) support. Reviewed by: jdp@FreeBSD.org
1999-03-24 23:37:35 +00:00
int
dladdr(const void *addr, Dl_info *info)
{
const Obj_Entry *obj;
const Elf_Sym *def;
void *symbol_addr;
unsigned long symoffset;
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
RtldLockState lockstate;
Allow threading libraries to register their own locking implementation in case default one provided by rtld is not suitable. Consolidate various identical MD lock implementation into a single file using appropriate machine/atomic.h. Approved by: re (scottl)
2003-05-29 22:58:26 +00:00
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
rlock_acquire(rtld_bind_lock, &lockstate);
- Commit the correct dladdr() implementation. Reviewed by: jdp@FreeBSD.org <This is the version he reviewed!>
1999-03-24 23:47:29 +00:00
obj = obj_from_addr(addr);
- Added dladdr(3) support. Reviewed by: jdp@FreeBSD.org
1999-03-24 23:37:35 +00:00
if (obj == NULL) {
- Commit the correct dladdr() implementation. Reviewed by: jdp@FreeBSD.org <This is the version he reviewed!>
1999-03-24 23:47:29 +00:00
_rtld_error("No shared object contains address");
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
lock_release(rtld_bind_lock, &lockstate);
- Added dladdr(3) support. Reviewed by: jdp@FreeBSD.org
1999-03-24 23:37:35 +00:00
return 0;
}
info->dli_fname = obj->path;
- Commit the correct dladdr() implementation. Reviewed by: jdp@FreeBSD.org <This is the version he reviewed!>
1999-03-24 23:47:29 +00:00
info->dli_fbase = obj->mapbase;
- Added dladdr(3) support. Reviewed by: jdp@FreeBSD.org
1999-03-24 23:37:35 +00:00
info->dli_saddr = (void *)0;
info->dli_sname = NULL;
/*
Fix a couple of typos in comments.
1999-04-07 02:48:43 +00:00
* Walk the symbol list looking for the symbol whose address is
- Added dladdr(3) support. Reviewed by: jdp@FreeBSD.org
1999-03-24 23:37:35 +00:00
* closest to the address sent in.
*/
Add GNU hash support for rtld. Based on dragonflybsd support for GNU hash by John Marino <draco marino st> Reviewed by: kan Tested by: bapt MFC after: 2 weeks
2012-04-30 13:31:10 +00:00
for (symoffset = 0; symoffset < obj->dynsymcount; symoffset++) {
- Added dladdr(3) support. Reviewed by: jdp@FreeBSD.org
1999-03-24 23:37:35 +00:00
def = obj->symtab + symoffset;
- Commit the correct dladdr() implementation. Reviewed by: jdp@FreeBSD.org <This is the version he reviewed!>
1999-03-24 23:47:29 +00:00
/*
* For skip the symbol if st_shndx is either SHN_UNDEF or
* SHN_COMMON.
*/
if (def->st_shndx == SHN_UNDEF || def->st_shndx == SHN_COMMON)
continue;
- Added dladdr(3) support. Reviewed by: jdp@FreeBSD.org
1999-03-24 23:37:35 +00:00
/*
Fix a couple of typos in comments.
1999-04-07 02:48:43 +00:00
* If the symbol is greater than the specified address, or if it
- Added dladdr(3) support. Reviewed by: jdp@FreeBSD.org
1999-03-24 23:37:35 +00:00
* is further away from addr than the current nearest symbol,
* then reject it.
*/
- Commit the correct dladdr() implementation. Reviewed by: jdp@FreeBSD.org <This is the version he reviewed!>
1999-03-24 23:47:29 +00:00
symbol_addr = obj->relocbase + def->st_value;
if (symbol_addr > addr || symbol_addr < info->dli_saddr)
- Added dladdr(3) support. Reviewed by: jdp@FreeBSD.org
1999-03-24 23:37:35 +00:00
continue;
/* Update our idea of the nearest symbol. */
info->dli_sname = obj->strtab + def->st_name;
info->dli_saddr = symbol_addr;
/* Exact match? */
if (info->dli_saddr == addr)
break;
}
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
lock_release(rtld_bind_lock, &lockstate);
- Added dladdr(3) support. Reviewed by: jdp@FreeBSD.org
1999-03-24 23:37:35 +00:00
return 1;
}
Implement dlinfo() function. Introdice RTLD_SELF special handle and properly process it within dlsym() and dlinfo() functions. The intention is to improve our compatibility with Solaris and to make a Java port easier. Partially submitted by: phantom
2003-02-13 17:47:44 +00:00
int
dlinfo(void *handle, int request, void *p)
{
const Obj_Entry *obj;
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
RtldLockState lockstate;
int error;
Implement dlinfo() function. Introdice RTLD_SELF special handle and properly process it within dlsym() and dlinfo() functions. The intention is to improve our compatibility with Solaris and to make a Java port easier. Partially submitted by: phantom
2003-02-13 17:47:44 +00:00
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
rlock_acquire(rtld_bind_lock, &lockstate);
Implement dlinfo() function. Introdice RTLD_SELF special handle and properly process it within dlsym() and dlinfo() functions. The intention is to improve our compatibility with Solaris and to make a Java port easier. Partially submitted by: phantom
2003-02-13 17:47:44 +00:00
if (handle == NULL || handle == RTLD_SELF) {
void *retaddr;
retaddr = __builtin_return_address(0); /* __GNUC__ only */
if ((obj = obj_from_addr(retaddr)) == NULL)
_rtld_error("Cannot determine caller's shared object");
} else
obj = dlcheck(handle);
if (obj == NULL) {
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
lock_release(rtld_bind_lock, &lockstate);
Implement dlinfo() function. Introdice RTLD_SELF special handle and properly process it within dlsym() and dlinfo() functions. The intention is to improve our compatibility with Solaris and to make a Java port easier. Partially submitted by: phantom
2003-02-13 17:47:44 +00:00
return (-1);
}
error = 0;
switch (request) {
case RTLD_DI_LINKMAP:
*((struct link_map const **)p) = &obj->linkmap;
break;
case RTLD_DI_ORIGIN:
error = rtld_dirname(obj->path, p);
break;
case RTLD_DI_SERINFOSIZE:
case RTLD_DI_SERINFO:
error = do_search_info(obj, request, (struct dl_serinfo *)p);
break;
default:
_rtld_error("Invalid request %d passed to dlinfo()", request);
error = -1;
}
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
lock_release(rtld_bind_lock, &lockstate);
Implement dlinfo() function. Introdice RTLD_SELF special handle and properly process it within dlsym() and dlinfo() functions. The intention is to improve our compatibility with Solaris and to make a Java port easier. Partially submitted by: phantom
2003-02-13 17:47:44 +00:00
return (error);
}
Introduce implementation-private rtld interface _rtld_addr_phdr, which fills struct dl_phdr_info for the shared object that contains the specified address, if any. Idea and reviewed by: kan MFC after: 3 weeks
2010-08-23 15:27:03 +00:00
static void
rtld_fill_dl_phdr_info(const Obj_Entry *obj, struct dl_phdr_info *phdr_info)
{
phdr_info->dlpi_addr = (Elf_Addr)obj->relocbase;
Always return pathname in dl_iterate_phdr's dlpi_name, as Linux does Linux LD_ITERATE_PHDR(3): The dlpi_name field is a null-terminated string giving the pathname from which the shared object was loaded. That functionality is much more useful than returning just the short name. Approved by: kan Sponsored by: The FreeBSD Foundation
2014-10-09 20:39:18 +00:00
phdr_info->dlpi_name = obj->path;
Introduce implementation-private rtld interface _rtld_addr_phdr, which fills struct dl_phdr_info for the shared object that contains the specified address, if any. Idea and reviewed by: kan MFC after: 3 weeks
2010-08-23 15:27:03 +00:00
phdr_info->dlpi_phdr = obj->phdr;
phdr_info->dlpi_phnum = obj->phsize / sizeof(obj->phdr[0]);
phdr_info->dlpi_tls_modid = obj->tlsindex;
phdr_info->dlpi_tls_data = obj->tlsinit;
phdr_info->dlpi_adds = obj_loads;
phdr_info->dlpi_subs = obj_loads - obj_count;
}
Implement dl_iterate_phdr function. Convert boolean flags in internal Obj_Entry structure into bitfields. Properly check for loaded segment alignment in map_object.
2007-04-03 18:31:20 +00:00
int
dl_iterate_phdr(__dl_iterate_hdr_callback callback, void *param)
{
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
struct dl_phdr_info phdr_info;
Obj_Entry *obj, marker;
RtldLockState bind_lockstate, phdr_lockstate;
int error;
Implement dl_iterate_phdr function. Convert boolean flags in internal Obj_Entry structure into bitfields. Properly check for loaded segment alignment in map_object.
2007-04-03 18:31:20 +00:00
rtld: Fix a couple of bugs around the unloading of ELF filters. - Pass the correct object to unload_filtees(). - Use a marker to restart iteration after unload_filtees() has returned. It calls dlclose() and may recursively remove entries from the global object list, so TAILQ_FOREACH_SAFE is not sufficient. Reviewed by: kib MFC after: 2 weeks Sponsored by: Dell EMC Isilon
2016-12-22 17:44:27 +00:00
init_marker(&marker);
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
error = 0;
Include rtld itself when iterating over loaded ELF objects in dl_iterate_phdr(3). Reviewed by: kib
2013-11-07 20:45:50 +00:00
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
wlock_acquire(rtld_phdr_lock, &phdr_lockstate);
Take write lock for rtld_bind before modifying obj_list in dl_iterate_phdr(). This avoids a race with readers such as dladdr(3)/dlinfo(3)/dlsym(3) and the atexit(3) handler. This race was introduced in r294373. Reviewed by: markj, kib, kan MFC after: 2 weeks Sponsored by: Dell EMC Isilon
2016-12-13 18:05:14 +00:00
wlock_acquire(rtld_bind_lock, &bind_lockstate);
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
for (obj = globallist_curr(TAILQ_FIRST(&obj_list)); obj != NULL;) {
TAILQ_INSERT_AFTER(&obj_list, obj, &marker, next);
rtld_fill_dl_phdr_info(obj, &phdr_info);
rtld: Fix a race between dl_iterate_phdr() and dlclose(). Add a transient reference count to ensure that the phdr argument to the callback remains valid while the bind lock is dropped. Reviewed by: kib MFC after: 2 weeks Sponsored by: Dell EMC Isilon
2016-12-22 17:37:39 +00:00
hold_object(obj);
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
lock_release(rtld_bind_lock, &bind_lockstate);
error = callback(&phdr_info, sizeof phdr_info, param);
Take write lock for rtld_bind before modifying obj_list in dl_iterate_phdr(). This avoids a race with readers such as dladdr(3)/dlinfo(3)/dlsym(3) and the atexit(3) handler. This race was introduced in r294373. Reviewed by: markj, kib, kan MFC after: 2 weeks Sponsored by: Dell EMC Isilon
2016-12-13 18:05:14 +00:00
wlock_acquire(rtld_bind_lock, &bind_lockstate);
rtld: Fix a race between dl_iterate_phdr() and dlclose(). Add a transient reference count to ensure that the phdr argument to the callback remains valid while the bind lock is dropped. Reviewed by: kib MFC after: 2 weeks Sponsored by: Dell EMC Isilon
2016-12-22 17:37:39 +00:00
unhold_object(obj);
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
obj = globallist_next(&marker);
TAILQ_REMOVE(&obj_list, &marker, next);
if (error != 0) {
lock_release(rtld_bind_lock, &bind_lockstate);
lock_release(rtld_phdr_lock, &phdr_lockstate);
return (error);
}
}
Implement dl_iterate_phdr function. Convert boolean flags in internal Obj_Entry structure into bitfields. Properly check for loaded segment alignment in map_object.
2007-04-03 18:31:20 +00:00
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
if (error == 0) {
rtld_fill_dl_phdr_info(&obj_rtld, &phdr_info);
lock_release(rtld_bind_lock, &bind_lockstate);
error = callback(&phdr_info, sizeof(phdr_info), param);
}
Do not unlock rtld_phdr_lock over callback invocations. The dl_iterate_phdr consumer code in libgcc does not expect multiple callbacks running concurrently. This was fixed once already in r178807, but accidentally got reverted in r294373.
2016-01-27 20:20:37 +00:00
lock_release(rtld_phdr_lock, &phdr_lockstate);
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
return (error);
Implement dl_iterate_phdr function. Convert boolean flags in internal Obj_Entry structure into bitfields. Properly check for loaded segment alignment in map_object.
2007-04-03 18:31:20 +00:00
}
Implement dlinfo() function. Introdice RTLD_SELF special handle and properly process it within dlsym() and dlinfo() functions. The intention is to improve our compatibility with Solaris and to make a Java port easier. Partially submitted by: phantom
2003-02-13 17:47:44 +00:00
static void *
fill_search_info(const char *dir, size_t dirlen, void *param)
{
struct fill_search_info_args *arg;
arg = param;
if (arg->request == RTLD_DI_SERINFOSIZE) {
arg->serinfo->dls_cnt ++;
Import the DragonFly BSD commit 4f0bc915b65fcf5a23214f6d221d65c80be68ad4 by John Marino <draco@marino.st>, with the following (edited) commit message Date: Sat, 24 Mar 2012 06:40:50 +0100 Subject: [PATCH 1/1] rtld: Implement DT_RUNPATH and -z nodefaultlib DT_RUNPATH is incorrectly being considered as an alias of DT_RPATH. The purpose of DT_RUNPATH is to have two different types of rpath: one that can be overridden by the environment variable LD_LIBRARY_PATH and one that can't. With the currently implementation, LD_LIBRARY_PATH will always trump any embedded rpath or runpath tags. Current path search order by rtld: ================================== LD_LIBRARY_PATH DT_RPATH / DT_RUNPATH (always the same) ldconfig hints file (default: /var/run/ld-elf.so.hints) /usr/lib New path search order by rtld: ============================== DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file /usr/lib The new path search matches how the linux runtime loader works. The other major added feature is support for linker flag "-z nodefaultlib". When this flag is passed to the linker, rtld will skip all references to the standard library search path ("/usr/lib" in this case but it could handle more color delimited paths) except in DT_RPATH and DT_RUNPATH. New path search order by rtld with -z nodefaultlib flag set: ============================================================ DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file (skips all references to /usr/lib) FreeBSD notes: - we fixed some bugs which were submitted to DragonFly and merged there as commit 1ff8a2bd3eb6e5587174c6a983303ea3a79e0002; - we added LD_LIBRARY_PATH_RPATH environment variable to switch to the previous behaviour of considering DT_RPATH a synonym for DT_RUNPATH; - the FreeBSD default search path is /lib:/usr/lib and not /usr/lib. Reviewed by: kan MFC after: 1 month MFC note: flip the ld_library_path_rpath default value for stable/9
2012-07-15 10:53:48 +00:00
arg->serinfo->dls_size += sizeof(struct dl_serpath) + dirlen + 1;
Implement dlinfo() function. Introdice RTLD_SELF special handle and properly process it within dlsym() and dlinfo() functions. The intention is to improve our compatibility with Solaris and to make a Java port easier. Partially submitted by: phantom
2003-02-13 17:47:44 +00:00
} else {
struct dl_serpath *s_entry;
s_entry = arg->serpath;
s_entry->dls_name = arg->strspace;
s_entry->dls_flags = arg->flags;
strncpy(arg->strspace, dir, dirlen);
arg->strspace[dirlen] = '\0';
arg->strspace += dirlen + 1;
arg->serpath++;
}
return (NULL);
}
static int
do_search_info(const Obj_Entry *obj, int request, struct dl_serinfo *info)
{
struct dl_serinfo _info;
struct fill_search_info_args args;
args.request = RTLD_DI_SERINFOSIZE;
args.serinfo = &_info;
_info.dls_size = __offsetof(struct dl_serinfo, dls_serpath);
_info.dls_cnt = 0;
path_enumerate(obj->rpath, fill_search_info, &args);
Import the DragonFly BSD commit 4f0bc915b65fcf5a23214f6d221d65c80be68ad4 by John Marino <draco@marino.st>, with the following (edited) commit message Date: Sat, 24 Mar 2012 06:40:50 +0100 Subject: [PATCH 1/1] rtld: Implement DT_RUNPATH and -z nodefaultlib DT_RUNPATH is incorrectly being considered as an alias of DT_RPATH. The purpose of DT_RUNPATH is to have two different types of rpath: one that can be overridden by the environment variable LD_LIBRARY_PATH and one that can't. With the currently implementation, LD_LIBRARY_PATH will always trump any embedded rpath or runpath tags. Current path search order by rtld: ================================== LD_LIBRARY_PATH DT_RPATH / DT_RUNPATH (always the same) ldconfig hints file (default: /var/run/ld-elf.so.hints) /usr/lib New path search order by rtld: ============================== DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file /usr/lib The new path search matches how the linux runtime loader works. The other major added feature is support for linker flag "-z nodefaultlib". When this flag is passed to the linker, rtld will skip all references to the standard library search path ("/usr/lib" in this case but it could handle more color delimited paths) except in DT_RPATH and DT_RUNPATH. New path search order by rtld with -z nodefaultlib flag set: ============================================================ DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file (skips all references to /usr/lib) FreeBSD notes: - we fixed some bugs which were submitted to DragonFly and merged there as commit 1ff8a2bd3eb6e5587174c6a983303ea3a79e0002; - we added LD_LIBRARY_PATH_RPATH environment variable to switch to the previous behaviour of considering DT_RPATH a synonym for DT_RUNPATH; - the FreeBSD default search path is /lib:/usr/lib and not /usr/lib. Reviewed by: kan MFC after: 1 month MFC note: flip the ld_library_path_rpath default value for stable/9
2012-07-15 10:53:48 +00:00
path_enumerate(ld_library_path, fill_search_info, &args);
path_enumerate(obj->runpath, fill_search_info, &args);
path_enumerate(gethints(obj->z_nodeflib), fill_search_info, &args);
if (!obj->z_nodeflib)
Rather than using the #define for path names, indirect through a char * variable that could change for different executable types detected.
2015-10-31 04:39:55 +00:00
path_enumerate(ld_standard_library_path, fill_search_info, &args);
Implement dlinfo() function. Introdice RTLD_SELF special handle and properly process it within dlsym() and dlinfo() functions. The intention is to improve our compatibility with Solaris and to make a Java port easier. Partially submitted by: phantom
2003-02-13 17:47:44 +00:00
if (request == RTLD_DI_SERINFOSIZE) {
info->dls_size = _info.dls_size;
info->dls_cnt = _info.dls_cnt;
return (0);
}
if (info->dls_cnt != _info.dls_cnt || info->dls_size != _info.dls_size) {
_rtld_error("Uninitialized Dl_serinfo struct passed to dlinfo()");
return (-1);
}
args.request = RTLD_DI_SERINFO;
args.serinfo = info;
args.serpath = &info->dls_serpath[0];
args.strspace = (char *)&info->dls_serpath[_info.dls_cnt];
Import the DragonFly BSD commit 4f0bc915b65fcf5a23214f6d221d65c80be68ad4 by John Marino <draco@marino.st>, with the following (edited) commit message Date: Sat, 24 Mar 2012 06:40:50 +0100 Subject: [PATCH 1/1] rtld: Implement DT_RUNPATH and -z nodefaultlib DT_RUNPATH is incorrectly being considered as an alias of DT_RPATH. The purpose of DT_RUNPATH is to have two different types of rpath: one that can be overridden by the environment variable LD_LIBRARY_PATH and one that can't. With the currently implementation, LD_LIBRARY_PATH will always trump any embedded rpath or runpath tags. Current path search order by rtld: ================================== LD_LIBRARY_PATH DT_RPATH / DT_RUNPATH (always the same) ldconfig hints file (default: /var/run/ld-elf.so.hints) /usr/lib New path search order by rtld: ============================== DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file /usr/lib The new path search matches how the linux runtime loader works. The other major added feature is support for linker flag "-z nodefaultlib". When this flag is passed to the linker, rtld will skip all references to the standard library search path ("/usr/lib" in this case but it could handle more color delimited paths) except in DT_RPATH and DT_RUNPATH. New path search order by rtld with -z nodefaultlib flag set: ============================================================ DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file (skips all references to /usr/lib) FreeBSD notes: - we fixed some bugs which were submitted to DragonFly and merged there as commit 1ff8a2bd3eb6e5587174c6a983303ea3a79e0002; - we added LD_LIBRARY_PATH_RPATH environment variable to switch to the previous behaviour of considering DT_RPATH a synonym for DT_RUNPATH; - the FreeBSD default search path is /lib:/usr/lib and not /usr/lib. Reviewed by: kan MFC after: 1 month MFC note: flip the ld_library_path_rpath default value for stable/9
2012-07-15 10:53:48 +00:00
args.flags = LA_SER_RUNPATH;
if (path_enumerate(obj->rpath, fill_search_info, &args) != NULL)
return (-1);
Implement dlinfo() function. Introdice RTLD_SELF special handle and properly process it within dlsym() and dlinfo() functions. The intention is to improve our compatibility with Solaris and to make a Java port easier. Partially submitted by: phantom
2003-02-13 17:47:44 +00:00
args.flags = LA_SER_LIBPATH;
if (path_enumerate(ld_library_path, fill_search_info, &args) != NULL)
return (-1);
args.flags = LA_SER_RUNPATH;
Import the DragonFly BSD commit 4f0bc915b65fcf5a23214f6d221d65c80be68ad4 by John Marino <draco@marino.st>, with the following (edited) commit message Date: Sat, 24 Mar 2012 06:40:50 +0100 Subject: [PATCH 1/1] rtld: Implement DT_RUNPATH and -z nodefaultlib DT_RUNPATH is incorrectly being considered as an alias of DT_RPATH. The purpose of DT_RUNPATH is to have two different types of rpath: one that can be overridden by the environment variable LD_LIBRARY_PATH and one that can't. With the currently implementation, LD_LIBRARY_PATH will always trump any embedded rpath or runpath tags. Current path search order by rtld: ================================== LD_LIBRARY_PATH DT_RPATH / DT_RUNPATH (always the same) ldconfig hints file (default: /var/run/ld-elf.so.hints) /usr/lib New path search order by rtld: ============================== DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file /usr/lib The new path search matches how the linux runtime loader works. The other major added feature is support for linker flag "-z nodefaultlib". When this flag is passed to the linker, rtld will skip all references to the standard library search path ("/usr/lib" in this case but it could handle more color delimited paths) except in DT_RPATH and DT_RUNPATH. New path search order by rtld with -z nodefaultlib flag set: ============================================================ DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file (skips all references to /usr/lib) FreeBSD notes: - we fixed some bugs which were submitted to DragonFly and merged there as commit 1ff8a2bd3eb6e5587174c6a983303ea3a79e0002; - we added LD_LIBRARY_PATH_RPATH environment variable to switch to the previous behaviour of considering DT_RPATH a synonym for DT_RUNPATH; - the FreeBSD default search path is /lib:/usr/lib and not /usr/lib. Reviewed by: kan MFC after: 1 month MFC note: flip the ld_library_path_rpath default value for stable/9
2012-07-15 10:53:48 +00:00
if (path_enumerate(obj->runpath, fill_search_info, &args) != NULL)
Implement dlinfo() function. Introdice RTLD_SELF special handle and properly process it within dlsym() and dlinfo() functions. The intention is to improve our compatibility with Solaris and to make a Java port easier. Partially submitted by: phantom
2003-02-13 17:47:44 +00:00
return (-1);
args.flags = LA_SER_CONFIG;
Import the DragonFly BSD commit 4f0bc915b65fcf5a23214f6d221d65c80be68ad4 by John Marino <draco@marino.st>, with the following (edited) commit message Date: Sat, 24 Mar 2012 06:40:50 +0100 Subject: [PATCH 1/1] rtld: Implement DT_RUNPATH and -z nodefaultlib DT_RUNPATH is incorrectly being considered as an alias of DT_RPATH. The purpose of DT_RUNPATH is to have two different types of rpath: one that can be overridden by the environment variable LD_LIBRARY_PATH and one that can't. With the currently implementation, LD_LIBRARY_PATH will always trump any embedded rpath or runpath tags. Current path search order by rtld: ================================== LD_LIBRARY_PATH DT_RPATH / DT_RUNPATH (always the same) ldconfig hints file (default: /var/run/ld-elf.so.hints) /usr/lib New path search order by rtld: ============================== DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file /usr/lib The new path search matches how the linux runtime loader works. The other major added feature is support for linker flag "-z nodefaultlib". When this flag is passed to the linker, rtld will skip all references to the standard library search path ("/usr/lib" in this case but it could handle more color delimited paths) except in DT_RPATH and DT_RUNPATH. New path search order by rtld with -z nodefaultlib flag set: ============================================================ DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file (skips all references to /usr/lib) FreeBSD notes: - we fixed some bugs which were submitted to DragonFly and merged there as commit 1ff8a2bd3eb6e5587174c6a983303ea3a79e0002; - we added LD_LIBRARY_PATH_RPATH environment variable to switch to the previous behaviour of considering DT_RPATH a synonym for DT_RUNPATH; - the FreeBSD default search path is /lib:/usr/lib and not /usr/lib. Reviewed by: kan MFC after: 1 month MFC note: flip the ld_library_path_rpath default value for stable/9
2012-07-15 10:53:48 +00:00
if (path_enumerate(gethints(obj->z_nodeflib), fill_search_info, &args)
!= NULL)
Implement dlinfo() function. Introdice RTLD_SELF special handle and properly process it within dlsym() and dlinfo() functions. The intention is to improve our compatibility with Solaris and to make a Java port easier. Partially submitted by: phantom
2003-02-13 17:47:44 +00:00
return (-1);
args.flags = LA_SER_DEFAULT;
Import the DragonFly BSD commit 4f0bc915b65fcf5a23214f6d221d65c80be68ad4 by John Marino <draco@marino.st>, with the following (edited) commit message Date: Sat, 24 Mar 2012 06:40:50 +0100 Subject: [PATCH 1/1] rtld: Implement DT_RUNPATH and -z nodefaultlib DT_RUNPATH is incorrectly being considered as an alias of DT_RPATH. The purpose of DT_RUNPATH is to have two different types of rpath: one that can be overridden by the environment variable LD_LIBRARY_PATH and one that can't. With the currently implementation, LD_LIBRARY_PATH will always trump any embedded rpath or runpath tags. Current path search order by rtld: ================================== LD_LIBRARY_PATH DT_RPATH / DT_RUNPATH (always the same) ldconfig hints file (default: /var/run/ld-elf.so.hints) /usr/lib New path search order by rtld: ============================== DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file /usr/lib The new path search matches how the linux runtime loader works. The other major added feature is support for linker flag "-z nodefaultlib". When this flag is passed to the linker, rtld will skip all references to the standard library search path ("/usr/lib" in this case but it could handle more color delimited paths) except in DT_RPATH and DT_RUNPATH. New path search order by rtld with -z nodefaultlib flag set: ============================================================ DT_RPATH of the calling object if no DT_RUNPATH DT_RPATH of the main binary if no DT_RUNPATH and binary isn't calling obj LD_LIBRARY_PATH DT_RUNPATH ldconfig hints file (skips all references to /usr/lib) FreeBSD notes: - we fixed some bugs which were submitted to DragonFly and merged there as commit 1ff8a2bd3eb6e5587174c6a983303ea3a79e0002; - we added LD_LIBRARY_PATH_RPATH environment variable to switch to the previous behaviour of considering DT_RPATH a synonym for DT_RUNPATH; - the FreeBSD default search path is /lib:/usr/lib and not /usr/lib. Reviewed by: kan MFC after: 1 month MFC note: flip the ld_library_path_rpath default value for stable/9
2012-07-15 10:53:48 +00:00
if (!obj->z_nodeflib &&
Rather than using the #define for path names, indirect through a char * variable that could change for different executable types detected.
2015-10-31 04:39:55 +00:00
path_enumerate(ld_standard_library_path, fill_search_info, &args) != NULL)
Implement dlinfo() function. Introdice RTLD_SELF special handle and properly process it within dlsym() and dlinfo() functions. The intention is to improve our compatibility with Solaris and to make a Java port easier. Partially submitted by: phantom
2003-02-13 17:47:44 +00:00
return (-1);
return (0);
}
static int
rtld_dirname(const char *path, char *bname)
{
const char *endp;
/* Empty or NULL string gets treated as "." */
if (path == NULL || *path == '\0') {
bname[0] = '.';
Fix a typo in rtld_dirname.
2003-02-13 22:47:41 +00:00
bname[1] = '\0';
Implement dlinfo() function. Introdice RTLD_SELF special handle and properly process it within dlsym() and dlinfo() functions. The intention is to improve our compatibility with Solaris and to make a Java port easier. Partially submitted by: phantom
2003-02-13 17:47:44 +00:00
return (0);
}
/* Strip trailing slashes */
endp = path + strlen(path) - 1;
while (endp > path && *endp == '/')
endp--;
/* Find the start of the dir */
while (endp > path && *endp != '/')
endp--;
/* Either the dir is "/" or there are no slashes */
if (endp == path) {
bname[0] = *endp == '/' ? '/' : '.';
bname[1] = '\0';
return (0);
} else {
do {
endp--;
} while (endp > path && *endp == '/');
}
if (endp - path + 2 > PATH_MAX)
{
_rtld_error("Filename is too long: %s", path);
return(-1);
}
strncpy(bname, path, endp - path + 1);
bname[endp - path + 1] = '\0';
return (0);
}
Implement the dynamic string token substitution in the rpath and soneeded pathes. The $ORIGIN, $OSNAME, $OSREL and $PLATFORM tokens are supported. Enabling the substitution requires DF_ORIGIN flag in DT_FLAGS or DF_1_ORIGIN if DF_FLAGS_1, that may be set with -z origin gnu ld flag. Translation is unconditionally disabled for setuid/setgid processes. The $ORIGIN translation relies on the AT_EXECPATH auxinfo supplied by kernel. Requested by: maho Tested by: maho, pho Reviewed by: kan
2009-03-18 13:40:37 +00:00
static int
rtld_dirname_abs(const char *path, char *base)
{
When a file is executed and the path starts with `/', AT_EXECPATH is set without any translation. If the file is a symbolic link, $ORIGIN may not be expanded to the actual origin. Use realpath(3) to properly expand $ORIGIN to its absolute path. Reviewed by: kib MFC after: 1 week
2015-02-27 19:05:23 +00:00
char *last;
Implement the dynamic string token substitution in the rpath and soneeded pathes. The $ORIGIN, $OSNAME, $OSREL and $PLATFORM tokens are supported. Enabling the substitution requires DF_ORIGIN flag in DT_FLAGS or DF_1_ORIGIN if DF_FLAGS_1, that may be set with -z origin gnu ld flag. Translation is unconditionally disabled for setuid/setgid processes. The $ORIGIN translation relies on the AT_EXECPATH auxinfo supplied by kernel. Requested by: maho Tested by: maho, pho Reviewed by: kan
2009-03-18 13:40:37 +00:00
When a file is executed and the path starts with `/', AT_EXECPATH is set without any translation. If the file is a symbolic link, $ORIGIN may not be expanded to the actual origin. Use realpath(3) to properly expand $ORIGIN to its absolute path. Reviewed by: kib MFC after: 1 week
2015-02-27 19:05:23 +00:00
if (realpath(path, base) == NULL)
Implement the dynamic string token substitution in the rpath and soneeded pathes. The $ORIGIN, $OSNAME, $OSREL and $PLATFORM tokens are supported. Enabling the substitution requires DF_ORIGIN flag in DT_FLAGS or DF_1_ORIGIN if DF_FLAGS_1, that may be set with -z origin gnu ld flag. Translation is unconditionally disabled for setuid/setgid processes. The $ORIGIN translation relies on the AT_EXECPATH auxinfo supplied by kernel. Requested by: maho Tested by: maho, pho Reviewed by: kan
2009-03-18 13:40:37 +00:00
return (-1);
When a file is executed and the path starts with `/', AT_EXECPATH is set without any translation. If the file is a symbolic link, $ORIGIN may not be expanded to the actual origin. Use realpath(3) to properly expand $ORIGIN to its absolute path. Reviewed by: kib MFC after: 1 week
2015-02-27 19:05:23 +00:00
dbg("%s -> %s", path, base);
last = strrchr(base, '/');
if (last == NULL)
Implement the dynamic string token substitution in the rpath and soneeded pathes. The $ORIGIN, $OSNAME, $OSREL and $PLATFORM tokens are supported. Enabling the substitution requires DF_ORIGIN flag in DT_FLAGS or DF_1_ORIGIN if DF_FLAGS_1, that may be set with -z origin gnu ld flag. Translation is unconditionally disabled for setuid/setgid processes. The $ORIGIN translation relies on the AT_EXECPATH auxinfo supplied by kernel. Requested by: maho Tested by: maho, pho Reviewed by: kan
2009-03-18 13:40:37 +00:00
return (-1);
When a file is executed and the path starts with `/', AT_EXECPATH is set without any translation. If the file is a symbolic link, $ORIGIN may not be expanded to the actual origin. Use realpath(3) to properly expand $ORIGIN to its absolute path. Reviewed by: kib MFC after: 1 week
2015-02-27 19:05:23 +00:00
if (last != base)
*last = '\0';
Implement the dynamic string token substitution in the rpath and soneeded pathes. The $ORIGIN, $OSNAME, $OSREL and $PLATFORM tokens are supported. Enabling the substitution requires DF_ORIGIN flag in DT_FLAGS or DF_1_ORIGIN if DF_FLAGS_1, that may be set with -z origin gnu ld flag. Translation is unconditionally disabled for setuid/setgid processes. The $ORIGIN translation relies on the AT_EXECPATH auxinfo supplied by kernel. Requested by: maho Tested by: maho, pho Reviewed by: kan
2009-03-18 13:40:37 +00:00
return (0);
}
Add GDB support. The method and some of the code came from NetBSD's elf runtime linker.
1998-04-30 07:48:02 +00:00
static void
linkmap_add(Obj_Entry *obj)
{
struct link_map *l = &obj->linkmap;
struct link_map *prev;
obj->linkmap.l_name = obj->path;
obj->linkmap.l_addr = obj->mapbase;
obj->linkmap.l_ld = obj->dynamic;
#ifdef __mips__
/* GDB needs load offset on MIPS to use the symbols */
obj->linkmap.l_offs = obj->relocbase;
#endif
if (r_debug.r_map == NULL) {
r_debug.r_map = l;
return;
}
Implement dlinfo() function. Introdice RTLD_SELF special handle and properly process it within dlsym() and dlinfo() functions. The intention is to improve our compatibility with Solaris and to make a Java port easier. Partially submitted by: phantom
2003-02-13 17:47:44 +00:00
Fix a bug that showed up when debugging dynamically linked programs. References from GDB to "printf" and various other functions would find the versions in the dynamic linker itself, rather than the versions in the program's libc. This fix moves the GDB link map entry for the dynamic linker to the end of the search list, where its symbols will be found only if they are not found anywhere else. It was suggested by Doug Rabson, though I implemented it a little differently. I personally would prefer to leave the dynamic linker's entry out of the GDB search list altogether. But Doug argues that it is handy there for such things as setting breakpoints on dlopen(). So it stays for now, at least. Note, if we ever integrate the dynamic linker with libc (which has several important benefits to recommend it), this whole problem goes away.
1998-09-16 02:54:08 +00:00
/*
* Scan to the end of the list, but not past the entry for the
* dynamic linker, which we want to keep at the very end.
*/
for (prev = r_debug.r_map;
prev->l_next != NULL && prev->l_next != &obj_rtld.linkmap;
prev = prev->l_next)
Add GDB support. The method and some of the code came from NetBSD's elf runtime linker.
1998-04-30 07:48:02 +00:00
;
Fix a bug that showed up when debugging dynamically linked programs. References from GDB to "printf" and various other functions would find the versions in the dynamic linker itself, rather than the versions in the program's libc. This fix moves the GDB link map entry for the dynamic linker to the end of the search list, where its symbols will be found only if they are not found anywhere else. It was suggested by Doug Rabson, though I implemented it a little differently. I personally would prefer to leave the dynamic linker's entry out of the GDB search list altogether. But Doug argues that it is handy there for such things as setting breakpoints on dlopen(). So it stays for now, at least. Note, if we ever integrate the dynamic linker with libc (which has several important benefits to recommend it), this whole problem goes away.
1998-09-16 02:54:08 +00:00
/* Link in the new entry. */
Add GDB support. The method and some of the code came from NetBSD's elf runtime linker.
1998-04-30 07:48:02 +00:00
l->l_prev = prev;
Fix a bug that showed up when debugging dynamically linked programs. References from GDB to "printf" and various other functions would find the versions in the dynamic linker itself, rather than the versions in the program's libc. This fix moves the GDB link map entry for the dynamic linker to the end of the search list, where its symbols will be found only if they are not found anywhere else. It was suggested by Doug Rabson, though I implemented it a little differently. I personally would prefer to leave the dynamic linker's entry out of the GDB search list altogether. But Doug argues that it is handy there for such things as setting breakpoints on dlopen(). So it stays for now, at least. Note, if we ever integrate the dynamic linker with libc (which has several important benefits to recommend it), this whole problem goes away.
1998-09-16 02:54:08 +00:00
l->l_next = prev->l_next;
if (l->l_next != NULL)
l->l_next->l_prev = l;
Add GDB support. The method and some of the code came from NetBSD's elf runtime linker.
1998-04-30 07:48:02 +00:00
prev->l_next = l;
}
Style fixes. If it seems like a lot of lines of changes, it's because I moved some functions. Mr. Tidy likes them to be in alphabetical order.
1998-09-02 02:00:20 +00:00
static void
linkmap_delete(Obj_Entry *obj)
Add GDB support. The method and some of the code came from NetBSD's elf runtime linker.
1998-04-30 07:48:02 +00:00
{
struct link_map *l = &obj->linkmap;
if (l->l_prev == NULL) {
if ((r_debug.r_map = l->l_next) != NULL)
l->l_next->l_prev = NULL;
return;
}
if ((l->l_prev->l_next = l->l_next) != NULL)
l->l_next->l_prev = l->l_prev;
}
Add support for ldd.
1998-05-01 08:39:27 +00:00
Style fixes. If it seems like a lot of lines of changes, it's because I moved some functions. Mr. Tidy likes them to be in alphabetical order.
1998-09-02 02:00:20 +00:00
/*
* Function for the debugger to set a breakpoint on to gain control.
Pass two pointer parameters to the r_debug_state() hook function, thus allowing a debugger or other trace tool to easily grab the addresses of the needed structures off the stack. This change is transparent to gdb, which locates the link_map list and transfers it to debugger memory for comparison purposes. A sample program will be committed showing how this can be used. Reviewed by: John Polstra <jdp@FreeBSD.org>
2000-08-26 05:13:29 +00:00
*
* The two parameters allow the debugger to easily find and determine
* what the runtime loader is doing and to whom it is doing it.
*
* When the loadhook trap is hit (r_debug_state, set at program
* initialization), the arguments can be found on the stack:
*
* +8 struct link_map *m
* +4 struct r_debug *rd
* +0 RetAddr
Style fixes. If it seems like a lot of lines of changes, it's because I moved some functions. Mr. Tidy likes them to be in alphabetical order.
1998-09-02 02:00:20 +00:00
*/
void
Pass two pointer parameters to the r_debug_state() hook function, thus allowing a debugger or other trace tool to easily grab the addresses of the needed structures off the stack. This change is transparent to gdb, which locates the link_map list and transfers it to debugger memory for comparison purposes. A sample program will be committed showing how this can be used. Reviewed by: John Polstra <jdp@FreeBSD.org>
2000-08-26 05:13:29 +00:00
r_debug_state(struct r_debug* rd, struct link_map *m)
Style fixes. If it seems like a lot of lines of changes, it's because I moved some functions. Mr. Tidy likes them to be in alphabetical order.
1998-09-02 02:00:20 +00:00
{
When libexec/rtld-elf/rtld.c is compiled with clang, the r_debug_state() function (a hook necessary for gdb support), is inlined, but since the function contains no code, no calls to it are generated. When gdb is debugging a dynamically linked program, this causes backtraces to be corrupted. Fix it by marking the function __noinline, and inserting an empty asm statement, that pretends to clobber memory. This forces the compiler to emit calls to r_debug_state() throughout rtld.c. Approved by: re (kib)
2011-09-03 11:41:00 +00:00
/*
* The following is a hack to force the compiler to emit calls to
* this function, even when optimizing. If the function is empty,
* the compiler is not obliged to emit any code for calls to it,
* even when marked __noinline. However, gdb depends on those
* calls being made.
*/
- Export the function added in r265456 rather than the non-existent _rtld_debug_postinit(). [1] - Use __compiler_membar() instead of inline asm in _r_debug_state() and _r_debug_postinit(). [2] Pointy hat to: markj [1] Reported by: attilio [2] Discussed with: kib X-MFC-With: r265456
2014-05-07 17:21:22 +00:00
__compiler_membar();
Style fixes. If it seems like a lot of lines of changes, it's because I moved some functions. Mr. Tidy likes them to be in alphabetical order.
1998-09-02 02:00:20 +00:00
}
Add a postinit debugger hook to rtld. This will be used by dtrace(1) to halt the victim process before its entry point is called, at which point probes and DOF data are registered with the kernel. The r_debug_state hook cannot be used for this purpose, as it is called before the program's init routines are invoked and in particular before DOF data is registered (via drti.o). Reviewed by: kib MFC after: 2 weeks
2014-05-06 18:07:58 +00:00
/*
* A function called after init routines have completed. This can be used to
* break before a program's entry routine is called, and can be used when
* main is not available in the symbol table.
*/
void
_r_debug_postinit(struct link_map *m)
{
/* See r_debug_state(). */
- Export the function added in r265456 rather than the non-existent _rtld_debug_postinit(). [1] - Use __compiler_membar() instead of inline asm in _r_debug_state() and _r_debug_postinit(). [2] Pointy hat to: markj [1] Reported by: attilio [2] Discussed with: kib X-MFC-With: r265456
2014-05-07 17:21:22 +00:00
__compiler_membar();
Add a postinit debugger hook to rtld. This will be used by dtrace(1) to halt the victim process before its entry point is called, at which point probes and DOF data are registered with the kernel. The r_debug_state hook cannot be used for this purpose, as it is called before the program's init routines are invoked and in particular before DOF data is registered (via drti.o). Reviewed by: kib MFC after: 2 weeks
2014-05-06 18:07:58 +00:00
}
rtld: Fix a race between dl_iterate_phdr() and dlclose(). Add a transient reference count to ensure that the phdr argument to the callback remains valid while the bind lock is dropped. Reviewed by: kib MFC after: 2 weeks Sponsored by: Dell EMC Isilon
2016-12-22 17:37:39 +00:00
static void
release_object(Obj_Entry *obj)
{
if (obj->holdcount > 0) {
obj->unholdfree = true;
return;
}
munmap(obj->mapbase, obj->mapsize);
linkmap_delete(obj);
obj_free(obj);
}
After relocating the main program, but before calling any of the _init() functions, initialize the global variables "__progname" and "environ". This makes it possible for the _init() functions to call things like getenv() and err().
1999-04-21 04:06:57 +00:00
/*
Allow ldd(1) be used on shared libraries in addition to executables.
2002-02-04 10:33:48 +00:00
* Get address of the pointer variable in the main program.
get_program_var_addr() must prefer the strong symbol to the weak one. Search global objects, together with main object and dependencies, for the requested symbol. Move the common code from symlook_default() into new helper symlook_global(), and use it both in symlook_global() and get_program_var_addr(). Supply lock state to get_program_var_addr(). Reviewed by: kan Tested by: Mykola Dzham <i levsha me>
2011-01-10 16:09:35 +00:00
* Prefer non-weak symbol over the weak one.
After relocating the main program, but before calling any of the _init() functions, initialize the global variables "__progname" and "environ". This makes it possible for the _init() functions to call things like getenv() and err().
1999-04-21 04:06:57 +00:00
*/
Allow ldd(1) be used on shared libraries in addition to executables.
2002-02-04 10:33:48 +00:00
static const void **
get_program_var_addr() must prefer the strong symbol to the weak one. Search global objects, together with main object and dependencies, for the requested symbol. Move the common code from symlook_default() into new helper symlook_global(), and use it both in symlook_global() and get_program_var_addr(). Supply lock state to get_program_var_addr(). Reviewed by: kan Tested by: Mykola Dzham <i levsha me>
2011-01-10 16:09:35 +00:00
get_program_var_addr(const char *name, RtldLockState *lockstate)
After relocating the main program, but before calling any of the _init() functions, initialize the global variables "__progname" and "environ". This makes it possible for the _init() functions to call things like getenv() and err().
1999-04-21 04:06:57 +00:00
{
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
SymLook req;
get_program_var_addr() must prefer the strong symbol to the weak one. Search global objects, together with main object and dependencies, for the requested symbol. Move the common code from symlook_default() into new helper symlook_global(), and use it both in symlook_global() and get_program_var_addr(). Supply lock state to get_program_var_addr(). Reviewed by: kan Tested by: Mykola Dzham <i levsha me>
2011-01-10 16:09:35 +00:00
DoneList donelist;
After relocating the main program, but before calling any of the _init() functions, initialize the global variables "__progname" and "environ". This makes it possible for the _init() functions to call things like getenv() and err().
1999-04-21 04:06:57 +00:00
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
symlook_init(&req, name);
get_program_var_addr() must prefer the strong symbol to the weak one. Search global objects, together with main object and dependencies, for the requested symbol. Move the common code from symlook_default() into new helper symlook_global(), and use it both in symlook_global() and get_program_var_addr(). Supply lock state to get_program_var_addr(). Reviewed by: kan Tested by: Mykola Dzham <i levsha me>
2011-01-10 16:09:35 +00:00
req.lockstate = lockstate;
donelist_init(&donelist);
if (symlook_global(&req, &donelist) != 0)
return (NULL);
Fix get_program_var_addr() when type of the resolved symbol is functional. Use make_function_pointer then, otherwise ia64 is broken. Reported and tested by: marcel
2011-01-25 21:04:55 +00:00
if (ELF_ST_TYPE(req.sym_out->st_info) == STT_FUNC)
return ((const void **)make_function_pointer(req.sym_out,
req.defobj_out));
Add support for STT_GNU_IFUNC and R_MACHINE_IRELATIVE GNU extensions to rtld on 386 and amd64. This adds runtime bits neccessary for the use of the dispatch functions from the dynamically-linked executables and shared libraries. To allow use of external references from the dispatch function, resolution of the R_MACHINE_IRESOLVE relocations in PLT is postponed until GOT entries for PLT are prepared, and normal resolution of the GOT entries is finished. Similar to how it is done by GNU, IRELATIVE relocations are resolved in advance, instead of normal lazy handling for PLT. Move the init_pltgot() call before the relocations for the object are processed. MFC after: 3 weeks
2011-12-12 11:03:14 +00:00
else if (ELF_ST_TYPE(req.sym_out->st_info) == STT_GNU_IFUNC)
return ((const void **)rtld_resolve_ifunc(req.defobj_out, req.sym_out));
Fix get_program_var_addr() when type of the resolved symbol is functional. Use make_function_pointer then, otherwise ia64 is broken. Reported and tested by: marcel
2011-01-25 21:04:55 +00:00
else
return ((const void **)(req.defobj_out->relocbase +
req.sym_out->st_value));
Allow ldd(1) be used on shared libraries in addition to executables.
2002-02-04 10:33:48 +00:00
}
/*
* Set a pointer variable in the main program to the given value. This
* is used to set key variables such as "environ" before any of the
* init functions are called.
*/
static void
set_program_var(const char *name, const void *value)
{
const void **addr;
get_program_var_addr() must prefer the strong symbol to the weak one. Search global objects, together with main object and dependencies, for the requested symbol. Move the common code from symlook_default() into new helper symlook_global(), and use it both in symlook_global() and get_program_var_addr(). Supply lock state to get_program_var_addr(). Reviewed by: kan Tested by: Mykola Dzham <i levsha me>
2011-01-10 16:09:35 +00:00
if ((addr = get_program_var_addr(name, NULL)) != NULL) {
Allow ldd(1) be used on shared libraries in addition to executables.
2002-02-04 10:33:48 +00:00
dbg("\"%s\": *%p <-- %p", name, addr, value);
*addr = value;
}
After relocating the main program, but before calling any of the _init() functions, initialize the global variables "__progname" and "environ". This makes it possible for the _init() functions to call things like getenv() and err().
1999-04-21 04:06:57 +00:00
}
get_program_var_addr() must prefer the strong symbol to the weak one. Search global objects, together with main object and dependencies, for the requested symbol. Move the common code from symlook_default() into new helper symlook_global(), and use it both in symlook_global() and get_program_var_addr(). Supply lock state to get_program_var_addr(). Reviewed by: kan Tested by: Mykola Dzham <i levsha me>
2011-01-10 16:09:35 +00:00
/*
* Search the global objects, including dependencies and main object,
* for the given symbol.
*/
static int
symlook_global(SymLook *req, DoneList *donelist)
{
SymLook req1;
const Objlist_Entry *elm;
int res;
symlook_init_from_req(&req1, req);
/* Search all objects loaded at program start up. */
if (req->defobj_out == NULL ||
ELF_ST_BIND(req->sym_out->st_info) == STB_WEAK) {
res = symlook_list(&req1, &list_main, donelist);
if (res == 0 && (req->defobj_out == NULL ||
ELF_ST_BIND(req1.sym_out->st_info) != STB_WEAK)) {
req->sym_out = req1.sym_out;
req->defobj_out = req1.defobj_out;
assert(req->defobj_out != NULL);
}
}
/* Search all DAGs whose roots are RTLD_GLOBAL objects. */
STAILQ_FOREACH(elm, &list_global, link) {
if (req->defobj_out != NULL &&
ELF_ST_BIND(req->sym_out->st_info) != STB_WEAK)
break;
res = symlook_list(&req1, &elm->obj->dagmembers, donelist);
if (res == 0 && (req->defobj_out == NULL ||
ELF_ST_BIND(req1.sym_out->st_info) != STB_WEAK)) {
req->sym_out = req1.sym_out;
req->defobj_out = req1.defobj_out;
assert(req->defobj_out != NULL);
}
}
return (req->sym_out != NULL ? 0 : ESRCH);
}
Add support for dlsym(RTLD_DEFAULT, ...).
2000-09-19 04:27:16 +00:00
/*
* Given a symbol name in a referencing object, find the corresponding
* definition of the symbol. Returns a pointer to the symbol, or NULL if
* no definition was found. Returns a pointer to the Obj_Entry of the
* defining object via the reference parameter DEFOBJ_OUT.
*/
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
static int
symlook_default(SymLook *req, const Obj_Entry *refobj)
Add support for dlsym(RTLD_DEFAULT, ...).
2000-09-19 04:27:16 +00:00
{
DoneList donelist;
const Objlist_Entry *elm;
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
SymLook req1;
int res;
get_program_var_addr() must prefer the strong symbol to the weak one. Search global objects, together with main object and dependencies, for the requested symbol. Move the common code from symlook_default() into new helper symlook_global(), and use it both in symlook_global() and get_program_var_addr(). Supply lock state to get_program_var_addr(). Reviewed by: kan Tested by: Mykola Dzham <i levsha me>
2011-01-10 16:09:35 +00:00
Add support for dlsym(RTLD_DEFAULT, ...).
2000-09-19 04:27:16 +00:00
donelist_init(&donelist);
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
symlook_init_from_req(&req1, req);
Add support for dlsym(RTLD_DEFAULT, ...).
2000-09-19 04:27:16 +00:00
Handle protected symbols in rtld. Protected symbol reference in GOT of the defining object must be resolved to itself, same as -Bsymbolic globally. Discussed with: emaste Sponsored by: The FreeBSD Foundation MFC after: 2 weeks Differential revision: https://reviews.freebsd.org/D9317
2017-02-09 23:33:06 +00:00
/*
* Look first in the referencing object if linked symbolically,
* and similarly handle protected symbols.
*/
res = symlook_obj(&req1, refobj);
if (res == 0 && (refobj->symbolic ||
ELF_ST_VISIBILITY(req1.sym_out->st_other) == STV_PROTECTED)) {
req->sym_out = req1.sym_out;
req->defobj_out = req1.defobj_out;
assert(req->defobj_out != NULL);
}
if (refobj->symbolic || req->defobj_out != NULL)
donelist_check(&donelist, refobj);
Add support for dlsym(RTLD_DEFAULT, ...).
2000-09-19 04:27:16 +00:00
get_program_var_addr() must prefer the strong symbol to the weak one. Search global objects, together with main object and dependencies, for the requested symbol. Move the common code from symlook_default() into new helper symlook_global(), and use it both in symlook_global() and get_program_var_addr(). Supply lock state to get_program_var_addr(). Reviewed by: kan Tested by: Mykola Dzham <i levsha me>
2011-01-10 16:09:35 +00:00
symlook_global(req, &donelist);
Add support for dlsym(RTLD_DEFAULT, ...).
2000-09-19 04:27:16 +00:00
Change the symbol lookup order to search RTLD_GLOBAL objects before referencing object's DAG. This makes it possible for C++ exceptions to work across shared libraries and brings us closer to the search order used by Solaris/Linux. Reviewed by: jdp Approved by: obrien MFC after: 1 month
2002-10-19 23:03:35 +00:00
/* Search all dlopened DAGs containing the referencing object. */
STAILQ_FOREACH(elm, &refobj->dldags, link) {
get_program_var_addr() must prefer the strong symbol to the weak one. Search global objects, together with main object and dependencies, for the requested symbol. Move the common code from symlook_default() into new helper symlook_global(), and use it both in symlook_global() and get_program_var_addr(). Supply lock state to get_program_var_addr(). Reviewed by: kan Tested by: Mykola Dzham <i levsha me>
2011-01-10 16:09:35 +00:00
if (req->sym_out != NULL &&
ELF_ST_BIND(req->sym_out->st_info) != STB_WEAK)
Change the symbol lookup order to search RTLD_GLOBAL objects before referencing object's DAG. This makes it possible for C++ exceptions to work across shared libraries and brings us closer to the search order used by Solaris/Linux. Reviewed by: jdp Approved by: obrien MFC after: 1 month
2002-10-19 23:03:35 +00:00
break;
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
res = symlook_list(&req1, &elm->obj->dagmembers, &donelist);
get_program_var_addr() must prefer the strong symbol to the weak one. Search global objects, together with main object and dependencies, for the requested symbol. Move the common code from symlook_default() into new helper symlook_global(), and use it both in symlook_global() and get_program_var_addr(). Supply lock state to get_program_var_addr(). Reviewed by: kan Tested by: Mykola Dzham <i levsha me>
2011-01-10 16:09:35 +00:00
if (res == 0 && (req->sym_out == NULL ||
ELF_ST_BIND(req1.sym_out->st_info) != STB_WEAK)) {
req->sym_out = req1.sym_out;
req->defobj_out = req1.defobj_out;
assert(req->defobj_out != NULL);
Add support for dlsym(RTLD_DEFAULT, ...).
2000-09-19 04:27:16 +00:00
}
}
/*
* Search the dynamic linker itself, and possibly resolve the
* symbol from there. This is how the application links to
Remove exports table. Export control by the version script is enough. Reviewed by: kan MFC after: 3 weeks
2010-08-24 13:01:14 +00:00
* dynamic linker services such as dlopen.
Add support for dlsym(RTLD_DEFAULT, ...).
2000-09-19 04:27:16 +00:00
*/
get_program_var_addr() must prefer the strong symbol to the weak one. Search global objects, together with main object and dependencies, for the requested symbol. Move the common code from symlook_default() into new helper symlook_global(), and use it both in symlook_global() and get_program_var_addr(). Supply lock state to get_program_var_addr(). Reviewed by: kan Tested by: Mykola Dzham <i levsha me>
2011-01-10 16:09:35 +00:00
if (req->sym_out == NULL ||
ELF_ST_BIND(req->sym_out->st_info) == STB_WEAK) {
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
res = symlook_obj(&req1, &obj_rtld);
if (res == 0) {
get_program_var_addr() must prefer the strong symbol to the weak one. Search global objects, together with main object and dependencies, for the requested symbol. Move the common code from symlook_default() into new helper symlook_global(), and use it both in symlook_global() and get_program_var_addr(). Supply lock state to get_program_var_addr(). Reviewed by: kan Tested by: Mykola Dzham <i levsha me>
2011-01-10 16:09:35 +00:00
req->sym_out = req1.sym_out;
req->defobj_out = req1.defobj_out;
assert(req->defobj_out != NULL);
Add support for dlsym(RTLD_DEFAULT, ...).
2000-09-19 04:27:16 +00:00
}
}
get_program_var_addr() must prefer the strong symbol to the weak one. Search global objects, together with main object and dependencies, for the requested symbol. Move the common code from symlook_default() into new helper symlook_global(), and use it both in symlook_global() and get_program_var_addr(). Supply lock state to get_program_var_addr(). Reviewed by: kan Tested by: Mykola Dzham <i levsha me>
2011-01-10 16:09:35 +00:00
return (req->sym_out != NULL ? 0 : ESRCH);
Add support for dlsym(RTLD_DEFAULT, ...).
2000-09-19 04:27:16 +00:00
}
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
static int
symlook_list(SymLook *req, const Objlist *objlist, DoneList *dlp)
Revamp the symbol lookup algorithm to cope better with objects loaded separately by dlopen that have global symbols with identical names. Viewing each dlopened object as a DAG which is linked by its DT_NEEDED entries in the dynamic table, the search order is as follows: * If the referencing object was linked with -Bsymbolic, search it internally. * Search all dlopened DAGs containing the referencing object. * Search all objects loaded at program start up. * Search all objects which were dlopened() using the RTLD_GLOBAL flag (which is now supported too). The search terminates as soon as a strong definition is found. Lacking that, the first weak definition is used. These rules match those of Solaris, as best I could determine them from its vague manual pages and the results of experiments I performed. PR: misc/12438
1999-08-30 01:48:19 +00:00
{
const Elf_Sym *def;
const Obj_Entry *defobj;
const Objlist_Entry *elm;
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
SymLook req1;
int res;
Revamp the symbol lookup algorithm to cope better with objects loaded separately by dlopen that have global symbols with identical names. Viewing each dlopened object as a DAG which is linked by its DT_NEEDED entries in the dynamic table, the search order is as follows: * If the referencing object was linked with -Bsymbolic, search it internally. * Search all dlopened DAGs containing the referencing object. * Search all objects loaded at program start up. * Search all objects which were dlopened() using the RTLD_GLOBAL flag (which is now supported too). The search terminates as soon as a strong definition is found. Lacking that, the first weak definition is used. These rules match those of Solaris, as best I could determine them from its vague manual pages and the results of experiments I performed. PR: misc/12438
1999-08-30 01:48:19 +00:00
def = NULL;
defobj = NULL;
STAILQ_FOREACH(elm, objlist, link) {
Solve the dynamic linker's problems with multithreaded programs once and for all (I hope). Packages such as wine, JDK, and linuxthreads should no longer have any problems with re-entering the dynamic linker. This commit replaces the locking used in the dynamic linker with a new spinlock-based reader/writer lock implementation. Brian Fundakowski Feldman <green> argued for this from the very beginning, but it took me a long time to come around to his point of view. Spinlocks are the only kinds of locks that work with all thread packages. But on uniprocessor systems they can be inefficient, because while a contender for the lock is spinning the holder of the lock cannot make any progress toward releasing it. To alleviate this disadvantage I have borrowed a trick from Sleepycat's Berkeley DB implementation. When spinning for a lock, the requester does a nanosleep() call for 1 usec. each time around the loop. This will generally yield the CPU to other threads, allowing the lock holder to finish its business and release the lock. I chose 1 usec. as the minimum sleep which would with reasonable certainty not be rounded down to 0. The formerly machine-independent file "lockdflt.c" has been moved into the architecture-specific subdirectories by repository copy. It now contains the machine-dependent spinlocking code. For the spinlocks I used the very nifty "simple, non-scalable reader-preference lock" which I found at <http://www.cs.rochester.edu/u/scott/synchronization/pseudocode/rw.html> on all CPUs except the 80386 (the specific CPU model, not the architecture). The 80386 CPU doesn't support the necessary "cmpxchg" instruction, so on that CPU a simple exclusive test-and-set lock is used instead. 80386 CPUs are detected at initialization time by trying to execute "cmpxchg" and catching the resulting SIGILL signal. To reduce contention for the locks, I have revamped a couple of key data structures, permitting all common operations to be done under non-exclusive (reader) locking. The only operations that require exclusive locking now are the rare intrusive operations such as dlopen() and dlclose(). The dllockinit() interface is now deprecated. It still exists, but only as a do-nothing stub. I plan to remove it as soon as is reasonably possible. (From the very beginning it was clearly labeled as experimental and subject to change.) As far as I know, only the linuxthreads port uses dllockinit(). This interface turned out to have several problems. As one example, when the dynamic linker called a client-supplied locking function, that function sometimes needed lazy binding, causing re-entry into the dynamic linker and a big looping mess. And in any case, it turned out to be too burdensome to require threads packages to register themselves with the dynamic linker.
2000-07-08 04:10:38 +00:00
if (donelist_check(dlp, elm->obj))
Revamp the symbol lookup algorithm to cope better with objects loaded separately by dlopen that have global symbols with identical names. Viewing each dlopened object as a DAG which is linked by its DT_NEEDED entries in the dynamic table, the search order is as follows: * If the referencing object was linked with -Bsymbolic, search it internally. * Search all dlopened DAGs containing the referencing object. * Search all objects loaded at program start up. * Search all objects which were dlopened() using the RTLD_GLOBAL flag (which is now supported too). The search terminates as soon as a strong definition is found. Lacking that, the first weak definition is used. These rules match those of Solaris, as best I could determine them from its vague manual pages and the results of experiments I performed. PR: misc/12438
1999-08-30 01:48:19 +00:00
continue;
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
symlook_init_from_req(&req1, req);
if ((res = symlook_obj(&req1, elm->obj)) == 0) {
if (def == NULL || ELF_ST_BIND(req1.sym_out->st_info) != STB_WEAK) {
def = req1.sym_out;
defobj = req1.defobj_out;
Revamp the symbol lookup algorithm to cope better with objects loaded separately by dlopen that have global symbols with identical names. Viewing each dlopened object as a DAG which is linked by its DT_NEEDED entries in the dynamic table, the search order is as follows: * If the referencing object was linked with -Bsymbolic, search it internally. * Search all dlopened DAGs containing the referencing object. * Search all objects loaded at program start up. * Search all objects which were dlopened() using the RTLD_GLOBAL flag (which is now supported too). The search terminates as soon as a strong definition is found. Lacking that, the first weak definition is used. These rules match those of Solaris, as best I could determine them from its vague manual pages and the results of experiments I performed. PR: misc/12438
1999-08-30 01:48:19 +00:00
if (ELF_ST_BIND(def->st_info) != STB_WEAK)
break;
}
}
}
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
if (def != NULL) {
req->sym_out = def;
req->defobj_out = defobj;
return (0);
}
return (ESRCH);
Revamp the symbol lookup algorithm to cope better with objects loaded separately by dlopen that have global symbols with identical names. Viewing each dlopened object as a DAG which is linked by its DT_NEEDED entries in the dynamic table, the search order is as follows: * If the referencing object was linked with -Bsymbolic, search it internally. * Search all dlopened DAGs containing the referencing object. * Search all objects loaded at program start up. * Search all objects which were dlopened() using the RTLD_GLOBAL flag (which is now supported too). The search terminates as soon as a strong definition is found. Lacking that, the first weak definition is used. These rules match those of Solaris, as best I could determine them from its vague manual pages and the results of experiments I performed. PR: misc/12438
1999-08-30 01:48:19 +00:00
}
Fix the buggy rev. 1.117. dagmembers are only initialized for dlopen'ed dso that are actually loading. If dso a.so depends on b.so, then dlsym with handle from dlopen("b.so") will fail unconditionally. Correct implementation shall use the Obj_Entry.needed list to walk dependencies DAG. Test provided by: jkim Tested (prev. version) by: jkim, Nicolas Blais <nb_root at videotron ca>, h.blanke at chello nl Pointy hat to: kib Approved by: kan (mentor)
2006-09-19 12:47:13 +00:00
/*
Eliminate the use of symlook_needed function in favor of DAGS. Place elements on DAG lists in breadth-first order. This allows us to walk pre-built list in all cases where breadth-first dependency chain enumeration is required. Fix dlsym on special handle obtained by dlopen(NULL, ...) to do what comment claims it does. Take advantage of recently added symlook_global function to iterate over main objects and global DAGs lists properly in search of a symbol. Since rtld itself provides part of the global namespace, search rtld_obj too. Remove recursion from init_dag and symlook_needed functions. Use symlook_needed for ELF filtee processing only and change lookup order used in the function to match the order used by Solaris runtime linker under same circumstances. While there, fix weak symbol handling in the loop so that we return the first weak symbol definition if no strong one was found, instead of the last one. Reviewed by: kib MFC after: 1 month
2011-01-28 23:44:57 +00:00
* Search the chain of DAGS cointed to by the given Needed_Entry
* for a symbol of the given name. Each DAG is scanned completely
* before advancing to the next one. Returns a pointer to the symbol,
* or NULL if no definition was found.
Fix the buggy rev. 1.117. dagmembers are only initialized for dlopen'ed dso that are actually loading. If dso a.so depends on b.so, then dlsym with handle from dlopen("b.so") will fail unconditionally. Correct implementation shall use the Obj_Entry.needed list to walk dependencies DAG. Test provided by: jkim Tested (prev. version) by: jkim, Nicolas Blais <nb_root at videotron ca>, h.blanke at chello nl Pointy hat to: kib Approved by: kan (mentor)
2006-09-19 12:47:13 +00:00
*/
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
static int
symlook_needed(SymLook *req, const Needed_Entry *needed, DoneList *dlp)
Fix the buggy rev. 1.117. dagmembers are only initialized for dlopen'ed dso that are actually loading. If dso a.so depends on b.so, then dlsym with handle from dlopen("b.so") will fail unconditionally. Correct implementation shall use the Obj_Entry.needed list to walk dependencies DAG. Test provided by: jkim Tested (prev. version) by: jkim, Nicolas Blais <nb_root at videotron ca>, h.blanke at chello nl Pointy hat to: kib Approved by: kan (mentor)
2006-09-19 12:47:13 +00:00
{
Eliminate the use of symlook_needed function in favor of DAGS. Place elements on DAG lists in breadth-first order. This allows us to walk pre-built list in all cases where breadth-first dependency chain enumeration is required. Fix dlsym on special handle obtained by dlopen(NULL, ...) to do what comment claims it does. Take advantage of recently added symlook_global function to iterate over main objects and global DAGs lists properly in search of a symbol. Since rtld itself provides part of the global namespace, search rtld_obj too. Remove recursion from init_dag and symlook_needed functions. Use symlook_needed for ELF filtee processing only and change lookup order used in the function to match the order used by Solaris runtime linker under same circumstances. While there, fix weak symbol handling in the loop so that we return the first weak symbol definition if no strong one was found, instead of the last one. Reviewed by: kib MFC after: 1 month
2011-01-28 23:44:57 +00:00
const Elf_Sym *def;
Fix the buggy rev. 1.117. dagmembers are only initialized for dlopen'ed dso that are actually loading. If dso a.so depends on b.so, then dlsym with handle from dlopen("b.so") will fail unconditionally. Correct implementation shall use the Obj_Entry.needed list to walk dependencies DAG. Test provided by: jkim Tested (prev. version) by: jkim, Nicolas Blais <nb_root at videotron ca>, h.blanke at chello nl Pointy hat to: kib Approved by: kan (mentor)
2006-09-19 12:47:13 +00:00
const Needed_Entry *n;
Eliminate the use of symlook_needed function in favor of DAGS. Place elements on DAG lists in breadth-first order. This allows us to walk pre-built list in all cases where breadth-first dependency chain enumeration is required. Fix dlsym on special handle obtained by dlopen(NULL, ...) to do what comment claims it does. Take advantage of recently added symlook_global function to iterate over main objects and global DAGs lists properly in search of a symbol. Since rtld itself provides part of the global namespace, search rtld_obj too. Remove recursion from init_dag and symlook_needed functions. Use symlook_needed for ELF filtee processing only and change lookup order used in the function to match the order used by Solaris runtime linker under same circumstances. While there, fix weak symbol handling in the loop so that we return the first weak symbol definition if no strong one was found, instead of the last one. Reviewed by: kib MFC after: 1 month
2011-01-28 23:44:57 +00:00
const Obj_Entry *defobj;
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
SymLook req1;
int res;
Clean up white spaces and fix style(9).
2006-09-19 16:48:08 +00:00
Eliminate the use of symlook_needed function in favor of DAGS. Place elements on DAG lists in breadth-first order. This allows us to walk pre-built list in all cases where breadth-first dependency chain enumeration is required. Fix dlsym on special handle obtained by dlopen(NULL, ...) to do what comment claims it does. Take advantage of recently added symlook_global function to iterate over main objects and global DAGs lists properly in search of a symbol. Since rtld itself provides part of the global namespace, search rtld_obj too. Remove recursion from init_dag and symlook_needed functions. Use symlook_needed for ELF filtee processing only and change lookup order used in the function to match the order used by Solaris runtime linker under same circumstances. While there, fix weak symbol handling in the loop so that we return the first weak symbol definition if no strong one was found, instead of the last one. Reviewed by: kib MFC after: 1 month
2011-01-28 23:44:57 +00:00
def = NULL;
Fix the buggy rev. 1.117. dagmembers are only initialized for dlopen'ed dso that are actually loading. If dso a.so depends on b.so, then dlsym with handle from dlopen("b.so") will fail unconditionally. Correct implementation shall use the Obj_Entry.needed list to walk dependencies DAG. Test provided by: jkim Tested (prev. version) by: jkim, Nicolas Blais <nb_root at videotron ca>, h.blanke at chello nl Pointy hat to: kib Approved by: kan (mentor)
2006-09-19 12:47:13 +00:00
defobj = NULL;
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
symlook_init_from_req(&req1, req);
Fix the buggy rev. 1.117. dagmembers are only initialized for dlopen'ed dso that are actually loading. If dso a.so depends on b.so, then dlsym with handle from dlopen("b.so") will fail unconditionally. Correct implementation shall use the Obj_Entry.needed list to walk dependencies DAG. Test provided by: jkim Tested (prev. version) by: jkim, Nicolas Blais <nb_root at videotron ca>, h.blanke at chello nl Pointy hat to: kib Approved by: kan (mentor)
2006-09-19 12:47:13 +00:00
for (n = needed; n != NULL; n = n->next) {
Eliminate the use of symlook_needed function in favor of DAGS. Place elements on DAG lists in breadth-first order. This allows us to walk pre-built list in all cases where breadth-first dependency chain enumeration is required. Fix dlsym on special handle obtained by dlopen(NULL, ...) to do what comment claims it does. Take advantage of recently added symlook_global function to iterate over main objects and global DAGs lists properly in search of a symbol. Since rtld itself provides part of the global namespace, search rtld_obj too. Remove recursion from init_dag and symlook_needed functions. Use symlook_needed for ELF filtee processing only and change lookup order used in the function to match the order used by Solaris runtime linker under same circumstances. While there, fix weak symbol handling in the loop so that we return the first weak symbol definition if no strong one was found, instead of the last one. Reviewed by: kib MFC after: 1 month
2011-01-28 23:44:57 +00:00
if (n->obj == NULL ||
(res = symlook_list(&req1, &n->obj->dagmembers, dlp)) != 0)
Clean up white spaces and fix style(9).
2006-09-19 16:48:08 +00:00
continue;
Eliminate the use of symlook_needed function in favor of DAGS. Place elements on DAG lists in breadth-first order. This allows us to walk pre-built list in all cases where breadth-first dependency chain enumeration is required. Fix dlsym on special handle obtained by dlopen(NULL, ...) to do what comment claims it does. Take advantage of recently added symlook_global function to iterate over main objects and global DAGs lists properly in search of a symbol. Since rtld itself provides part of the global namespace, search rtld_obj too. Remove recursion from init_dag and symlook_needed functions. Use symlook_needed for ELF filtee processing only and change lookup order used in the function to match the order used by Solaris runtime linker under same circumstances. While there, fix weak symbol handling in the loop so that we return the first weak symbol definition if no strong one was found, instead of the last one. Reviewed by: kib MFC after: 1 month
2011-01-28 23:44:57 +00:00
if (def == NULL || ELF_ST_BIND(req1.sym_out->st_info) != STB_WEAK) {
def = req1.sym_out;
defobj = req1.defobj_out;
if (ELF_ST_BIND(def->st_info) != STB_WEAK)
break;
Clean up white spaces and fix style(9).
2006-09-19 16:48:08 +00:00
}
Fix the buggy rev. 1.117. dagmembers are only initialized for dlopen'ed dso that are actually loading. If dso a.so depends on b.so, then dlsym with handle from dlopen("b.so") will fail unconditionally. Correct implementation shall use the Obj_Entry.needed list to walk dependencies DAG. Test provided by: jkim Tested (prev. version) by: jkim, Nicolas Blais <nb_root at videotron ca>, h.blanke at chello nl Pointy hat to: kib Approved by: kan (mentor)
2006-09-19 12:47:13 +00:00
}
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
if (def != NULL) {
req->sym_out = def;
req->defobj_out = defobj;
return (0);
}
return (ESRCH);
Fix the buggy rev. 1.117. dagmembers are only initialized for dlopen'ed dso that are actually loading. If dso a.so depends on b.so, then dlsym with handle from dlopen("b.so") will fail unconditionally. Correct implementation shall use the Obj_Entry.needed list to walk dependencies DAG. Test provided by: jkim Tested (prev. version) by: jkim, Nicolas Blais <nb_root at videotron ca>, h.blanke at chello nl Pointy hat to: kib Approved by: kan (mentor)
2006-09-19 12:47:13 +00:00
}
Style fixes. If it seems like a lot of lines of changes, it's because I moved some functions. Mr. Tidy likes them to be in alphabetical order.
1998-09-02 02:00:20 +00:00
/*
* Search the symbol table of a single shared object for a symbol of
Implement ELF symbol versioning using GNU semantics. This code aims to be compatible with symbol versioning support as implemented by GNU libc and documented by http://people.redhat.com/~drepper/symbol-versioning and LSB 3.0. Implement dlvsym() function to allow lookups for a specific version of a given symbol.
2005-12-18 19:43:33 +00:00
* the given name and version, if requested. Returns a pointer to the
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
* symbol, or NULL if no definition was found. If the object is
* filter, return filtered symbol from filtee.
Style fixes. If it seems like a lot of lines of changes, it's because I moved some functions. Mr. Tidy likes them to be in alphabetical order.
1998-09-02 02:00:20 +00:00
*
* The symbol's hash value is passed in for efficiency reasons; that
* eliminates many recomputations of the hash value.
*/
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
int
symlook_obj(SymLook *req, const Obj_Entry *obj)
{
DoneList donelist;
SymLook req1;
Fix several problems with our ELF filters implementation. Do not relocate twice an object which happens to be needed by loaded binary (or dso) and some filtee opened due to symbol resolution when relocating need objects. Record the state of the relocation processing in Obj_Entry and short-circuit relocate_objects() if current object already processed. Do not call constructors for filtees loaded during the early relocation processing before image is initialized enough to run user-provided code. Filtees are loaded using dlopen_object(), which normally performs relocation and initialization. If filtee is lazy-loaded during the relocation of dso needed by the main object, dlopen_object() runs too earlier, when most runtime services are not yet ready. Postpone the constructors call to the time when main binary and depended libraries constructors are run, passing the new flag RTLD_LO_EARLY to dlopen_object(). Symbol lookups callers inform symlook_* functions about early stage of initialization with SYMLOOK_EARLY. Pass flags through all functions participating in object relocation. Use the opportunity and fix flags argument to find_symdef() in arch-specific reloc.c to use proper name SYMLOOK_IN_PLT instead of true, which happen to have the same numeric value. Reported and tested by: theraven Reviewed by: kan MFC after: 2 weeks
2012-03-20 13:20:49 +00:00
int flags, res, mres;
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
Add GNU hash support for rtld. Based on dragonflybsd support for GNU hash by John Marino <draco marino st> Reviewed by: kan Tested by: bapt MFC after: 2 weeks
2012-04-30 13:31:10 +00:00
/*
Work around a situation where symlook_obj() could be called for the object for which digest_dynamic1() was not done yet. Just return EINVAL and do not try to dereference NULL buckets hash array. This seems to happen on ia64 for rtld object itself, where the R_IA_64_FPTR64LSB relocations require symbol lookup. The dynamic linker itself does not rely on identity of the C-level function pointers (i.e. function descriptors). Reported and reviewed by: marcel MFC after: 8 days
2012-05-05 11:26:08 +00:00
* If there is at least one valid hash at this point, we prefer to
* use the faster GNU version if available.
Add GNU hash support for rtld. Based on dragonflybsd support for GNU hash by John Marino <draco marino st> Reviewed by: kan Tested by: bapt MFC after: 2 weeks
2012-04-30 13:31:10 +00:00
*/
if (obj->valid_hash_gnu)
mres = symlook_obj1_gnu(req, obj);
Work around a situation where symlook_obj() could be called for the object for which digest_dynamic1() was not done yet. Just return EINVAL and do not try to dereference NULL buckets hash array. This seems to happen on ia64 for rtld object itself, where the R_IA_64_FPTR64LSB relocations require symbol lookup. The dynamic linker itself does not rely on identity of the C-level function pointers (i.e. function descriptors). Reported and reviewed by: marcel MFC after: 8 days
2012-05-05 11:26:08 +00:00
else if (obj->valid_hash_sysv)
Add GNU hash support for rtld. Based on dragonflybsd support for GNU hash by John Marino <draco marino st> Reviewed by: kan Tested by: bapt MFC after: 2 weeks
2012-04-30 13:31:10 +00:00
mres = symlook_obj1_sysv(req, obj);
Work around a situation where symlook_obj() could be called for the object for which digest_dynamic1() was not done yet. Just return EINVAL and do not try to dereference NULL buckets hash array. This seems to happen on ia64 for rtld object itself, where the R_IA_64_FPTR64LSB relocations require symbol lookup. The dynamic linker itself does not rely on identity of the C-level function pointers (i.e. function descriptors). Reported and reviewed by: marcel MFC after: 8 days
2012-05-05 11:26:08 +00:00
else
return (EINVAL);
Add GNU hash support for rtld. Based on dragonflybsd support for GNU hash by John Marino <draco marino st> Reviewed by: kan Tested by: bapt MFC after: 2 weeks
2012-04-30 13:31:10 +00:00
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
if (mres == 0) {
if (obj->needed_filtees != NULL) {
Fix several problems with our ELF filters implementation. Do not relocate twice an object which happens to be needed by loaded binary (or dso) and some filtee opened due to symbol resolution when relocating need objects. Record the state of the relocation processing in Obj_Entry and short-circuit relocate_objects() if current object already processed. Do not call constructors for filtees loaded during the early relocation processing before image is initialized enough to run user-provided code. Filtees are loaded using dlopen_object(), which normally performs relocation and initialization. If filtee is lazy-loaded during the relocation of dso needed by the main object, dlopen_object() runs too earlier, when most runtime services are not yet ready. Postpone the constructors call to the time when main binary and depended libraries constructors are run, passing the new flag RTLD_LO_EARLY to dlopen_object(). Symbol lookups callers inform symlook_* functions about early stage of initialization with SYMLOOK_EARLY. Pass flags through all functions participating in object relocation. Use the opportunity and fix flags argument to find_symdef() in arch-specific reloc.c to use proper name SYMLOOK_IN_PLT instead of true, which happen to have the same numeric value. Reported and tested by: theraven Reviewed by: kan MFC after: 2 weeks
2012-03-20 13:20:49 +00:00
flags = (req->flags & SYMLOOK_EARLY) ? RTLD_LO_EARLY : 0;
load_filtees(__DECONST(Obj_Entry *, obj), flags, req->lockstate);
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
donelist_init(&donelist);
symlook_init_from_req(&req1, req);
res = symlook_needed(&req1, obj->needed_filtees, &donelist);
if (res == 0) {
req->sym_out = req1.sym_out;
req->defobj_out = req1.defobj_out;
}
return (res);
}
if (obj->needed_aux_filtees != NULL) {
Fix several problems with our ELF filters implementation. Do not relocate twice an object which happens to be needed by loaded binary (or dso) and some filtee opened due to symbol resolution when relocating need objects. Record the state of the relocation processing in Obj_Entry and short-circuit relocate_objects() if current object already processed. Do not call constructors for filtees loaded during the early relocation processing before image is initialized enough to run user-provided code. Filtees are loaded using dlopen_object(), which normally performs relocation and initialization. If filtee is lazy-loaded during the relocation of dso needed by the main object, dlopen_object() runs too earlier, when most runtime services are not yet ready. Postpone the constructors call to the time when main binary and depended libraries constructors are run, passing the new flag RTLD_LO_EARLY to dlopen_object(). Symbol lookups callers inform symlook_* functions about early stage of initialization with SYMLOOK_EARLY. Pass flags through all functions participating in object relocation. Use the opportunity and fix flags argument to find_symdef() in arch-specific reloc.c to use proper name SYMLOOK_IN_PLT instead of true, which happen to have the same numeric value. Reported and tested by: theraven Reviewed by: kan MFC after: 2 weeks
2012-03-20 13:20:49 +00:00
flags = (req->flags & SYMLOOK_EARLY) ? RTLD_LO_EARLY : 0;
load_filtees(__DECONST(Obj_Entry *, obj), flags, req->lockstate);
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
donelist_init(&donelist);
symlook_init_from_req(&req1, req);
res = symlook_needed(&req1, obj->needed_aux_filtees, &donelist);
if (res == 0) {
req->sym_out = req1.sym_out;
req->defobj_out = req1.defobj_out;
return (res);
}
}
}
return (mres);
}
Split the symlook_obj1 into a loop iterating over the ELF object symbol hash elements, and a helper matched_symbol() which match the given hash entry and request, performing needed type and version checks. Based on dragonflybsd support for GNU hash by John Marino <draco marino st> Reviewed by: kan Tested by: bapt MFC after: 2 weeks
2012-04-30 13:29:21 +00:00
/* Symbol match routine common to both hash functions */
static bool
matched_symbol(SymLook *req, const Obj_Entry *obj, Sym_Match_Result *result,
const unsigned long symnum)
Style fixes. If it seems like a lot of lines of changes, it's because I moved some functions. Mr. Tidy likes them to be in alphabetical order.
1998-09-02 02:00:20 +00:00
{
Split the symlook_obj1 into a loop iterating over the ELF object symbol hash elements, and a helper matched_symbol() which match the given hash entry and request, performing needed type and version checks. Based on dragonflybsd support for GNU hash by John Marino <draco marino st> Reviewed by: kan Tested by: bapt MFC after: 2 weeks
2012-04-30 13:29:21 +00:00
Elf_Versym verndx;
Implement ELF symbol versioning using GNU semantics. This code aims to be compatible with symbol versioning support as implemented by GNU libc and documented by http://people.redhat.com/~drepper/symbol-versioning and LSB 3.0. Implement dlvsym() function to allow lookups for a specific version of a given symbol.
2005-12-18 19:43:33 +00:00
const Elf_Sym *symp;
const char *strp;
symp = obj->symtab + symnum;
strp = obj->strtab + symp->st_name;
switch (ELF_ST_TYPE(symp->st_info)) {
case STT_FUNC:
case STT_NOTYPE:
case STT_OBJECT:
Split the symlook_obj1 into a loop iterating over the ELF object symbol hash elements, and a helper matched_symbol() which match the given hash entry and request, performing needed type and version checks. Based on dragonflybsd support for GNU hash by John Marino <draco marino st> Reviewed by: kan Tested by: bapt MFC after: 2 weeks
2012-04-30 13:29:21 +00:00
case STT_COMMON:
Add support for STT_GNU_IFUNC and R_MACHINE_IRELATIVE GNU extensions to rtld on 386 and amd64. This adds runtime bits neccessary for the use of the dispatch functions from the dynamically-linked executables and shared libraries. To allow use of external references from the dispatch function, resolution of the R_MACHINE_IRESOLVE relocations in PLT is postponed until GOT entries for PLT are prepared, and normal resolution of the GOT entries is finished. Similar to how it is done by GNU, IRELATIVE relocations are resolved in advance, instead of normal lazy handling for PLT. Move the init_pltgot() call before the relocations for the object are processed. MFC after: 3 weeks
2011-12-12 11:03:14 +00:00
case STT_GNU_IFUNC:
Split the symlook_obj1 into a loop iterating over the ELF object symbol hash elements, and a helper matched_symbol() which match the given hash entry and request, performing needed type and version checks. Based on dragonflybsd support for GNU hash by John Marino <draco marino st> Reviewed by: kan Tested by: bapt MFC after: 2 weeks
2012-04-30 13:29:21 +00:00
if (symp->st_value == 0)
return (false);
Implement ELF symbol versioning using GNU semantics. This code aims to be compatible with symbol versioning support as implemented by GNU libc and documented by http://people.redhat.com/~drepper/symbol-versioning and LSB 3.0. Implement dlvsym() function to allow lookups for a specific version of a given symbol.
2005-12-18 19:43:33 +00:00
/* fallthrough */
case STT_TLS:
Split the symlook_obj1 into a loop iterating over the ELF object symbol hash elements, and a helper matched_symbol() which match the given hash entry and request, performing needed type and version checks. Based on dragonflybsd support for GNU hash by John Marino <draco marino st> Reviewed by: kan Tested by: bapt MFC after: 2 weeks
2012-04-30 13:29:21 +00:00
if (symp->st_shndx != SHN_UNDEF)
break;
MFp4: Add mips support for dynamic linking. This code came from the merged mips2 and Juniper mips repositories. Warner Losh, Randall Seager, Oleksandr Tymoshenko and Olivier Houchard worked to merge, debug and integrate this code. This code may also contain code derived from NetBSD.
2008-04-04 20:59:26 +00:00
#ifndef __mips__
Split the symlook_obj1 into a loop iterating over the ELF object symbol hash elements, and a helper matched_symbol() which match the given hash entry and request, performing needed type and version checks. Based on dragonflybsd support for GNU hash by John Marino <draco marino st> Reviewed by: kan Tested by: bapt MFC after: 2 weeks
2012-04-30 13:29:21 +00:00
else if (((req->flags & SYMLOOK_IN_PLT) == 0) &&
(ELF_ST_TYPE(symp->st_info) == STT_FUNC))
break;
Implement ELF symbol versioning using GNU semantics. This code aims to be compatible with symbol versioning support as implemented by GNU libc and documented by http://people.redhat.com/~drepper/symbol-versioning and LSB 3.0. Implement dlvsym() function to allow lookups for a specific version of a given symbol.
2005-12-18 19:43:33 +00:00
/* fallthrough */
MFp4: Add mips support for dynamic linking. This code came from the merged mips2 and Juniper mips repositories. Warner Losh, Randall Seager, Oleksandr Tymoshenko and Olivier Houchard worked to merge, debug and integrate this code. This code may also contain code derived from NetBSD.
2008-04-04 20:59:26 +00:00
#endif
Implement ELF symbol versioning using GNU semantics. This code aims to be compatible with symbol versioning support as implemented by GNU libc and documented by http://people.redhat.com/~drepper/symbol-versioning and LSB 3.0. Implement dlvsym() function to allow lookups for a specific version of a given symbol.
2005-12-18 19:43:33 +00:00
default:
Split the symlook_obj1 into a loop iterating over the ELF object symbol hash elements, and a helper matched_symbol() which match the given hash entry and request, performing needed type and version checks. Based on dragonflybsd support for GNU hash by John Marino <draco marino st> Reviewed by: kan Tested by: bapt MFC after: 2 weeks
2012-04-30 13:29:21 +00:00
return (false);
Implement ELF symbol versioning using GNU semantics. This code aims to be compatible with symbol versioning support as implemented by GNU libc and documented by http://people.redhat.com/~drepper/symbol-versioning and LSB 3.0. Implement dlvsym() function to allow lookups for a specific version of a given symbol.
2005-12-18 19:43:33 +00:00
}
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
if (req->name[0] != strp[0] || strcmp(req->name, strp) != 0)
Split the symlook_obj1 into a loop iterating over the ELF object symbol hash elements, and a helper matched_symbol() which match the given hash entry and request, performing needed type and version checks. Based on dragonflybsd support for GNU hash by John Marino <draco marino st> Reviewed by: kan Tested by: bapt MFC after: 2 weeks
2012-04-30 13:29:21 +00:00
return (false);
Style fixes. If it seems like a lot of lines of changes, it's because I moved some functions. Mr. Tidy likes them to be in alphabetical order.
1998-09-02 02:00:20 +00:00
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
if (req->ventry == NULL) {
Split the symlook_obj1 into a loop iterating over the ELF object symbol hash elements, and a helper matched_symbol() which match the given hash entry and request, performing needed type and version checks. Based on dragonflybsd support for GNU hash by John Marino <draco marino st> Reviewed by: kan Tested by: bapt MFC after: 2 weeks
2012-04-30 13:29:21 +00:00
if (obj->versyms != NULL) {
verndx = VER_NDX(obj->versyms[symnum]);
if (verndx > obj->vernum) {
_rtld_error(
"%s: symbol %s references wrong version %d",
obj->path, obj->strtab + symnum, verndx);
return (false);
}
/*
* If we are not called from dlsym (i.e. this
* is a normal relocation from unversioned
* binary), accept the symbol immediately if
* it happens to have first version after this
* shared object became versioned. Otherwise,
* if symbol is versioned and not hidden,
* remember it. If it is the only symbol with
* this name exported by the shared object, it
* will be returned as a match by the calling
* function. If symbol is global (verndx < 2)
* accept it unconditionally.
*/
if ((req->flags & SYMLOOK_DLSYM) == 0 &&
verndx == VER_NDX_GIVEN) {
result->sym_out = symp;
return (true);
}
else if (verndx >= VER_NDX_GIVEN) {
if ((obj->versyms[symnum] & VER_NDX_HIDDEN)
== 0) {
if (result->vsymp == NULL)
result->vsymp = symp;
result->vcount++;
}
return (false);
}
Implement ELF symbol versioning using GNU semantics. This code aims to be compatible with symbol versioning support as implemented by GNU libc and documented by http://people.redhat.com/~drepper/symbol-versioning and LSB 3.0. Implement dlvsym() function to allow lookups for a specific version of a given symbol.
2005-12-18 19:43:33 +00:00
}
Split the symlook_obj1 into a loop iterating over the ELF object symbol hash elements, and a helper matched_symbol() which match the given hash entry and request, performing needed type and version checks. Based on dragonflybsd support for GNU hash by John Marino <draco marino st> Reviewed by: kan Tested by: bapt MFC after: 2 weeks
2012-04-30 13:29:21 +00:00
result->sym_out = symp;
return (true);
}
if (obj->versyms == NULL) {
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
if (object_match_name(obj, req->ventry->name)) {
Split the symlook_obj1 into a loop iterating over the ELF object symbol hash elements, and a helper matched_symbol() which match the given hash entry and request, performing needed type and version checks. Based on dragonflybsd support for GNU hash by John Marino <draco marino st> Reviewed by: kan Tested by: bapt MFC after: 2 weeks
2012-04-30 13:29:21 +00:00
_rtld_error("%s: object %s should provide version %s "
"for symbol %s", obj_rtld.path, obj->path,
req->ventry->name, obj->strtab + symnum);
return (false);
Implement ELF symbol versioning using GNU semantics. This code aims to be compatible with symbol versioning support as implemented by GNU libc and documented by http://people.redhat.com/~drepper/symbol-versioning and LSB 3.0. Implement dlvsym() function to allow lookups for a specific version of a given symbol.
2005-12-18 19:43:33 +00:00
}
Split the symlook_obj1 into a loop iterating over the ELF object symbol hash elements, and a helper matched_symbol() which match the given hash entry and request, performing needed type and version checks. Based on dragonflybsd support for GNU hash by John Marino <draco marino st> Reviewed by: kan Tested by: bapt MFC after: 2 weeks
2012-04-30 13:29:21 +00:00
} else {
Implement ELF symbol versioning using GNU semantics. This code aims to be compatible with symbol versioning support as implemented by GNU libc and documented by http://people.redhat.com/~drepper/symbol-versioning and LSB 3.0. Implement dlvsym() function to allow lookups for a specific version of a given symbol.
2005-12-18 19:43:33 +00:00
verndx = VER_NDX(obj->versyms[symnum]);
if (verndx > obj->vernum) {
Split the symlook_obj1 into a loop iterating over the ELF object symbol hash elements, and a helper matched_symbol() which match the given hash entry and request, performing needed type and version checks. Based on dragonflybsd support for GNU hash by John Marino <draco marino st> Reviewed by: kan Tested by: bapt MFC after: 2 weeks
2012-04-30 13:29:21 +00:00
_rtld_error("%s: symbol %s references wrong version %d",
obj->path, obj->strtab + symnum, verndx);
return (false);
Implement ELF symbol versioning using GNU semantics. This code aims to be compatible with symbol versioning support as implemented by GNU libc and documented by http://people.redhat.com/~drepper/symbol-versioning and LSB 3.0. Implement dlvsym() function to allow lookups for a specific version of a given symbol.
2005-12-18 19:43:33 +00:00
}
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
if (obj->vertab[verndx].hash != req->ventry->hash ||
strcmp(obj->vertab[verndx].name, req->ventry->name)) {
Split the symlook_obj1 into a loop iterating over the ELF object symbol hash elements, and a helper matched_symbol() which match the given hash entry and request, performing needed type and version checks. Based on dragonflybsd support for GNU hash by John Marino <draco marino st> Reviewed by: kan Tested by: bapt MFC after: 2 weeks
2012-04-30 13:29:21 +00:00
/*
* Version does not match. Look if this is a
* global symbol and if it is not hidden. If
* global symbol (verndx < 2) is available,
* use it. Do not return symbol if we are
* called by dlvsym, because dlvsym looks for
* a specific version and default one is not
* what dlvsym wants.
*/
if ((req->flags & SYMLOOK_DLSYM) ||
(verndx >= VER_NDX_GIVEN) ||
(obj->versyms[symnum] & VER_NDX_HIDDEN))
return (false);
Implement ELF symbol versioning using GNU semantics. This code aims to be compatible with symbol versioning support as implemented by GNU libc and documented by http://people.redhat.com/~drepper/symbol-versioning and LSB 3.0. Implement dlvsym() function to allow lookups for a specific version of a given symbol.
2005-12-18 19:43:33 +00:00
}
Eliminate all machine-dependent code from the main source body and the Makefile, and move it down into the architecture-specific subdirectories. Eliminate an asm() statement for the i386. Make the dynamic linker work if it is built as an executable instead of as a shared library. See i386/Makefile.inc to find out how to do it. Note, this change is not enabled and it might never be enabled. But it might be useful in the future. Building the dynamic linker as an executable should make it start up faster, because it won't have any relocations. But in practice I suspect the difference is negligible.
1999-04-09 00:28:43 +00:00
}
Split the symlook_obj1 into a loop iterating over the ELF object symbol hash elements, and a helper matched_symbol() which match the given hash entry and request, performing needed type and version checks. Based on dragonflybsd support for GNU hash by John Marino <draco marino st> Reviewed by: kan Tested by: bapt MFC after: 2 weeks
2012-04-30 13:29:21 +00:00
result->sym_out = symp;
return (true);
}
Add GNU hash support for rtld. Based on dragonflybsd support for GNU hash by John Marino <draco marino st> Reviewed by: kan Tested by: bapt MFC after: 2 weeks
2012-04-30 13:31:10 +00:00
/*
* Search for symbol using SysV hash function.
* obj->buckets is known not to be NULL at this point; the test for this was
* performed with the obj->valid_hash_sysv assignment.
*/
Split the symlook_obj1 into a loop iterating over the ELF object symbol hash elements, and a helper matched_symbol() which match the given hash entry and request, performing needed type and version checks. Based on dragonflybsd support for GNU hash by John Marino <draco marino st> Reviewed by: kan Tested by: bapt MFC after: 2 weeks
2012-04-30 13:29:21 +00:00
static int
Add GNU hash support for rtld. Based on dragonflybsd support for GNU hash by John Marino <draco marino st> Reviewed by: kan Tested by: bapt MFC after: 2 weeks
2012-04-30 13:31:10 +00:00
symlook_obj1_sysv(SymLook *req, const Obj_Entry *obj)
Split the symlook_obj1 into a loop iterating over the ELF object symbol hash elements, and a helper matched_symbol() which match the given hash entry and request, performing needed type and version checks. Based on dragonflybsd support for GNU hash by John Marino <draco marino st> Reviewed by: kan Tested by: bapt MFC after: 2 weeks
2012-04-30 13:29:21 +00:00
{
unsigned long symnum;
Sym_Match_Result matchres;
matchres.sym_out = NULL;
matchres.vsymp = NULL;
matchres.vcount = 0;
for (symnum = obj->buckets[req->hash % obj->nbuckets];
symnum != STN_UNDEF; symnum = obj->chains[symnum]) {
if (symnum >= obj->nchains)
return (ESRCH); /* Bad object */
if (matched_symbol(req, obj, &matchres, symnum)) {
req->sym_out = matchres.sym_out;
req->defobj_out = obj;
return (0);
}
}
if (matchres.vcount == 1) {
req->sym_out = matchres.vsymp;
req->defobj_out = obj;
return (0);
}
return (ESRCH);
Style fixes. If it seems like a lot of lines of changes, it's because I moved some functions. Mr. Tidy likes them to be in alphabetical order.
1998-09-02 02:00:20 +00:00
}
Add GNU hash support for rtld. Based on dragonflybsd support for GNU hash by John Marino <draco marino st> Reviewed by: kan Tested by: bapt MFC after: 2 weeks
2012-04-30 13:31:10 +00:00
/* Search for symbol using GNU hash function */
static int
symlook_obj1_gnu(SymLook *req, const Obj_Entry *obj)
{
Elf_Addr bloom_word;
const Elf32_Word *hashval;
Elf32_Word bucket;
Sym_Match_Result matchres;
unsigned int h1, h2;
unsigned long symnum;
matchres.sym_out = NULL;
matchres.vsymp = NULL;
matchres.vcount = 0;
/* Pick right bitmask word from Bloom filter array */
bloom_word = obj->bloom_gnu[(req->hash_gnu / __ELF_WORD_SIZE) &
obj->maskwords_bm_gnu];
/* Calculate modulus word size of gnu hash and its derivative */
h1 = req->hash_gnu & (__ELF_WORD_SIZE - 1);
h2 = ((req->hash_gnu >> obj->shift2_gnu) & (__ELF_WORD_SIZE - 1));
/* Filter out the "definitely not in set" queries */
if (((bloom_word >> h1) & (bloom_word >> h2) & 1) == 0)
return (ESRCH);
/* Locate hash chain and corresponding value element*/
bucket = obj->buckets_gnu[req->hash_gnu % obj->nbuckets_gnu];
if (bucket == 0)
return (ESRCH);
hashval = &obj->chain_zero_gnu[bucket];
do {
if (((*hashval ^ req->hash_gnu) >> 1) == 0) {
symnum = hashval - obj->chain_zero_gnu;
if (matched_symbol(req, obj, &matchres, symnum)) {
req->sym_out = matchres.sym_out;
req->defobj_out = obj;
return (0);
}
}
} while ((*hashval++ & 1) == 0);
if (matchres.vcount == 1) {
req->sym_out = matchres.vsymp;
req->defobj_out = obj;
return (0);
}
return (ESRCH);
}
Style fixes. If it seems like a lot of lines of changes, it's because I moved some functions. Mr. Tidy likes them to be in alphabetical order.
1998-09-02 02:00:20 +00:00
static void
trace_loaded_objects(Obj_Entry *obj)
Add support for ldd.
1998-05-01 08:39:27 +00:00
{
Add support such that if LD_TRACE_LOADED_OBJECTS_ALL is defined to a non-empty string in the environment; we indicate which objects caused each object to be loaded. PR: 30908 Submitted-by: Mike Meyer <mwm@mired.org>
2002-02-17 07:04:32 +00:00
char *fmt1, *fmt2, *fmt, *main_local, *list_containers;
Add support for ldd.
1998-05-01 08:39:27 +00:00
int c;
Use a macro to create the names for the library path names. This will allow later substitution at run time instead of compile time of the environment variable name prefix. Differential Review: https://reviews.freebsd.org/D2718
2015-12-27 23:04:12 +00:00
if ((main_local = getenv(_LD("TRACE_LOADED_OBJECTS_PROGNAME"))) == NULL)
Add support for ldd.
1998-05-01 08:39:27 +00:00
main_local = "";
Use a macro to create the names for the library path names. This will allow later substitution at run time instead of compile time of the environment variable name prefix. Differential Review: https://reviews.freebsd.org/D2718
2015-12-27 23:04:12 +00:00
if ((fmt1 = getenv(_LD("TRACE_LOADED_OBJECTS_FMT1"))) == NULL)
Add support for ldd.
1998-05-01 08:39:27 +00:00
fmt1 = "\t%o => %p (%x)\n";
Use a macro to create the names for the library path names. This will allow later substitution at run time instead of compile time of the environment variable name prefix. Differential Review: https://reviews.freebsd.org/D2718
2015-12-27 23:04:12 +00:00
if ((fmt2 = getenv(_LD("TRACE_LOADED_OBJECTS_FMT2"))) == NULL)
Add support for ldd.
1998-05-01 08:39:27 +00:00
fmt2 = "\t%o (%x)\n";
Use a macro to create the names for the library path names. This will allow later substitution at run time instead of compile time of the environment variable name prefix. Differential Review: https://reviews.freebsd.org/D2718
2015-12-27 23:04:12 +00:00
list_containers = getenv(_LD("TRACE_LOADED_OBJECTS_ALL"));
Add support such that if LD_TRACE_LOADED_OBJECTS_ALL is defined to a non-empty string in the environment; we indicate which objects caused each object to be loaded. PR: 30908 Submitted-by: Mike Meyer <mwm@mired.org>
2002-02-17 07:04:32 +00:00
Remove all remaining uses of TAILQ_FOREACH_FROM() from rtld-elf. Sponsored by: The FreeBSD Foundation MFC after: 1 week
2016-08-12 18:29:11 +00:00
for (; obj != NULL; obj = TAILQ_NEXT(obj, next)) {
Add support for ldd.
1998-05-01 08:39:27 +00:00
Needed_Entry *needed;
char *name, *path;
bool is_lib;
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
if (obj->marker)
continue;
Add support such that if LD_TRACE_LOADED_OBJECTS_ALL is defined to a non-empty string in the environment; we indicate which objects caused each object to be loaded. PR: 30908 Submitted-by: Mike Meyer <mwm@mired.org>
2002-02-17 07:04:32 +00:00
if (list_containers && obj->needed != NULL)
Rtld links with the specially built pic static libc library to get some C runtime services, like printf(). Unfortunately, the multithread-safeness measures in the libc do not work in rtld environment. Rip the kernel printf() implementation and use it in the rtld instead of libc version. This printf does not require any shared global data and thus is mt-safe. Systematically use rtld_printf() and related functions, remove the calls to err(3). Note that stdio is still pulled from libc due to libmap implementaion using fopen(). This is safe but unoptimal, and can be changed later. Reported and tested by: pgj Diagnosed and reviewed by: kan (previous version) Approved by: re (bz)
2011-08-24 20:05:13 +00:00
rtld_printf("%s:\n", obj->path);
Add support for ldd.
1998-05-01 08:39:27 +00:00
for (needed = obj->needed; needed; needed = needed->next) {
Suppress duplicate entries in ldd output.
1998-09-02 02:51:12 +00:00
if (needed->obj != NULL) {
Add support such that if LD_TRACE_LOADED_OBJECTS_ALL is defined to a non-empty string in the environment; we indicate which objects caused each object to be loaded. PR: 30908 Submitted-by: Mike Meyer <mwm@mired.org>
2002-02-17 07:04:32 +00:00
if (needed->obj->traced && !list_containers)
Suppress duplicate entries in ldd output.
1998-09-02 02:51:12 +00:00
continue;
needed->obj->traced = true;
Add support for ldd.
1998-05-01 08:39:27 +00:00
path = needed->obj->path;
Suppress duplicate entries in ldd output.
1998-09-02 02:51:12 +00:00
} else
path = "not found";
name = (char *)obj->strtab + needed->name;
is_lib = strncmp(name, "lib", 3) == 0; /* XXX - bogus */
Add support for ldd.
1998-05-01 08:39:27 +00:00
fmt = is_lib ? fmt1 : fmt2;
while ((c = *fmt++) != '\0') {
switch (c) {
default:
Rtld links with the specially built pic static libc library to get some C runtime services, like printf(). Unfortunately, the multithread-safeness measures in the libc do not work in rtld environment. Rip the kernel printf() implementation and use it in the rtld instead of libc version. This printf does not require any shared global data and thus is mt-safe. Systematically use rtld_printf() and related functions, remove the calls to err(3). Note that stdio is still pulled from libc due to libmap implementaion using fopen(). This is safe but unoptimal, and can be changed later. Reported and tested by: pgj Diagnosed and reviewed by: kan (previous version) Approved by: re (bz)
2011-08-24 20:05:13 +00:00
rtld_putchar(c);
Add support for ldd.
1998-05-01 08:39:27 +00:00
continue;
case '\\':
switch (c = *fmt) {
case '\0':
continue;
case 'n':
Rtld links with the specially built pic static libc library to get some C runtime services, like printf(). Unfortunately, the multithread-safeness measures in the libc do not work in rtld environment. Rip the kernel printf() implementation and use it in the rtld instead of libc version. This printf does not require any shared global data and thus is mt-safe. Systematically use rtld_printf() and related functions, remove the calls to err(3). Note that stdio is still pulled from libc due to libmap implementaion using fopen(). This is safe but unoptimal, and can be changed later. Reported and tested by: pgj Diagnosed and reviewed by: kan (previous version) Approved by: re (bz)
2011-08-24 20:05:13 +00:00
rtld_putchar('\n');
Add support for ldd.
1998-05-01 08:39:27 +00:00
break;
case 't':
Rtld links with the specially built pic static libc library to get some C runtime services, like printf(). Unfortunately, the multithread-safeness measures in the libc do not work in rtld environment. Rip the kernel printf() implementation and use it in the rtld instead of libc version. This printf does not require any shared global data and thus is mt-safe. Systematically use rtld_printf() and related functions, remove the calls to err(3). Note that stdio is still pulled from libc due to libmap implementaion using fopen(). This is safe but unoptimal, and can be changed later. Reported and tested by: pgj Diagnosed and reviewed by: kan (previous version) Approved by: re (bz)
2011-08-24 20:05:13 +00:00
rtld_putchar('\t');
Add support for ldd.
1998-05-01 08:39:27 +00:00
break;
}
break;
case '%':
switch (c = *fmt) {
case '\0':
continue;
case '%':
default:
Rtld links with the specially built pic static libc library to get some C runtime services, like printf(). Unfortunately, the multithread-safeness measures in the libc do not work in rtld environment. Rip the kernel printf() implementation and use it in the rtld instead of libc version. This printf does not require any shared global data and thus is mt-safe. Systematically use rtld_printf() and related functions, remove the calls to err(3). Note that stdio is still pulled from libc due to libmap implementaion using fopen(). This is safe but unoptimal, and can be changed later. Reported and tested by: pgj Diagnosed and reviewed by: kan (previous version) Approved by: re (bz)
2011-08-24 20:05:13 +00:00
rtld_putchar(c);
Add support for ldd.
1998-05-01 08:39:27 +00:00
break;
case 'A':
Rtld links with the specially built pic static libc library to get some C runtime services, like printf(). Unfortunately, the multithread-safeness measures in the libc do not work in rtld environment. Rip the kernel printf() implementation and use it in the rtld instead of libc version. This printf does not require any shared global data and thus is mt-safe. Systematically use rtld_printf() and related functions, remove the calls to err(3). Note that stdio is still pulled from libc due to libmap implementaion using fopen(). This is safe but unoptimal, and can be changed later. Reported and tested by: pgj Diagnosed and reviewed by: kan (previous version) Approved by: re (bz)
2011-08-24 20:05:13 +00:00
rtld_putstr(main_local);
Add support for ldd.
1998-05-01 08:39:27 +00:00
break;
case 'a':
Rtld links with the specially built pic static libc library to get some C runtime services, like printf(). Unfortunately, the multithread-safeness measures in the libc do not work in rtld environment. Rip the kernel printf() implementation and use it in the rtld instead of libc version. This printf does not require any shared global data and thus is mt-safe. Systematically use rtld_printf() and related functions, remove the calls to err(3). Note that stdio is still pulled from libc due to libmap implementaion using fopen(). This is safe but unoptimal, and can be changed later. Reported and tested by: pgj Diagnosed and reviewed by: kan (previous version) Approved by: re (bz)
2011-08-24 20:05:13 +00:00
rtld_putstr(obj_main->path);
Add support for ldd.
1998-05-01 08:39:27 +00:00
break;
case 'o':
Rtld links with the specially built pic static libc library to get some C runtime services, like printf(). Unfortunately, the multithread-safeness measures in the libc do not work in rtld environment. Rip the kernel printf() implementation and use it in the rtld instead of libc version. This printf does not require any shared global data and thus is mt-safe. Systematically use rtld_printf() and related functions, remove the calls to err(3). Note that stdio is still pulled from libc due to libmap implementaion using fopen(). This is safe but unoptimal, and can be changed later. Reported and tested by: pgj Diagnosed and reviewed by: kan (previous version) Approved by: re (bz)
2011-08-24 20:05:13 +00:00
rtld_putstr(name);
Add support for ldd.
1998-05-01 08:39:27 +00:00
break;
#if 0
case 'm':
Rtld links with the specially built pic static libc library to get some C runtime services, like printf(). Unfortunately, the multithread-safeness measures in the libc do not work in rtld environment. Rip the kernel printf() implementation and use it in the rtld instead of libc version. This printf does not require any shared global data and thus is mt-safe. Systematically use rtld_printf() and related functions, remove the calls to err(3). Note that stdio is still pulled from libc due to libmap implementaion using fopen(). This is safe but unoptimal, and can be changed later. Reported and tested by: pgj Diagnosed and reviewed by: kan (previous version) Approved by: re (bz)
2011-08-24 20:05:13 +00:00
rtld_printf("%d", sodp->sod_major);
Add support for ldd.
1998-05-01 08:39:27 +00:00
break;
case 'n':
Rtld links with the specially built pic static libc library to get some C runtime services, like printf(). Unfortunately, the multithread-safeness measures in the libc do not work in rtld environment. Rip the kernel printf() implementation and use it in the rtld instead of libc version. This printf does not require any shared global data and thus is mt-safe. Systematically use rtld_printf() and related functions, remove the calls to err(3). Note that stdio is still pulled from libc due to libmap implementaion using fopen(). This is safe but unoptimal, and can be changed later. Reported and tested by: pgj Diagnosed and reviewed by: kan (previous version) Approved by: re (bz)
2011-08-24 20:05:13 +00:00
rtld_printf("%d", sodp->sod_minor);
Add support for ldd.
1998-05-01 08:39:27 +00:00
break;
#endif
case 'p':
Rtld links with the specially built pic static libc library to get some C runtime services, like printf(). Unfortunately, the multithread-safeness measures in the libc do not work in rtld environment. Rip the kernel printf() implementation and use it in the rtld instead of libc version. This printf does not require any shared global data and thus is mt-safe. Systematically use rtld_printf() and related functions, remove the calls to err(3). Note that stdio is still pulled from libc due to libmap implementaion using fopen(). This is safe but unoptimal, and can be changed later. Reported and tested by: pgj Diagnosed and reviewed by: kan (previous version) Approved by: re (bz)
2011-08-24 20:05:13 +00:00
rtld_putstr(path);
Add support for ldd.
1998-05-01 08:39:27 +00:00
break;
case 'x':
Rtld links with the specially built pic static libc library to get some C runtime services, like printf(). Unfortunately, the multithread-safeness measures in the libc do not work in rtld environment. Rip the kernel printf() implementation and use it in the rtld instead of libc version. This printf does not require any shared global data and thus is mt-safe. Systematically use rtld_printf() and related functions, remove the calls to err(3). Note that stdio is still pulled from libc due to libmap implementaion using fopen(). This is safe but unoptimal, and can be changed later. Reported and tested by: pgj Diagnosed and reviewed by: kan (previous version) Approved by: re (bz)
2011-08-24 20:05:13 +00:00
rtld_printf("%p", needed->obj ? needed->obj->mapbase :
0);
Add support for ldd.
1998-05-01 08:39:27 +00:00
break;
}
break;
}
++fmt;
}
}
}
}
Style fixes. If it seems like a lot of lines of changes, it's because I moved some functions. Mr. Tidy likes them to be in alphabetical order.
1998-09-02 02:00:20 +00:00
Revamp the symbol lookup algorithm to cope better with objects loaded separately by dlopen that have global symbols with identical names. Viewing each dlopened object as a DAG which is linked by its DT_NEEDED entries in the dynamic table, the search order is as follows: * If the referencing object was linked with -Bsymbolic, search it internally. * Search all dlopened DAGs containing the referencing object. * Search all objects loaded at program start up. * Search all objects which were dlopened() using the RTLD_GLOBAL flag (which is now supported too). The search terminates as soon as a strong definition is found. Lacking that, the first weak definition is used. These rules match those of Solaris, as best I could determine them from its vague manual pages and the results of experiments I performed. PR: misc/12438
1999-08-30 01:48:19 +00:00
/*
Revamp the mechanism for enumerating and calling shared objects' init and fini functions. Now the code is very careful to hold no locks when calling these functions. Thus the dynamic linker cannot be re-entered with a lock already held. Remove the tolerance for recursive locking that I added in revision 1.2 of dllockinit.c. Recursive locking shouldn't happen any more. Mozilla and JDK users: I'd appreciate confirmation that things still work right (or at least the same) with these changes.
2000-01-09 21:13:48 +00:00
* Unload a dlopened object and its dependencies from memory and from
* our data structures. It is assumed that the DAG rooted in the
* object has already been unreferenced, and that the object has a
* reference count of 0.
Revamp the symbol lookup algorithm to cope better with objects loaded separately by dlopen that have global symbols with identical names. Viewing each dlopened object as a DAG which is linked by its DT_NEEDED entries in the dynamic table, the search order is as follows: * If the referencing object was linked with -Bsymbolic, search it internally. * Search all dlopened DAGs containing the referencing object. * Search all objects loaded at program start up. * Search all objects which were dlopened() using the RTLD_GLOBAL flag (which is now supported too). The search terminates as soon as a strong definition is found. Lacking that, the first weak definition is used. These rules match those of Solaris, as best I could determine them from its vague manual pages and the results of experiments I performed. PR: misc/12438
1999-08-30 01:48:19 +00:00
*/
Fix bug: if a dlopen() failed (e.g., because of undefined symbols), the dynamic linker didn't clean up properly. A subsequent dlopen() of the same object would appear to succeed. Another excellent fix from Max Khon. PR: bin/12471 Submitted by: Max Khon <fjoe@iclub.nsu.ru>
1999-07-09 16:22:55 +00:00
static void
Avoid bind lock recursion. When dlclose(3) unloads an object with filtees, it recursively calls dlclose(3) on each filtee in free_needed_filtees(). Introduce dlclose_locked() helper, called from free_needed_filtees() instead of dlclose(), and pass the bind lockstate down to avoid recursing. Reported and tested by: jhibbits Sponsored by: The FreeBSD Foundation MFC after: 1 week
2017-03-09 21:05:47 +00:00
unload_object(Obj_Entry *root, RtldLockState *lockstate)
Fix bug: if a dlopen() failed (e.g., because of undefined symbols), the dynamic linker didn't clean up properly. A subsequent dlopen() of the same object would appear to succeed. Another excellent fix from Max Khon. PR: bin/12471 Submitted by: Max Khon <fjoe@iclub.nsu.ru>
1999-07-09 16:22:55 +00:00
{
rtld: Fix a couple of bugs around the unloading of ELF filters. - Pass the correct object to unload_filtees(). - Use a marker to restart iteration after unload_filtees() has returned. It calls dlclose() and may recursively remove entries from the global object list, so TAILQ_FOREACH_SAFE is not sufficient. Reviewed by: kib MFC after: 2 weeks Sponsored by: Dell EMC Isilon
2016-12-22 17:44:27 +00:00
Obj_Entry marker, *obj, *next;
Revamp the mechanism for enumerating and calling shared objects' init and fini functions. Now the code is very careful to hold no locks when calling these functions. Thus the dynamic linker cannot be re-entered with a lock already held. Remove the tolerance for recursive locking that I added in revision 1.2 of dllockinit.c. Recursive locking shouldn't happen any more. Mozilla and JDK users: I'd appreciate confirmation that things still work right (or at least the same) with these changes.
2000-01-09 21:13:48 +00:00
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
assert(root->refcount == 0);
Revamp the mechanism for enumerating and calling shared objects' init and fini functions. Now the code is very careful to hold no locks when calling these functions. Thus the dynamic linker cannot be re-entered with a lock already held. Remove the tolerance for recursive locking that I added in revision 1.2 of dllockinit.c. Recursive locking shouldn't happen any more. Mozilla and JDK users: I'd appreciate confirmation that things still work right (or at least the same) with these changes.
2000-01-09 21:13:48 +00:00
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
/*
* Pass over the DAG removing unreferenced objects from
* appropriate lists.
*/
unlink_object(root);
/* Unmap all objects that are no longer referenced. */
rtld: Fix a couple of bugs around the unloading of ELF filters. - Pass the correct object to unload_filtees(). - Use a marker to restart iteration after unload_filtees() has returned. It calls dlclose() and may recursively remove entries from the global object list, so TAILQ_FOREACH_SAFE is not sufficient. Reviewed by: kib MFC after: 2 weeks Sponsored by: Dell EMC Isilon
2016-12-22 17:44:27 +00:00
for (obj = TAILQ_FIRST(&obj_list); obj != NULL; obj = next) {
next = TAILQ_NEXT(obj, next);
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
if (obj->marker || obj->refcount != 0)
continue;
LD_UTRACE(UTRACE_UNLOAD_OBJECT, obj, obj->mapbase,
obj->mapsize, 0, obj->path);
dbg("unloading \"%s\"", obj->path);
rtld: Fix a race between dl_iterate_phdr() and dlclose(). Add a transient reference count to ensure that the phdr argument to the callback remains valid while the bind lock is dropped. Reviewed by: kib MFC after: 2 weeks Sponsored by: Dell EMC Isilon
2016-12-22 17:37:39 +00:00
/*
* Unlink the object now to prevent new references from
* being acquired while the bind lock is dropped in
* recursive dlclose() invocations.
*/
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
TAILQ_REMOVE(&obj_list, obj, next);
obj_count--;
rtld: Fix a race between dl_iterate_phdr() and dlclose(). Add a transient reference count to ensure that the phdr argument to the callback remains valid while the bind lock is dropped. Reviewed by: kib MFC after: 2 weeks Sponsored by: Dell EMC Isilon
2016-12-22 17:37:39 +00:00
rtld: Fix a couple of bugs around the unloading of ELF filters. - Pass the correct object to unload_filtees(). - Use a marker to restart iteration after unload_filtees() has returned. It calls dlclose() and may recursively remove entries from the global object list, so TAILQ_FOREACH_SAFE is not sufficient. Reviewed by: kib MFC after: 2 weeks Sponsored by: Dell EMC Isilon
2016-12-22 17:44:27 +00:00
if (obj->filtees_loaded) {
if (next != NULL) {
init_marker(&marker);
TAILQ_INSERT_BEFORE(next, &marker, next);
Avoid bind lock recursion. When dlclose(3) unloads an object with filtees, it recursively calls dlclose(3) on each filtee in free_needed_filtees(). Introduce dlclose_locked() helper, called from free_needed_filtees() instead of dlclose(), and pass the bind lockstate down to avoid recursing. Reported and tested by: jhibbits Sponsored by: The FreeBSD Foundation MFC after: 1 week
2017-03-09 21:05:47 +00:00
unload_filtees(obj, lockstate);
rtld: Fix a couple of bugs around the unloading of ELF filters. - Pass the correct object to unload_filtees(). - Use a marker to restart iteration after unload_filtees() has returned. It calls dlclose() and may recursively remove entries from the global object list, so TAILQ_FOREACH_SAFE is not sufficient. Reviewed by: kib MFC after: 2 weeks Sponsored by: Dell EMC Isilon
2016-12-22 17:44:27 +00:00
next = TAILQ_NEXT(&marker, next);
TAILQ_REMOVE(&obj_list, &marker, next);
} else
Avoid bind lock recursion. When dlclose(3) unloads an object with filtees, it recursively calls dlclose(3) on each filtee in free_needed_filtees(). Introduce dlclose_locked() helper, called from free_needed_filtees() instead of dlclose(), and pass the bind lockstate down to avoid recursing. Reported and tested by: jhibbits Sponsored by: The FreeBSD Foundation MFC after: 1 week
2017-03-09 21:05:47 +00:00
unload_filtees(obj, lockstate);
rtld: Fix a couple of bugs around the unloading of ELF filters. - Pass the correct object to unload_filtees(). - Use a marker to restart iteration after unload_filtees() has returned. It calls dlclose() and may recursively remove entries from the global object list, so TAILQ_FOREACH_SAFE is not sufficient. Reviewed by: kib MFC after: 2 weeks Sponsored by: Dell EMC Isilon
2016-12-22 17:44:27 +00:00
}
rtld: Fix a race between dl_iterate_phdr() and dlclose(). Add a transient reference count to ensure that the phdr argument to the callback remains valid while the bind lock is dropped. Reviewed by: kib MFC after: 2 weeks Sponsored by: Dell EMC Isilon
2016-12-22 17:37:39 +00:00
release_object(obj);
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
}
Fix bug: if a dlopen() failed (e.g., because of undefined symbols), the dynamic linker didn't clean up properly. A subsequent dlopen() of the same object would appear to succeed. Another excellent fix from Max Khon. PR: bin/12471 Submitted by: Max Khon <fjoe@iclub.nsu.ru>
1999-07-09 16:22:55 +00:00
}
Style fixes. If it seems like a lot of lines of changes, it's because I moved some functions. Mr. Tidy likes them to be in alphabetical order.
1998-09-02 02:00:20 +00:00
static void
Do not remove object from the lists at the unref_dag() stage. Introduce a new unlink_object() function and call it in unload_object() instead. Removing the object in unref_dag() is too early, rtld calls _fini() function after that and shared objects might fail resolve their own symbols.
2003-02-17 20:58:27 +00:00
unlink_object(Obj_Entry *root)
Style fixes. If it seems like a lot of lines of changes, it's because I moved some functions. Mr. Tidy likes them to be in alphabetical order.
1998-09-02 02:00:20 +00:00
{
When unloading dependencies make sure they are removed from all the associated lists: remove RTLD_GLOBAL objects from global objects list; remove the parent object from dldags list of its children. Previosly we were doing that only to the top-level object OF the DAG being unloaded and all its dependencies were ignored, leading to mysterious crashes later. Submitted by: peter (partially)
2003-02-10 23:15:07 +00:00
Objlist_Entry *elm;
Add a NULL pointer check whose absence could cause segmentation violations in certain obscure cases involving failed dlopens. Many thanks to Archie Cobbs for providing me with a good test case. Eliminate a block that existed only to localize a declaration.
1999-08-20 22:33:44 +00:00
When unloading dependencies make sure they are removed from all the associated lists: remove RTLD_GLOBAL objects from global objects list; remove the parent object from dldags list of its children. Previosly we were doing that only to the top-level object OF the DAG being unloaded and all its dependencies were ignored, leading to mysterious crashes later. Submitted by: peter (partially)
2003-02-10 23:15:07 +00:00
if (root->refcount == 0) {
/* Remove the object from the RTLD_GLOBAL list. */
objlist_remove(&list_global, root);
/* Remove the object from all objects' DAG lists. */
Implement ELF symbol versioning using GNU semantics. This code aims to be compatible with symbol versioning support as implemented by GNU libc and documented by http://people.redhat.com/~drepper/symbol-versioning and LSB 3.0. Implement dlvsym() function to allow lookups for a specific version of a given symbol.
2005-12-18 19:43:33 +00:00
STAILQ_FOREACH(elm, &root->dagmembers, link) {
When unloading dependencies make sure they are removed from all the associated lists: remove RTLD_GLOBAL objects from global objects list; remove the parent object from dldags list of its children. Previosly we were doing that only to the top-level object OF the DAG being unloaded and all its dependencies were ignored, leading to mysterious crashes later. Submitted by: peter (partially)
2003-02-10 23:15:07 +00:00
objlist_remove(&elm->obj->dldags, root);
Rethink the way we count module references. Simply following DT_NEEDED links is not flexible enough for cases where dynamically loaded modules form a dependency cycle. This should fix an infinite recursion problem encountered by Yahoo. Approved by: re (jhb)
2003-05-08 01:31:36 +00:00
if (elm->obj != root)
unlink_object(elm->obj);
}
When unloading dependencies make sure they are removed from all the associated lists: remove RTLD_GLOBAL objects from global objects list; remove the parent object from dldags list of its children. Previosly we were doing that only to the top-level object OF the DAG being unloaded and all its dependencies were ignored, leading to mysterious crashes later. Submitted by: peter (partially)
2003-02-10 23:15:07 +00:00
}
Rethink the way we count module references. Simply following DT_NEEDED links is not flexible enough for cases where dynamically loaded modules form a dependency cycle. This should fix an infinite recursion problem encountered by Yahoo. Approved by: re (jhb)
2003-05-08 01:31:36 +00:00
}
Do not remove object from the lists at the unref_dag() stage. Introduce a new unlink_object() function and call it in unload_object() instead. Removing the object in unref_dag() is too early, rtld calls _fini() function after that and shared objects might fail resolve their own symbols.
2003-02-17 20:58:27 +00:00
Rethink the way we count module references. Simply following DT_NEEDED links is not flexible enough for cases where dynamically loaded modules form a dependency cycle. This should fix an infinite recursion problem encountered by Yahoo. Approved by: re (jhb)
2003-05-08 01:31:36 +00:00
static void
ref_dag(Obj_Entry *root)
{
Objlist_Entry *elm;
In r214728, if dlopen() is called for the object that has been already loaded as a dependency and marked -z nodlopen, object' DAG is already initialized by load_needed_objects(). Due to this, the init_dag() call from dlopen() does not increment refcount for the object [1]. Change init_dag() to not increment DAG refcount. Require explicit calls to ref_dag() to increment, and assert that ref_dag() and unref_dag() are called for root that has dag initialized. To fix the noted issue, unconditionally call both init_dag() and ref_dag() in dlopen() for the case when the object was already loaded, making it similar to the case of newly loaded object. Noted by: jh [1] Reviewed by: jh, kan MFC after: 6 days
2010-11-04 09:29:00 +00:00
assert(root->dag_inited);
Implement ELF symbol versioning using GNU semantics. This code aims to be compatible with symbol versioning support as implemented by GNU libc and documented by http://people.redhat.com/~drepper/symbol-versioning and LSB 3.0. Implement dlvsym() function to allow lookups for a specific version of a given symbol.
2005-12-18 19:43:33 +00:00
STAILQ_FOREACH(elm, &root->dagmembers, link)
Rethink the way we count module references. Simply following DT_NEEDED links is not flexible enough for cases where dynamically loaded modules form a dependency cycle. This should fix an infinite recursion problem encountered by Yahoo. Approved by: re (jhb)
2003-05-08 01:31:36 +00:00
elm->obj->refcount++;
Do not remove object from the lists at the unref_dag() stage. Introduce a new unlink_object() function and call it in unload_object() instead. Removing the object in unref_dag() is too early, rtld calls _fini() function after that and shared objects might fail resolve their own symbols.
2003-02-17 20:58:27 +00:00
}
static void
unref_dag(Obj_Entry *root)
{
Rethink the way we count module references. Simply following DT_NEEDED links is not flexible enough for cases where dynamically loaded modules form a dependency cycle. This should fix an infinite recursion problem encountered by Yahoo. Approved by: re (jhb)
2003-05-08 01:31:36 +00:00
Objlist_Entry *elm;
Do not remove object from the lists at the unref_dag() stage. Introduce a new unlink_object() function and call it in unload_object() instead. Removing the object in unref_dag() is too early, rtld calls _fini() function after that and shared objects might fail resolve their own symbols.
2003-02-17 20:58:27 +00:00
In r214728, if dlopen() is called for the object that has been already loaded as a dependency and marked -z nodlopen, object' DAG is already initialized by load_needed_objects(). Due to this, the init_dag() call from dlopen() does not increment refcount for the object [1]. Change init_dag() to not increment DAG refcount. Require explicit calls to ref_dag() to increment, and assert that ref_dag() and unref_dag() are called for root that has dag initialized. To fix the noted issue, unconditionally call both init_dag() and ref_dag() in dlopen() for the case when the object was already loaded, making it similar to the case of newly loaded object. Noted by: jh [1] Reviewed by: jh, kan MFC after: 6 days
2010-11-04 09:29:00 +00:00
assert(root->dag_inited);
Implement ELF symbol versioning using GNU semantics. This code aims to be compatible with symbol versioning support as implemented by GNU libc and documented by http://people.redhat.com/~drepper/symbol-versioning and LSB 3.0. Implement dlvsym() function to allow lookups for a specific version of a given symbol.
2005-12-18 19:43:33 +00:00
STAILQ_FOREACH(elm, &root->dagmembers, link)
Rethink the way we count module references. Simply following DT_NEEDED links is not flexible enough for cases where dynamically loaded modules form a dependency cycle. This should fix an infinite recursion problem encountered by Yahoo. Approved by: re (jhb)
2003-05-08 01:31:36 +00:00
elm->obj->refcount--;
Style fixes. If it seems like a lot of lines of changes, it's because I moved some functions. Mr. Tidy likes them to be in alphabetical order.
1998-09-02 02:00:20 +00:00
}
Allow threading libraries to register their own locking implementation in case default one provided by rtld is not suitable. Consolidate various identical MD lock implementation into a single file using appropriate machine/atomic.h. Approved by: re (scottl)
2003-05-29 22:58:26 +00:00
Add support for Thread Local Storage.
2004-08-03 08:51:00 +00:00
/*
* Common code for MD __tls_get_addr().
*/
Optimize tls_get_addr_common(). The change provides around 30% speedup for TLS microbenchmark using global-dynamic TLS model on amd64 (which is default for PIC dso objects). Split the slow path into tls_get_addr_slow(), for which inlining is disabled. This prevents the registers spill on tls_get_addr_common() entry. Provide static branch hint to the compiler, indicating that slow path is not likely to be taken. While there, do some minimal style adjustments. Reported and tested by: davidxu MFC after: 1 week
2012-03-10 08:49:44 +00:00
static void *tls_get_addr_slow(Elf_Addr **, int, size_t) __noinline;
static void *
tls_get_addr_slow(Elf_Addr **dtvp, int index, size_t offset)
Add support for Thread Local Storage.
2004-08-03 08:51:00 +00:00
{
Optimize tls_get_addr_common(). The change provides around 30% speedup for TLS microbenchmark using global-dynamic TLS model on amd64 (which is default for PIC dso objects). Split the slow path into tls_get_addr_slow(), for which inlining is disabled. This prevents the registers spill on tls_get_addr_common() entry. Provide static branch hint to the compiler, indicating that slow path is not likely to be taken. While there, do some minimal style adjustments. Reported and tested by: davidxu MFC after: 1 week
2012-03-10 08:49:44 +00:00
Elf_Addr *newdtv, *dtv;
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
RtldLockState lockstate;
Optimize tls_get_addr_common(). The change provides around 30% speedup for TLS microbenchmark using global-dynamic TLS model on amd64 (which is default for PIC dso objects). Split the slow path into tls_get_addr_slow(), for which inlining is disabled. This prevents the registers spill on tls_get_addr_common() entry. Provide static branch hint to the compiler, indicating that slow path is not likely to be taken. While there, do some minimal style adjustments. Reported and tested by: davidxu MFC after: 1 week
2012-03-10 08:49:44 +00:00
int to_copy;
Add support for Thread Local Storage.
2004-08-03 08:51:00 +00:00
Optimize tls_get_addr_common(). The change provides around 30% speedup for TLS microbenchmark using global-dynamic TLS model on amd64 (which is default for PIC dso objects). Split the slow path into tls_get_addr_slow(), for which inlining is disabled. This prevents the registers spill on tls_get_addr_common() entry. Provide static branch hint to the compiler, indicating that slow path is not likely to be taken. While there, do some minimal style adjustments. Reported and tested by: davidxu MFC after: 1 week
2012-03-10 08:49:44 +00:00
dtv = *dtvp;
Add support for Thread Local Storage.
2004-08-03 08:51:00 +00:00
/* Check dtv generation in case new modules have arrived */
if (dtv[0] != tls_dtv_generation) {
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
wlock_acquire(rtld_bind_lock, &lockstate);
Use xmalloc() instead of malloc() in the places where malloc() calls are assumed to not fail. Make the xcalloc() calling conventions follow the calloc(3) calling conventions and replace unchecked calls to calloc() with calls to xcalloc(). Remove redundand declarations from xmalloc.c, which are already present in rtld.h. Reviewed by: kan Discussed with: bde MFC after: 2 weeks
2012-03-22 14:20:51 +00:00
newdtv = xcalloc(tls_max_index + 2, sizeof(Elf_Addr));
Add support for Thread Local Storage.
2004-08-03 08:51:00 +00:00
to_copy = dtv[1];
if (to_copy > tls_max_index)
to_copy = tls_max_index;
memcpy(&newdtv[2], &dtv[2], to_copy * sizeof(Elf_Addr));
newdtv[0] = tls_dtv_generation;
newdtv[1] = tls_max_index;
free(dtv);
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
lock_release(rtld_bind_lock, &lockstate);
Use the proper dynamic tls block to calculate the tls variable address in case tls data generation was updated. PR: misc/160721 Submitted by: "Thinker K.F. Li" <thinker codemud net> Tested by: flo Approved by: re (bz) MFC after: 1 week
2011-09-15 11:50:09 +00:00
dtv = *dtvp = newdtv;
Add support for Thread Local Storage.
2004-08-03 08:51:00 +00:00
}
/* Dynamically allocate module TLS if necessary */
Optimize tls_get_addr_common(). The change provides around 30% speedup for TLS microbenchmark using global-dynamic TLS model on amd64 (which is default for PIC dso objects). Split the slow path into tls_get_addr_slow(), for which inlining is disabled. This prevents the registers spill on tls_get_addr_common() entry. Provide static branch hint to the compiler, indicating that slow path is not likely to be taken. While there, do some minimal style adjustments. Reported and tested by: davidxu MFC after: 1 week
2012-03-10 08:49:44 +00:00
if (dtv[index + 1] == 0) {
Add locking code for tls routines.
2005-03-20 23:28:25 +00:00
/* Signal safe, wlock will block out signals. */
Optimize tls_get_addr_common(). The change provides around 30% speedup for TLS microbenchmark using global-dynamic TLS model on amd64 (which is default for PIC dso objects). Split the slow path into tls_get_addr_slow(), for which inlining is disabled. This prevents the registers spill on tls_get_addr_common() entry. Provide static branch hint to the compiler, indicating that slow path is not likely to be taken. While there, do some minimal style adjustments. Reported and tested by: davidxu MFC after: 1 week
2012-03-10 08:49:44 +00:00
wlock_acquire(rtld_bind_lock, &lockstate);
Add locking code for tls routines.
2005-03-20 23:28:25 +00:00
if (!dtv[index + 1])
dtv[index + 1] = (Elf_Addr)allocate_module_tls(index);
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
lock_release(rtld_bind_lock, &lockstate);
Add locking code for tls routines.
2005-03-20 23:28:25 +00:00
}
Optimize tls_get_addr_common(). The change provides around 30% speedup for TLS microbenchmark using global-dynamic TLS model on amd64 (which is default for PIC dso objects). Split the slow path into tls_get_addr_slow(), for which inlining is disabled. This prevents the registers spill on tls_get_addr_common() entry. Provide static branch hint to the compiler, indicating that slow path is not likely to be taken. While there, do some minimal style adjustments. Reported and tested by: davidxu MFC after: 1 week
2012-03-10 08:49:44 +00:00
return ((void *)(dtv[index + 1] + offset));
}
void *
tls_get_addr_common(Elf_Addr **dtvp, int index, size_t offset)
{
Elf_Addr *dtv;
dtv = *dtvp;
/* Check dtv generation in case new modules have arrived */
if (__predict_true(dtv[0] == tls_dtv_generation &&
dtv[index + 1] != 0))
return ((void *)(dtv[index + 1] + offset));
return (tls_get_addr_slow(dtvp, index, offset));
Add support for Thread Local Storage.
2004-08-03 08:51:00 +00:00
}
Add the arm64 code to the runtime linker. It's not able to be built as we still need libc_pic for a few things, but this is expected to be ready soon. Differential Revision: https://reviews.freebsd.org/D2136 Reviewed by: kib Sponsored by: The FreeBSD Foundation
2015-03-31 09:51:19 +00:00
#if defined(__aarch64__) || defined(__arm__) || defined(__mips__) || \
Add support for RISC-V architecture. Reviewed by: andrew, kib Sponsored by: DARPA, AFRL Sponsored by: HEIF5 Differential Revision: https://reviews.freebsd.org/D4679
2015-12-24 15:47:51 +00:00
defined(__powerpc__) || defined(__riscv__)
Add support for Thread Local Storage.
2004-08-03 08:51:00 +00:00
/*
* Allocate Static TLS using the Variant I method.
*/
void *
Fix the variant I allocation for KSE: Allow a larger TCB and assume that the documented TCB is at the tail of the extended TCB. In other words, the base of the TCB has a negative offset from the TLS.
2006-09-01 06:13:16 +00:00
allocate_tls(Obj_Entry *objs, void *oldtcb, size_t tcbsize, size_t tcbalign)
Add support for Thread Local Storage.
2004-08-03 08:51:00 +00:00
{
Obj_Entry *obj;
Fix the variant I allocation for KSE: Allow a larger TCB and assume that the documented TCB is at the tail of the extended TCB. In other words, the base of the TCB has a negative offset from the TLS.
2006-09-01 06:13:16 +00:00
char *tcb;
Elf_Addr **tls;
Elf_Addr *dtv;
Add support for Thread Local Storage.
2004-08-03 08:51:00 +00:00
Elf_Addr addr;
int i;
Fix the variant I allocation for KSE: Allow a larger TCB and assume that the documented TCB is at the tail of the extended TCB. In other words, the base of the TCB has a negative offset from the TLS.
2006-09-01 06:13:16 +00:00
if (oldtcb != NULL && tcbsize == TLS_TCB_SIZE)
return (oldtcb);
Add support for Thread Local Storage.
2004-08-03 08:51:00 +00:00
Fix the variant I allocation for KSE: Allow a larger TCB and assume that the documented TCB is at the tail of the extended TCB. In other words, the base of the TCB has a negative offset from the TLS.
2006-09-01 06:13:16 +00:00
assert(tcbsize >= TLS_TCB_SIZE);
Use xmalloc() instead of malloc() in the places where malloc() calls are assumed to not fail. Make the xcalloc() calling conventions follow the calloc(3) calling conventions and replace unchecked calls to calloc() with calls to xcalloc(). Remove redundand declarations from xmalloc.c, which are already present in rtld.h. Reviewed by: kan Discussed with: bde MFC after: 2 weeks
2012-03-22 14:20:51 +00:00
tcb = xcalloc(1, tls_static_space - TLS_TCB_SIZE + tcbsize);
Fix the variant I allocation for KSE: Allow a larger TCB and assume that the documented TCB is at the tail of the extended TCB. In other words, the base of the TCB has a negative offset from the TLS.
2006-09-01 06:13:16 +00:00
tls = (Elf_Addr **)(tcb + tcbsize - TLS_TCB_SIZE);
Add support for Thread Local Storage.
2004-08-03 08:51:00 +00:00
Fix the variant I allocation for KSE: Allow a larger TCB and assume that the documented TCB is at the tail of the extended TCB. In other words, the base of the TCB has a negative offset from the TLS.
2006-09-01 06:13:16 +00:00
if (oldtcb != NULL) {
memcpy(tls, oldtcb, tls_static_space);
free(oldtcb);
Add support for Thread Local Storage.
2004-08-03 08:51:00 +00:00
Fix the variant I allocation for KSE: Allow a larger TCB and assume that the documented TCB is at the tail of the extended TCB. In other words, the base of the TCB has a negative offset from the TLS.
2006-09-01 06:13:16 +00:00
/* Adjust the DTV. */
dtv = tls[0];
for (i = 0; i < dtv[1]; i++) {
if (dtv[i+2] >= (Elf_Addr)oldtcb &&
dtv[i+2] < (Elf_Addr)oldtcb + tls_static_space) {
dtv[i+2] = dtv[i+2] - (Elf_Addr)oldtcb + (Elf_Addr)tls;
Add support for Thread Local Storage.
2004-08-03 08:51:00 +00:00
}
}
} else {
Use xmalloc() instead of malloc() in the places where malloc() calls are assumed to not fail. Make the xcalloc() calling conventions follow the calloc(3) calling conventions and replace unchecked calls to calloc() with calls to xcalloc(). Remove redundand declarations from xmalloc.c, which are already present in rtld.h. Reviewed by: kan Discussed with: bde MFC after: 2 weeks
2012-03-22 14:20:51 +00:00
dtv = xcalloc(tls_max_index + 2, sizeof(Elf_Addr));
Fix the variant I allocation for KSE: Allow a larger TCB and assume that the documented TCB is at the tail of the extended TCB. In other words, the base of the TCB has a negative offset from the TLS.
2006-09-01 06:13:16 +00:00
tls[0] = dtv;
dtv[0] = tls_dtv_generation;
dtv[1] = tls_max_index;
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
for (obj = globallist_curr(objs); obj != NULL;
obj = globallist_next(obj)) {
Improve TLS variant I: o Use obj->tlsinitsize to determine whether there's initialized data. o If obj->tlssize > obj->tlsinitsize, then bzero uninitialized data. o Don't exclude variant I from the work-around in free_tls_offset().
2010-02-16 02:48:11 +00:00
if (obj->tlsoffset > 0) {
Add support for Thread Local Storage.
2004-08-03 08:51:00 +00:00
addr = (Elf_Addr)tls + obj->tlsoffset;
Improve TLS variant I: o Use obj->tlsinitsize to determine whether there's initialized data. o If obj->tlssize > obj->tlsinitsize, then bzero uninitialized data. o Don't exclude variant I from the work-around in free_tls_offset().
2010-02-16 02:48:11 +00:00
if (obj->tlsinitsize > 0)
memcpy((void*) addr, obj->tlsinit, obj->tlsinitsize);
if (obj->tlssize > obj->tlsinitsize)
memset((void*) (addr + obj->tlsinitsize), 0,
obj->tlssize - obj->tlsinitsize);
Add support for Thread Local Storage.
2004-08-03 08:51:00 +00:00
dtv[obj->tlsindex + 1] = addr;
}
}
}
Fix the variant I allocation for KSE: Allow a larger TCB and assume that the documented TCB is at the tail of the extended TCB. In other words, the base of the TCB has a negative offset from the TLS.
2006-09-01 06:13:16 +00:00
return (tcb);
Add support for Thread Local Storage.
2004-08-03 08:51:00 +00:00
}
void
Fix the variant I allocation for KSE: Allow a larger TCB and assume that the documented TCB is at the tail of the extended TCB. In other words, the base of the TCB has a negative offset from the TLS.
2006-09-01 06:13:16 +00:00
free_tls(void *tcb, size_t tcbsize, size_t tcbalign)
Add support for Thread Local Storage.
2004-08-03 08:51:00 +00:00
{
Fix the variant I allocation for KSE: Allow a larger TCB and assume that the documented TCB is at the tail of the extended TCB. In other words, the base of the TCB has a negative offset from the TLS.
2006-09-01 06:13:16 +00:00
Elf_Addr *dtv;
Add support for Thread Local Storage.
2004-08-03 08:51:00 +00:00
Elf_Addr tlsstart, tlsend;
Fix the variant I allocation for KSE: Allow a larger TCB and assume that the documented TCB is at the tail of the extended TCB. In other words, the base of the TCB has a negative offset from the TLS.
2006-09-01 06:13:16 +00:00
int dtvsize, i;
Add support for Thread Local Storage.
2004-08-03 08:51:00 +00:00
Fix the variant I allocation for KSE: Allow a larger TCB and assume that the documented TCB is at the tail of the extended TCB. In other words, the base of the TCB has a negative offset from the TLS.
2006-09-01 06:13:16 +00:00
assert(tcbsize >= TLS_TCB_SIZE);
Add support for Thread Local Storage.
2004-08-03 08:51:00 +00:00
Fix the variant I allocation for KSE: Allow a larger TCB and assume that the documented TCB is at the tail of the extended TCB. In other words, the base of the TCB has a negative offset from the TLS.
2006-09-01 06:13:16 +00:00
tlsstart = (Elf_Addr)tcb + tcbsize - TLS_TCB_SIZE;
tlsend = tlsstart + tls_static_space;
dtv = *(Elf_Addr **)tlsstart;
Add support for Thread Local Storage.
2004-08-03 08:51:00 +00:00
dtvsize = dtv[1];
for (i = 0; i < dtvsize; i++) {
Fix the variant I allocation for KSE: Allow a larger TCB and assume that the documented TCB is at the tail of the extended TCB. In other words, the base of the TCB has a negative offset from the TLS.
2006-09-01 06:13:16 +00:00
if (dtv[i+2] && (dtv[i+2] < tlsstart || dtv[i+2] >= tlsend)) {
free((void*)dtv[i+2]);
Add support for Thread Local Storage.
2004-08-03 08:51:00 +00:00
}
}
Fix the variant I allocation for KSE: Allow a larger TCB and assume that the documented TCB is at the tail of the extended TCB. In other words, the base of the TCB has a negative offset from the TLS.
2006-09-01 06:13:16 +00:00
free(dtv);
free(tcb);
Add support for Thread Local Storage.
2004-08-03 08:51:00 +00:00
}
#endif
Add thread-local storage support for ARM to rtld-elf Reviewed by: cognet Obtained from: NetBSD
2012-02-14 00:16:34 +00:00
#if defined(__i386__) || defined(__amd64__) || defined(__sparc64__)
Add support for Thread Local Storage.
2004-08-03 08:51:00 +00:00
/*
* Allocate Static TLS using the Variant II method.
*/
void *
allocate_tls(Obj_Entry *objs, void *oldtls, size_t tcbsize, size_t tcbalign)
{
Obj_Entry *obj;
For variant II static TLS, properly align tls segments. Pre-calculate the max required alignment for the static tls segments, and honor it when carving the pieces for next module, from the static space. Use aligned allocator to get properly-aligned dynamic blocks. Reported by: dt71@gmx.com Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 1 week
2013-12-06 21:39:45 +00:00
size_t size, ralign;
Add support for Thread Local Storage.
2004-08-03 08:51:00 +00:00
char *tls;
Elf_Addr *dtv, *olddtv;
Elf_Addr segbase, oldsegbase, addr;
int i;
For variant II static TLS, properly align tls segments. Pre-calculate the max required alignment for the static tls segments, and honor it when carving the pieces for next module, from the static space. Use aligned allocator to get properly-aligned dynamic blocks. Reported by: dt71@gmx.com Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 1 week
2013-12-06 21:39:45 +00:00
ralign = tcbalign;
if (tls_static_max_align > ralign)
ralign = tls_static_max_align;
size = round(tls_static_space, ralign) + round(tcbsize, ralign);
Add support for Thread Local Storage.
2004-08-03 08:51:00 +00:00
assert(tcbsize >= 2*sizeof(Elf_Addr));
For variant II static TLS, properly align tls segments. Pre-calculate the max required alignment for the static tls segments, and honor it when carving the pieces for next module, from the static space. Use aligned allocator to get properly-aligned dynamic blocks. Reported by: dt71@gmx.com Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 1 week
2013-12-06 21:39:45 +00:00
tls = malloc_aligned(size, ralign);
Use xmalloc() instead of malloc() in the places where malloc() calls are assumed to not fail. Make the xcalloc() calling conventions follow the calloc(3) calling conventions and replace unchecked calls to calloc() with calls to xcalloc(). Remove redundand declarations from xmalloc.c, which are already present in rtld.h. Reviewed by: kan Discussed with: bde MFC after: 2 weeks
2012-03-22 14:20:51 +00:00
dtv = xcalloc(tls_max_index + 2, sizeof(Elf_Addr));
Add support for Thread Local Storage.
2004-08-03 08:51:00 +00:00
For variant II static TLS, properly align tls segments. Pre-calculate the max required alignment for the static tls segments, and honor it when carving the pieces for next module, from the static space. Use aligned allocator to get properly-aligned dynamic blocks. Reported by: dt71@gmx.com Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 1 week
2013-12-06 21:39:45 +00:00
segbase = (Elf_Addr)(tls + round(tls_static_space, ralign));
Add support for Thread Local Storage.
2004-08-03 08:51:00 +00:00
((Elf_Addr*)segbase)[0] = segbase;
((Elf_Addr*)segbase)[1] = (Elf_Addr) dtv;
dtv[0] = tls_dtv_generation;
dtv[1] = tls_max_index;
if (oldtls) {
/*
* Copy the static TLS block over whole.
*/
oldsegbase = (Elf_Addr) oldtls;
memcpy((void *)(segbase - tls_static_space),
(const void *)(oldsegbase - tls_static_space),
tls_static_space);
/*
* If any dynamic TLS blocks have been created tls_get_addr(),
* move them over.
*/
olddtv = ((Elf_Addr**)oldsegbase)[1];
for (i = 0; i < olddtv[1]; i++) {
if (olddtv[i+2] < oldsegbase - size || olddtv[i+2] > oldsegbase) {
dtv[i+2] = olddtv[i+2];
olddtv[i+2] = 0;
}
}
/*
* We assume that this block was the one we created with
* allocate_initial_tls().
*/
free_tls(oldtls, 2*sizeof(Elf_Addr), sizeof(Elf_Addr));
} else {
Remove all remaining uses of TAILQ_FOREACH_FROM() from rtld-elf. Sponsored by: The FreeBSD Foundation MFC after: 1 week
2016-08-12 18:29:11 +00:00
for (obj = objs; obj != NULL; obj = TAILQ_NEXT(obj, next)) {
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
if (obj->marker || obj->tlsoffset == 0)
continue;
Add support for Thread Local Storage.
2004-08-03 08:51:00 +00:00
addr = segbase - obj->tlsoffset;
memset((void*) (addr + obj->tlsinitsize),
0, obj->tlssize - obj->tlsinitsize);
if (obj->tlsinit)
memcpy((void*) addr, obj->tlsinit, obj->tlsinitsize);
dtv[obj->tlsindex + 1] = addr;
}
}
return (void*) segbase;
}
void
free_tls(void *tls, size_t tcbsize, size_t tcbalign)
{
Elf_Addr* dtv;
For variant II static TLS, properly align tls segments. Pre-calculate the max required alignment for the static tls segments, and honor it when carving the pieces for next module, from the static space. Use aligned allocator to get properly-aligned dynamic blocks. Reported by: dt71@gmx.com Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 1 week
2013-12-06 21:39:45 +00:00
size_t size, ralign;
Add support for Thread Local Storage.
2004-08-03 08:51:00 +00:00
int dtvsize, i;
Elf_Addr tlsstart, tlsend;
/*
* Figure out the size of the initial TLS block so that we can
* find stuff which ___tls_get_addr() allocated dynamically.
*/
For variant II static TLS, properly align tls segments. Pre-calculate the max required alignment for the static tls segments, and honor it when carving the pieces for next module, from the static space. Use aligned allocator to get properly-aligned dynamic blocks. Reported by: dt71@gmx.com Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 1 week
2013-12-06 21:39:45 +00:00
ralign = tcbalign;
if (tls_static_max_align > ralign)
ralign = tls_static_max_align;
size = round(tls_static_space, ralign);
Add support for Thread Local Storage.
2004-08-03 08:51:00 +00:00
dtv = ((Elf_Addr**)tls)[1];
dtvsize = dtv[1];
tlsend = (Elf_Addr) tls;
tlsstart = tlsend - size;
for (i = 0; i < dtvsize; i++) {
For variant II static TLS, properly align tls segments. Pre-calculate the max required alignment for the static tls segments, and honor it when carving the pieces for next module, from the static space. Use aligned allocator to get properly-aligned dynamic blocks. Reported by: dt71@gmx.com Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 1 week
2013-12-06 21:39:45 +00:00
if (dtv[i + 2] != 0 && (dtv[i + 2] < tlsstart || dtv[i + 2] > tlsend)) {
Cast Elf_Addr to void * to match the free_aligned() argument type. Found by: gcc Sponsored by: The FreeBSD Foundation MFC after: 6 days
2013-12-07 15:49:16 +00:00
free_aligned((void *)dtv[i + 2]);
Add support for Thread Local Storage.
2004-08-03 08:51:00 +00:00
}
}
Cast Elf_Addr to void * to match the free_aligned() argument type. Found by: gcc Sponsored by: The FreeBSD Foundation MFC after: 6 days
2013-12-07 15:49:16 +00:00
free_aligned((void *)tlsstart);
Fix a TLS memory leak. PR: threads/112297 MFC: 1 week
2007-05-05 08:44:59 +00:00
free((void*) dtv);
Add support for Thread Local Storage.
2004-08-03 08:51:00 +00:00
}
#endif
Allow threading libraries to register their own locking implementation in case default one provided by rtld is not suitable. Consolidate various identical MD lock implementation into a single file using appropriate machine/atomic.h. Approved by: re (scottl)
2003-05-29 22:58:26 +00:00
Add support for Thread Local Storage.
2004-08-03 08:51:00 +00:00
/*
* Allocate TLS block for module with given index.
*/
void *
allocate_module_tls(int index)
{
Obj_Entry* obj;
char* p;
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
TAILQ_FOREACH(obj, &obj_list, next) {
if (obj->marker)
continue;
Add support for Thread Local Storage.
2004-08-03 08:51:00 +00:00
if (obj->tlsindex == index)
break;
}
if (!obj) {
_rtld_error("Can't find module with TLS index %d", index);
Make die available as rtld_die for use by MD relocation code Reviewed by: kib Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D2213
2015-04-02 21:35:36 +00:00
rtld_die();
Add support for Thread Local Storage.
2004-08-03 08:51:00 +00:00
}
For variant II static TLS, properly align tls segments. Pre-calculate the max required alignment for the static tls segments, and honor it when carving the pieces for next module, from the static space. Use aligned allocator to get properly-aligned dynamic blocks. Reported by: dt71@gmx.com Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 1 week
2013-12-06 21:39:45 +00:00
p = malloc_aligned(obj->tlssize, obj->tlsalign);
Add support for Thread Local Storage.
2004-08-03 08:51:00 +00:00
memcpy(p, obj->tlsinit, obj->tlsinitsize);
memset(p + obj->tlsinitsize, 0, obj->tlssize - obj->tlsinitsize);
return p;
}
bool
allocate_tls_offset(Obj_Entry *obj)
{
size_t off;
if (obj->tls_done)
return true;
if (obj->tlssize == 0) {
obj->tls_done = true;
return true;
}
Ensure we use calculate_first_tls_offset, even if the main program doesn't have TLS program header. This is needed on architectures with Variant I tls, that is arm, arm64, mips, and powerpc. These place the thread control block at the start of the buffer and, without this, this data may be trashed. This appears to not be an issue on mips or powerpc as they include a second adjustment to move the thread local data, however this is on arm64 (with a future change to fix placing this data), and should be on arm. I am unable to trigger this on arm, even after changing the code to move the data around to make it more likely to be hit. This is most likely because my tests didn't use the variable in offset 0. Reviewed by: kib MFC after: 1 week Sponsored by: ABT Systems Ltd
2015-09-01 15:43:56 +00:00
if (tls_last_offset == 0)
Add support for Thread Local Storage.
2004-08-03 08:51:00 +00:00
off = calculate_first_tls_offset(obj->tlssize, obj->tlsalign);
else
off = calculate_tls_offset(tls_last_offset, tls_last_size,
obj->tlssize, obj->tlsalign);
/*
* If we have already fixed the size of the static TLS block, we
* must stay within that size. When allocating the static TLS, we
* leave a small amount of space spare to be used for dynamically
* loading modules which use static TLS.
*/
For variant II static TLS, properly align tls segments. Pre-calculate the max required alignment for the static tls segments, and honor it when carving the pieces for next module, from the static space. Use aligned allocator to get properly-aligned dynamic blocks. Reported by: dt71@gmx.com Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 1 week
2013-12-06 21:39:45 +00:00
if (tls_static_space != 0) {
Add support for Thread Local Storage.
2004-08-03 08:51:00 +00:00
if (calculate_tls_end(off, obj->tlssize) > tls_static_space)
return false;
For variant II static TLS, properly align tls segments. Pre-calculate the max required alignment for the static tls segments, and honor it when carving the pieces for next module, from the static space. Use aligned allocator to get properly-aligned dynamic blocks. Reported by: dt71@gmx.com Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 1 week
2013-12-06 21:39:45 +00:00
} else if (obj->tlsalign > tls_static_max_align) {
tls_static_max_align = obj->tlsalign;
Add support for Thread Local Storage.
2004-08-03 08:51:00 +00:00
}
tls_last_offset = obj->tlsoffset = off;
tls_last_size = obj->tlssize;
obj->tls_done = true;
return true;
}
Attempt to free any static TLS space used by a shared library when it is unloaded. This allows applications which load and unload libraries like libGL.so.1 several times to work properly. MFC after: 2 days
2005-02-27 12:55:40 +00:00
void
free_tls_offset(Obj_Entry *obj)
{
Improve TLS variant I: o Use obj->tlsinitsize to determine whether there's initialized data. o If obj->tlssize > obj->tlsinitsize, then bzero uninitialized data. o Don't exclude variant I from the work-around in free_tls_offset().
2010-02-16 02:48:11 +00:00
Attempt to free any static TLS space used by a shared library when it is unloaded. This allows applications which load and unload libraries like libGL.so.1 several times to work properly. MFC after: 2 days
2005-02-27 12:55:40 +00:00
/*
* If we were the last thing to allocate out of the static TLS
* block, we give our space back to the 'allocator'. This is a
* simplistic workaround to allow libGL.so.1 to be loaded and
Improve TLS variant I: o Use obj->tlsinitsize to determine whether there's initialized data. o If obj->tlssize > obj->tlsinitsize, then bzero uninitialized data. o Don't exclude variant I from the work-around in free_tls_offset().
2010-02-16 02:48:11 +00:00
* unloaded multiple times.
Attempt to free any static TLS space used by a shared library when it is unloaded. This allows applications which load and unload libraries like libGL.so.1 several times to work properly. MFC after: 2 days
2005-02-27 12:55:40 +00:00
*/
if (calculate_tls_end(obj->tlsoffset, obj->tlssize)
== calculate_tls_end(tls_last_offset, tls_last_size)) {
tls_last_offset -= obj->tlssize;
tls_last_size = 0;
}
}
Add support for Thread Local Storage.
2004-08-03 08:51:00 +00:00
void *
_rtld_allocate_tls(void *oldtls, size_t tcbsize, size_t tcbalign)
{
Add locking code for tls routines.
2005-03-20 23:28:25 +00:00
void *ret;
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
RtldLockState lockstate;
Add locking code for tls routines.
2005-03-20 23:28:25 +00:00
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
wlock_acquire(rtld_bind_lock, &lockstate);
Do not call callbacks for dl_iterate_phdr(3) with the rtld bind and phdr locks locked. This allows to call rtld services from the callback, which is only reasonable for dlopen(path, RTLD_NOLOAD) to test existence of the library in the image, and for dlsym(). The later might still be not quite safe, due to the lazy resolution of filters. To allow dropping the locks around iteration in dl_iterate_phdr(3), we insert markers to track current position between relocks. The global objects list is converted to tailq and all iterators skip markers, globallist_next() and globallist_curr() helpers are added. Reported and tested by: davide Reviewed by: kan Sponsored by: The FreeBSD Foundation MFC after: 3 weeks
2016-01-20 07:21:33 +00:00
ret = allocate_tls(globallist_curr(TAILQ_FIRST(&obj_list)), oldtls,
tcbsize, tcbalign);
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
lock_release(rtld_bind_lock, &lockstate);
Add locking code for tls routines.
2005-03-20 23:28:25 +00:00
return (ret);
Add support for Thread Local Storage.
2004-08-03 08:51:00 +00:00
}
void
_rtld_free_tls(void *tcb, size_t tcbsize, size_t tcbalign)
{
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
RtldLockState lockstate;
Add locking code for tls routines.
2005-03-20 23:28:25 +00:00
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
wlock_acquire(rtld_bind_lock, &lockstate);
Add support for Thread Local Storage.
2004-08-03 08:51:00 +00:00
free_tls(tcb, tcbsize, tcbalign);
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
lock_release(rtld_bind_lock, &lockstate);
Add support for Thread Local Storage.
2004-08-03 08:51:00 +00:00
}
Implement ELF symbol versioning using GNU semantics. This code aims to be compatible with symbol versioning support as implemented by GNU libc and documented by http://people.redhat.com/~drepper/symbol-versioning and LSB 3.0. Implement dlvsym() function to allow lookups for a specific version of a given symbol.
2005-12-18 19:43:33 +00:00
static void
object_add_name(Obj_Entry *obj, const char *name)
{
Name_Entry *entry;
size_t len;
len = strlen(name);
entry = malloc(sizeof(Name_Entry) + len);
if (entry != NULL) {
strcpy(entry->name, name);
STAILQ_INSERT_TAIL(&obj->names, entry, link);
}
}
static int
object_match_name(const Obj_Entry *obj, const char *name)
{
Name_Entry *entry;
STAILQ_FOREACH(entry, &obj->names, link) {
if (strcmp(name, entry->name) == 0)
return (1);
}
return (0);
}
static Obj_Entry *
locate_dependency(const Obj_Entry *obj, const char *name)
{
const Objlist_Entry *entry;
const Needed_Entry *needed;
STAILQ_FOREACH(entry, &list_main, link) {
if (object_match_name(entry->obj, name))
return entry->obj;
}
for (needed = obj->needed; needed != NULL; needed = needed->next) {
Make ldd(1) work when versioned dependency file is cannot be loaded. Instead of aborting in locate_dependency(), propagate the error to caller. The rtld startup function does the right thing with an error from rtld_verify_versions(), depending on the mode of operation. Reported by: maho In collaboration with: kan MFC after: 1 week
2011-01-30 16:14:09 +00:00
if (strcmp(obj->strtab + needed->name, name) == 0 ||
(needed->obj != NULL && object_match_name(needed->obj, name))) {
/*
* If there is DT_NEEDED for the name we are looking for,
* we are all set. Note that object might not be found if
* dependency was not loaded yet, so the function can
* return NULL here. This is expected and handled
Fix grammar in comment. Pointy hat to: kib MFC after: 1 week
2011-01-30 16:21:25 +00:00
* properly by the caller.
Make ldd(1) work when versioned dependency file is cannot be loaded. Instead of aborting in locate_dependency(), propagate the error to caller. The rtld startup function does the right thing with an error from rtld_verify_versions(), depending on the mode of operation. Reported by: maho In collaboration with: kan MFC after: 1 week
2011-01-30 16:14:09 +00:00
*/
return (needed->obj);
}
Implement ELF symbol versioning using GNU semantics. This code aims to be compatible with symbol versioning support as implemented by GNU libc and documented by http://people.redhat.com/~drepper/symbol-versioning and LSB 3.0. Implement dlvsym() function to allow lookups for a specific version of a given symbol.
2005-12-18 19:43:33 +00:00
}
Revert r209544. It papers over a binutils bug, and is not even a reliable fix for it. The bug occurs when using the --as-needed flag to ld in the presence of synthetic linker-generated symbols that reference symbols defined in linked-to shared libraries with versioned symbols. When the only symbols used from a library fall into this category, ld will drop the DT_NEEDED entry for it, but retain the versioning information. This bug is best fixed/hacked around in binutils, not in rtld. Discussed with: kan
2010-06-28 01:40:56 +00:00
_rtld_error("%s: Unexpected inconsistency: dependency %s not found",
obj->path, name);
Make die available as rtld_die for use by MD relocation code Reviewed by: kib Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D2213
2015-04-02 21:35:36 +00:00
rtld_die();
Implement ELF symbol versioning using GNU semantics. This code aims to be compatible with symbol versioning support as implemented by GNU libc and documented by http://people.redhat.com/~drepper/symbol-versioning and LSB 3.0. Implement dlvsym() function to allow lookups for a specific version of a given symbol.
2005-12-18 19:43:33 +00:00
}
static int
check_object_provided_version(Obj_Entry *refobj, const Obj_Entry *depobj,
const Elf_Vernaux *vna)
{
const Elf_Verdef *vd;
const char *vername;
vername = refobj->strtab + vna->vna_name;
vd = depobj->verdef;
if (vd == NULL) {
Make lookups for relocations from old unversioned binaries return oldest versioned symbol available. Do not accept hidden symbols for all other versions. Use "<obj->path>: <error message>" for all error messages in new functions to make them more consistent.
2005-12-24 15:37:30 +00:00
_rtld_error("%s: version %s required by %s not defined",
depobj->path, vername, refobj->path);
Implement ELF symbol versioning using GNU semantics. This code aims to be compatible with symbol versioning support as implemented by GNU libc and documented by http://people.redhat.com/~drepper/symbol-versioning and LSB 3.0. Implement dlvsym() function to allow lookups for a specific version of a given symbol.
2005-12-18 19:43:33 +00:00
return (-1);
}
for (;;) {
if (vd->vd_version != VER_DEF_CURRENT) {
Make lookups for relocations from old unversioned binaries return oldest versioned symbol available. Do not accept hidden symbols for all other versions. Use "<obj->path>: <error message>" for all error messages in new functions to make them more consistent.
2005-12-24 15:37:30 +00:00
_rtld_error("%s: Unsupported version %d of Elf_Verdef entry",
Implement ELF symbol versioning using GNU semantics. This code aims to be compatible with symbol versioning support as implemented by GNU libc and documented by http://people.redhat.com/~drepper/symbol-versioning and LSB 3.0. Implement dlvsym() function to allow lookups for a specific version of a given symbol.
2005-12-18 19:43:33 +00:00
depobj->path, vd->vd_version);
return (-1);
}
if (vna->vna_hash == vd->vd_hash) {
const Elf_Verdaux *aux = (const Elf_Verdaux *)
((char *)vd + vd->vd_aux);
if (strcmp(vername, depobj->strtab + aux->vda_name) == 0)
return (0);
}
if (vd->vd_next == 0)
break;
vd = (const Elf_Verdef *) ((char *)vd + vd->vd_next);
}
if (vna->vna_flags & VER_FLG_WEAK)
return (0);
Make lookups for relocations from old unversioned binaries return oldest versioned symbol available. Do not accept hidden symbols for all other versions. Use "<obj->path>: <error message>" for all error messages in new functions to make them more consistent.
2005-12-24 15:37:30 +00:00
_rtld_error("%s: version %s required by %s not found",
depobj->path, vername, refobj->path);
Implement ELF symbol versioning using GNU semantics. This code aims to be compatible with symbol versioning support as implemented by GNU libc and documented by http://people.redhat.com/~drepper/symbol-versioning and LSB 3.0. Implement dlvsym() function to allow lookups for a specific version of a given symbol.
2005-12-18 19:43:33 +00:00
return (-1);
}
static int
rtld_verify_object_versions(Obj_Entry *obj)
{
const Elf_Verneed *vn;
const Elf_Verdef *vd;
const Elf_Verdaux *vda;
const Elf_Vernaux *vna;
const Obj_Entry *depobj;
int maxvernum, vernum;
Prevent rtld_verify_object_versions() from being called several times for the same object. This can happen when object is a dependency of the dlopen()ed dso. When called several times, we waste time due to unneeded processing, and memory, because obj->vertab is allocated anew on each iteration. Reviewed by: kan MFC after: 2 weeks
2012-03-27 14:10:15 +00:00
if (obj->ver_checked)
return (0);
obj->ver_checked = true;
Implement ELF symbol versioning using GNU semantics. This code aims to be compatible with symbol versioning support as implemented by GNU libc and documented by http://people.redhat.com/~drepper/symbol-versioning and LSB 3.0. Implement dlvsym() function to allow lookups for a specific version of a given symbol.
2005-12-18 19:43:33 +00:00
maxvernum = 0;
/*
* Walk over defined and required version records and figure out
* max index used by any of them. Do very basic sanity checking
* while there.
*/
vn = obj->verneed;
while (vn != NULL) {
if (vn->vn_version != VER_NEED_CURRENT) {
Make lookups for relocations from old unversioned binaries return oldest versioned symbol available. Do not accept hidden symbols for all other versions. Use "<obj->path>: <error message>" for all error messages in new functions to make them more consistent.
2005-12-24 15:37:30 +00:00
_rtld_error("%s: Unsupported version %d of Elf_Verneed entry",
Implement ELF symbol versioning using GNU semantics. This code aims to be compatible with symbol versioning support as implemented by GNU libc and documented by http://people.redhat.com/~drepper/symbol-versioning and LSB 3.0. Implement dlvsym() function to allow lookups for a specific version of a given symbol.
2005-12-18 19:43:33 +00:00
obj->path, vn->vn_version);
return (-1);
}
vna = (const Elf_Vernaux *) ((char *)vn + vn->vn_aux);
for (;;) {
vernum = VER_NEED_IDX(vna->vna_other);
if (vernum > maxvernum)
maxvernum = vernum;
if (vna->vna_next == 0)
break;
vna = (const Elf_Vernaux *) ((char *)vna + vna->vna_next);
}
if (vn->vn_next == 0)
break;
vn = (const Elf_Verneed *) ((char *)vn + vn->vn_next);
}
vd = obj->verdef;
while (vd != NULL) {
if (vd->vd_version != VER_DEF_CURRENT) {
Make lookups for relocations from old unversioned binaries return oldest versioned symbol available. Do not accept hidden symbols for all other versions. Use "<obj->path>: <error message>" for all error messages in new functions to make them more consistent.
2005-12-24 15:37:30 +00:00
_rtld_error("%s: Unsupported version %d of Elf_Verdef entry",
obj->path, vd->vd_version);
Implement ELF symbol versioning using GNU semantics. This code aims to be compatible with symbol versioning support as implemented by GNU libc and documented by http://people.redhat.com/~drepper/symbol-versioning and LSB 3.0. Implement dlvsym() function to allow lookups for a specific version of a given symbol.
2005-12-18 19:43:33 +00:00
return (-1);
}
vernum = VER_DEF_IDX(vd->vd_ndx);
if (vernum > maxvernum)
maxvernum = vernum;
if (vd->vd_next == 0)
break;
vd = (const Elf_Verdef *) ((char *)vd + vd->vd_next);
}
if (maxvernum == 0)
return (0);
/*
* Store version information in array indexable by version index.
* Verify that object version requirements are satisfied along the
* way.
*/
obj->vernum = maxvernum + 1;
Use xmalloc() instead of malloc() in the places where malloc() calls are assumed to not fail. Make the xcalloc() calling conventions follow the calloc(3) calling conventions and replace unchecked calls to calloc() with calls to xcalloc(). Remove redundand declarations from xmalloc.c, which are already present in rtld.h. Reviewed by: kan Discussed with: bde MFC after: 2 weeks
2012-03-22 14:20:51 +00:00
obj->vertab = xcalloc(obj->vernum, sizeof(Ver_Entry));
Implement ELF symbol versioning using GNU semantics. This code aims to be compatible with symbol versioning support as implemented by GNU libc and documented by http://people.redhat.com/~drepper/symbol-versioning and LSB 3.0. Implement dlvsym() function to allow lookups for a specific version of a given symbol.
2005-12-18 19:43:33 +00:00
vd = obj->verdef;
while (vd != NULL) {
if ((vd->vd_flags & VER_FLG_BASE) == 0) {
vernum = VER_DEF_IDX(vd->vd_ndx);
assert(vernum <= maxvernum);
vda = (const Elf_Verdaux *)((char *)vd + vd->vd_aux);
obj->vertab[vernum].hash = vd->vd_hash;
obj->vertab[vernum].name = obj->strtab + vda->vda_name;
obj->vertab[vernum].file = NULL;
obj->vertab[vernum].flags = 0;
}
if (vd->vd_next == 0)
break;
vd = (const Elf_Verdef *) ((char *)vd + vd->vd_next);
}
vn = obj->verneed;
while (vn != NULL) {
depobj = locate_dependency(obj, obj->strtab + vn->vn_file);
Make ldd(1) work when versioned dependency file is cannot be loaded. Instead of aborting in locate_dependency(), propagate the error to caller. The rtld startup function does the right thing with an error from rtld_verify_versions(), depending on the mode of operation. Reported by: maho In collaboration with: kan MFC after: 1 week
2011-01-30 16:14:09 +00:00
if (depobj == NULL)
return (-1);
Implement ELF symbol versioning using GNU semantics. This code aims to be compatible with symbol versioning support as implemented by GNU libc and documented by http://people.redhat.com/~drepper/symbol-versioning and LSB 3.0. Implement dlvsym() function to allow lookups for a specific version of a given symbol.
2005-12-18 19:43:33 +00:00
vna = (const Elf_Vernaux *) ((char *)vn + vn->vn_aux);
for (;;) {
if (check_object_provided_version(obj, depobj, vna))
return (-1);
vernum = VER_NEED_IDX(vna->vna_other);
assert(vernum <= maxvernum);
obj->vertab[vernum].hash = vna->vna_hash;
obj->vertab[vernum].name = obj->strtab + vna->vna_name;
obj->vertab[vernum].file = obj->strtab + vn->vn_file;
obj->vertab[vernum].flags = (vna->vna_other & VER_NEED_HIDDEN) ?
VER_INFO_HIDDEN : 0;
if (vna->vna_next == 0)
break;
vna = (const Elf_Vernaux *) ((char *)vna + vna->vna_next);
}
if (vn->vn_next == 0)
break;
vn = (const Elf_Verneed *) ((char *)vn + vn->vn_next);
}
return 0;
}
static int
rtld_verify_versions(const Objlist *objlist)
{
Objlist_Entry *entry;
int rc;
rc = 0;
STAILQ_FOREACH(entry, objlist, link) {
/*
* Skip dummy objects or objects that have their version requirements
* already checked.
*/
if (entry->obj->strtab == NULL || entry->obj->vertab != NULL)
continue;
if (rtld_verify_object_versions(entry->obj) == -1) {
rc = -1;
if (ld_tracing == NULL)
break;
}
}
Catch up on rtld's special status. Since it does not appear on main object list, its versioning information needs to be examined separately. This hopefully fixes problems that people running with SYMVER_ENABLED are experiencing.
2007-04-07 23:17:00 +00:00
if (rc == 0 || ld_tracing != NULL)
rc = rtld_verify_object_versions(&obj_rtld);
Implement ELF symbol versioning using GNU semantics. This code aims to be compatible with symbol versioning support as implemented by GNU libc and documented by http://people.redhat.com/~drepper/symbol-versioning and LSB 3.0. Implement dlvsym() function to allow lookups for a specific version of a given symbol.
2005-12-18 19:43:33 +00:00
return rc;
}
const Ver_Entry *
fetch_ventry(const Obj_Entry *obj, unsigned long symnum)
{
Elf_Versym vernum;
if (obj->vertab) {
vernum = VER_NDX(obj->versyms[symnum]);
if (vernum >= obj->vernum) {
_rtld_error("%s: symbol %s has wrong verneed value %d",
obj->path, obj->strtab + symnum, vernum);
} else if (obj->vertab[vernum].hash != 0) {
return &obj->vertab[vernum];
}
}
return NULL;
}
Use the newly provided aux vectors to get pagesize and osreldate information. Use local version of getpagesize(), rtld_getpagesize() in private allocator. Override the __getosreldate() previously fetched from libc_pic.a with local version that uses aux value if present. Note that __getosreldate() is used by rtld indirectly, by mmap(2) libc wrapper. To be able to utilize aux, split digest_dynamic() for use by init_rtld() into two parts, where the first one does not call malloc(), and the second part uses it. init_rtld() is able to initialize global variables before digest_dynamic2() calls. In particular, pagesize and osreldate are set up from the aux values. Now, rtld avoids (two) sysctl calls in startup. Tested by: marius (sparc64) MFC after: 1 month
2010-08-17 09:05:39 +00:00
In rtld, read the initial stack access mode from AT_STACKPROT as set by kernel, and parse PT_GNU_STACK phdr from linked and loaded dsos. If the loaded dso requires executable stack, as specified by PF_X bit of p_flags of PT_GNU_STACK phdr, but current stack protection does not permit execution, the __pthread_map_stacks_exec symbol is looked up and called. It should be implemented in libc or threading library and change the protection mode of all thread stacks to be executable. Provide a private interface _rtld_get_stack_prot() to export the stack access mode as calculated by rtld. Reviewed by: kan
2011-01-08 17:11:49 +00:00
int
_rtld_get_stack_prot(void)
{
return (stack_prot);
}
Add rtld private interface for dso to detect dynamic loading vs. static linking. Tested by: pho, antoine (exp-run) Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-01-03 18:09:53 +00:00
int
_rtld_is_dlopened(void *arg)
{
Obj_Entry *obj;
RtldLockState lockstate;
int res;
rlock_acquire(rtld_bind_lock, &lockstate);
obj = dlcheck(arg);
if (obj == NULL)
obj = obj_from_addr(arg);
if (obj == NULL) {
_rtld_error("No shared object contains address");
lock_release(rtld_bind_lock, &lockstate);
return (-1);
}
res = obj->dlopened ? 1 : 0;
lock_release(rtld_bind_lock, &lockstate);
return (res);
}
For the main binary, postpone enforcing relro read-only protection until copy relocations are done. Newer binutils and lld seems to output copy into relro-protected range. Reported by: Rafael Espц╜ndola via emaste Sponsored by: The FreeBSD Foundation MFC after: 1 week
2017-01-12 15:54:03 +00:00
int
obj_enforce_relro(Obj_Entry *obj)
{
if (obj->relro_size > 0 && mprotect(obj->relro_page, obj->relro_size,
PROT_READ) == -1) {
_rtld_error("%s: Cannot enforce relro protection: %s",
obj->path, rtld_strerror(errno));
return (-1);
}
return (0);
}
In rtld, read the initial stack access mode from AT_STACKPROT as set by kernel, and parse PT_GNU_STACK phdr from linked and loaded dsos. If the loaded dso requires executable stack, as specified by PF_X bit of p_flags of PT_GNU_STACK phdr, but current stack protection does not permit execution, the __pthread_map_stacks_exec symbol is looked up and called. It should be implemented in libc or threading library and change the protection mode of all thread stacks to be executable. Provide a private interface _rtld_get_stack_prot() to export the stack access mode as calculated by rtld. Reviewed by: kan
2011-01-08 17:11:49 +00:00
static void
get_program_var_addr() must prefer the strong symbol to the weak one. Search global objects, together with main object and dependencies, for the requested symbol. Move the common code from symlook_default() into new helper symlook_global(), and use it both in symlook_global() and get_program_var_addr(). Supply lock state to get_program_var_addr(). Reviewed by: kan Tested by: Mykola Dzham <i levsha me>
2011-01-10 16:09:35 +00:00
map_stacks_exec(RtldLockState *lockstate)
In rtld, read the initial stack access mode from AT_STACKPROT as set by kernel, and parse PT_GNU_STACK phdr from linked and loaded dsos. If the loaded dso requires executable stack, as specified by PF_X bit of p_flags of PT_GNU_STACK phdr, but current stack protection does not permit execution, the __pthread_map_stacks_exec symbol is looked up and called. It should be implemented in libc or threading library and change the protection mode of all thread stacks to be executable. Provide a private interface _rtld_get_stack_prot() to export the stack access mode as calculated by rtld. Reviewed by: kan
2011-01-08 17:11:49 +00:00
{
void (*thr_map_stacks_exec)(void);
if ((max_stack_flags & PF_X) == 0 || (stack_prot & PROT_EXEC) != 0)
return;
thr_map_stacks_exec = (void (*)(void))(uintptr_t)
get_program_var_addr() must prefer the strong symbol to the weak one. Search global objects, together with main object and dependencies, for the requested symbol. Move the common code from symlook_default() into new helper symlook_global(), and use it both in symlook_global() and get_program_var_addr(). Supply lock state to get_program_var_addr(). Reviewed by: kan Tested by: Mykola Dzham <i levsha me>
2011-01-10 16:09:35 +00:00
get_program_var_addr("__pthread_map_stacks_exec", lockstate);
In rtld, read the initial stack access mode from AT_STACKPROT as set by kernel, and parse PT_GNU_STACK phdr from linked and loaded dsos. If the loaded dso requires executable stack, as specified by PF_X bit of p_flags of PT_GNU_STACK phdr, but current stack protection does not permit execution, the __pthread_map_stacks_exec symbol is looked up and called. It should be implemented in libc or threading library and change the protection mode of all thread stacks to be executable. Provide a private interface _rtld_get_stack_prot() to export the stack access mode as calculated by rtld. Reviewed by: kan
2011-01-08 17:11:49 +00:00
if (thr_map_stacks_exec != NULL) {
stack_prot |= PROT_EXEC;
thr_map_stacks_exec();
}
}
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
void
symlook_init(SymLook *dst, const char *name)
{
bzero(dst, sizeof(*dst));
dst->name = name;
dst->hash = elf_hash(name);
Add GNU hash support for rtld. Based on dragonflybsd support for GNU hash by John Marino <draco marino st> Reviewed by: kan Tested by: bapt MFC after: 2 weeks
2012-04-30 13:31:10 +00:00
dst->hash_gnu = gnu_hash(name);
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
}
static void
symlook_init_from_req(SymLook *dst, const SymLook *src)
{
dst->name = src->name;
dst->hash = src->hash;
Add GNU hash support for rtld. Based on dragonflybsd support for GNU hash by John Marino <draco marino st> Reviewed by: kan Tested by: bapt MFC after: 2 weeks
2012-04-30 13:31:10 +00:00
dst->hash_gnu = src->hash_gnu;
Implement support for ELF filters in rtld. Both normal and auxillary filters are implemented. Filtees are loaded on demand, unless LD_LOADFLTR environment variable is set or -z loadfltr was specified during the linking. This forces rtld to upgrade read-locked rtld_bind_lock to write lock when it encounters an object with filter during symbol lookup. Consolidate common arguments of the symbol lookup functions in the SymLook structure. Track the state of the rtld locks in the RtldLockState structure. Pass local RtldLockState through the rtld symbol lookup calls to allow lock upgrades. Reviewed by: kan Tested by: Mykola Dzham <i levsha me>, nwhitehorn (powerpc)
2010-12-25 08:51:20 +00:00
dst->ventry = src->ventry;
dst->flags = src->flags;
dst->defobj_out = NULL;
dst->sym_out = NULL;
dst->lockstate = src->lockstate;
}
For ld.so direct execution mode, implement -p option: search for the binary in $PATH. Sponsored by: The FreeBSD Foundation MFC after: 2 weeks Differential revision: https://reviews.freebsd.org/D10790
2017-05-23 10:00:52 +00:00
static int
open_binary_fd(const char *argv0, bool search_in_path)
{
char *pathenv, *pe, binpath[PATH_MAX];
int fd;
if (search_in_path && strchr(argv0, '/') == NULL) {
pathenv = getenv("PATH");
if (pathenv == NULL) {
rtld_printf("-p and no PATH environment variable\n");
rtld_die();
}
pathenv = strdup(pathenv);
if (pathenv == NULL) {
rtld_printf("Cannot allocate memory\n");
rtld_die();
}
fd = -1;
errno = ENOENT;
while ((pe = strsep(&pathenv, ":")) != NULL) {
In open_binary_fd: when using buffer size for strl* and snprintf, always use >= instead of > to avoid truncation. Reviewed by: kib Differential Revision: https://reviews.freebsd.org/D11474 MFC after: 3 days
2017-07-05 06:12:21 +00:00
if (strlcpy(binpath, pe, sizeof(binpath)) >=
For ld.so direct execution mode, implement -p option: search for the binary in $PATH. Sponsored by: The FreeBSD Foundation MFC after: 2 weeks Differential revision: https://reviews.freebsd.org/D10790
2017-05-23 10:00:52 +00:00
sizeof(binpath))
continue;
if (binpath[0] != '\0' &&
In open_binary_fd: when using buffer size for strl* and snprintf, always use >= instead of > to avoid truncation. Reviewed by: kib Differential Revision: https://reviews.freebsd.org/D11474 MFC after: 3 days
2017-07-05 06:12:21 +00:00
strlcat(binpath, "/", sizeof(binpath)) >=
For ld.so direct execution mode, implement -p option: search for the binary in $PATH. Sponsored by: The FreeBSD Foundation MFC after: 2 weeks Differential revision: https://reviews.freebsd.org/D10790
2017-05-23 10:00:52 +00:00
sizeof(binpath))
continue;
In open_binary_fd: when using buffer size for strl* and snprintf, always use >= instead of > to avoid truncation. Reviewed by: kib Differential Revision: https://reviews.freebsd.org/D11474 MFC after: 3 days
2017-07-05 06:12:21 +00:00
if (strlcat(binpath, argv0, sizeof(binpath)) >=
For ld.so direct execution mode, implement -p option: search for the binary in $PATH. Sponsored by: The FreeBSD Foundation MFC after: 2 weeks Differential revision: https://reviews.freebsd.org/D10790
2017-05-23 10:00:52 +00:00
sizeof(binpath))
continue;
fd = open(binpath, O_RDONLY | O_CLOEXEC | O_VERIFY);
if (fd != -1 || errno != ENOENT)
break;
}
free(pathenv);
} else {
fd = open(argv0, O_RDONLY | O_CLOEXEC | O_VERIFY);
}
if (fd == -1) {
rtld_printf("Opening %s: %s\n", argv0,
rtld_strerror(errno));
rtld_die();
}
return (fd);
}
Add the LD_LIBRARY_PATH_FDS environmental variable. This variable allows the loading of shared libraries via directory descriptors rather than via library paths. If LD_LIBRARY_PATH_FDS=3:4:12, the directories represented by file descriptors 3, 4 and 12 will searched for shared libraries before the normal path-based mechanisms are used. This allows us to execute unprivileged binaries from within a Capsicum sandbox even if they require shared libraries. Approved by: rwatson (mentor) Reviewed by: kib MFC after: 3 weeks Sponsored by: DARPA/AFRL
2014-06-20 17:08:32 +00:00
Allow rtld direct-exec to take a file descriptor. When executing rtld directly, allow a file descriptor to be explicitly specified rather than opened from the given path. This, together with the LD_LIBRARY_PATH_FDS environment variable, allows dynamically-linked applications to be executed from within capability mode. Also add some rudimentary argument parsing (without pulling in getopt or the like) to accept this file descriptor, a help (-h) option and a basic usage string. Reviewed by: kib Sponsored by: NSERC, RDC Differential Revision: https://reviews.freebsd.org/D10751
2017-05-17 22:51:28 +00:00
/*
* Parse a set of command-line arguments.
*/
static int
parse_args(char* argv[], int argc, bool *use_pathp, int *fdp)
{
const char *arg;
int fd, i, j, arglen;
char opt;
dbg("Parsing command-line arguments");
*use_pathp = false;
*fdp = -1;
for (i = 1; i < argc; i++ ) {
arg = argv[i];
dbg("argv[%d]: '%s'", i, arg);
/*
* rtld arguments end with an explicit "--" or with the first
* non-prefixed argument.
*/
if (strcmp(arg, "--") == 0) {
i++;
break;
}
if (arg[0] != '-')
break;
/*
* All other arguments are single-character options that can
* be combined, so we need to search through `arg` for them.
*/
arglen = strlen(arg);
for (j = 1; j < arglen; j++) {
opt = arg[j];
if (opt == 'h') {
print_usage(argv[0]);
rtld_die();
} else if (opt == 'f') {
/*
* -f XX can be used to specify a descriptor for the
* binary named at the command line (i.e., the later
* argument will specify the process name but the
* descriptor is what will actually be executed)
*/
if (j != arglen - 1) {
/* -f must be the last option in, e.g., -abcf */
_rtld_error("invalid options: %s", arg);
rtld_die();
}
i++;
fd = parse_integer(argv[i]);
if (fd == -1) {
_rtld_error("invalid file descriptor: '%s'",
argv[i]);
rtld_die();
}
*fdp = fd;
break;
} else if (opt == 'p') {
*use_pathp = true;
} else {
rtld_printf("invalid argument: '%s'\n", arg);
print_usage(argv[0]);
rtld_die();
}
}
}
return (i);
}
Add the LD_LIBRARY_PATH_FDS environmental variable. This variable allows the loading of shared libraries via directory descriptors rather than via library paths. If LD_LIBRARY_PATH_FDS=3:4:12, the directories represented by file descriptors 3, 4 and 12 will searched for shared libraries before the normal path-based mechanisms are used. This allows us to execute unprivileged binaries from within a Capsicum sandbox even if they require shared libraries. Approved by: rwatson (mentor) Reviewed by: kib MFC after: 3 weeks Sponsored by: DARPA/AFRL
2014-06-20 17:08:32 +00:00
/*
* Parse a file descriptor number without pulling in more of libc (e.g. atoi).
*/
static int
Rename rtld's parse_libdir to parse_integer. This is a more accurate name, as the integer doesn't have to be a library directory descriptor. It is also a prerequisite for more argument parsing coming in the near future (e.g., parsing explicit binary descriptors). Reviewed by: kib MFC after: 2 weeks Sponsored by: NSERC
2017-05-16 13:27:44 +00:00
parse_integer(const char *str)
Add the LD_LIBRARY_PATH_FDS environmental variable. This variable allows the loading of shared libraries via directory descriptors rather than via library paths. If LD_LIBRARY_PATH_FDS=3:4:12, the directories represented by file descriptors 3, 4 and 12 will searched for shared libraries before the normal path-based mechanisms are used. This allows us to execute unprivileged binaries from within a Capsicum sandbox even if they require shared libraries. Approved by: rwatson (mentor) Reviewed by: kib MFC after: 3 weeks Sponsored by: DARPA/AFRL
2014-06-20 17:08:32 +00:00
{
static const int RADIX = 10; /* XXXJA: possibly support hex? */
const char *orig;
Rename rtld's parse_libdir to parse_integer. This is a more accurate name, as the integer doesn't have to be a library directory descriptor. It is also a prerequisite for more argument parsing coming in the near future (e.g., parsing explicit binary descriptors). Reviewed by: kib MFC after: 2 weeks Sponsored by: NSERC
2017-05-16 13:27:44 +00:00
int n;
Add the LD_LIBRARY_PATH_FDS environmental variable. This variable allows the loading of shared libraries via directory descriptors rather than via library paths. If LD_LIBRARY_PATH_FDS=3:4:12, the directories represented by file descriptors 3, 4 and 12 will searched for shared libraries before the normal path-based mechanisms are used. This allows us to execute unprivileged binaries from within a Capsicum sandbox even if they require shared libraries. Approved by: rwatson (mentor) Reviewed by: kib MFC after: 3 weeks Sponsored by: DARPA/AFRL
2014-06-20 17:08:32 +00:00
char c;
orig = str;
Rename rtld's parse_libdir to parse_integer. This is a more accurate name, as the integer doesn't have to be a library directory descriptor. It is also a prerequisite for more argument parsing coming in the near future (e.g., parsing explicit binary descriptors). Reviewed by: kib MFC after: 2 weeks Sponsored by: NSERC
2017-05-16 13:27:44 +00:00
n = 0;
Add the LD_LIBRARY_PATH_FDS environmental variable. This variable allows the loading of shared libraries via directory descriptors rather than via library paths. If LD_LIBRARY_PATH_FDS=3:4:12, the directories represented by file descriptors 3, 4 and 12 will searched for shared libraries before the normal path-based mechanisms are used. This allows us to execute unprivileged binaries from within a Capsicum sandbox even if they require shared libraries. Approved by: rwatson (mentor) Reviewed by: kib MFC after: 3 weeks Sponsored by: DARPA/AFRL
2014-06-20 17:08:32 +00:00
for (c = *str; c != '\0'; c = *++str) {
if (c < '0' || c > '9')
return (-1);
Rename rtld's parse_libdir to parse_integer. This is a more accurate name, as the integer doesn't have to be a library directory descriptor. It is also a prerequisite for more argument parsing coming in the near future (e.g., parsing explicit binary descriptors). Reviewed by: kib MFC after: 2 weeks Sponsored by: NSERC
2017-05-16 13:27:44 +00:00
n *= RADIX;
n += c - '0';
Add the LD_LIBRARY_PATH_FDS environmental variable. This variable allows the loading of shared libraries via directory descriptors rather than via library paths. If LD_LIBRARY_PATH_FDS=3:4:12, the directories represented by file descriptors 3, 4 and 12 will searched for shared libraries before the normal path-based mechanisms are used. This allows us to execute unprivileged binaries from within a Capsicum sandbox even if they require shared libraries. Approved by: rwatson (mentor) Reviewed by: kib MFC after: 3 weeks Sponsored by: DARPA/AFRL
2014-06-20 17:08:32 +00:00
}
/* Make sure we actually parsed something. */
Rename rtld's parse_libdir to parse_integer. This is a more accurate name, as the integer doesn't have to be a library directory descriptor. It is also a prerequisite for more argument parsing coming in the near future (e.g., parsing explicit binary descriptors). Reviewed by: kib MFC after: 2 weeks Sponsored by: NSERC
2017-05-16 13:27:44 +00:00
if (str == orig)
Add the LD_LIBRARY_PATH_FDS environmental variable. This variable allows the loading of shared libraries via directory descriptors rather than via library paths. If LD_LIBRARY_PATH_FDS=3:4:12, the directories represented by file descriptors 3, 4 and 12 will searched for shared libraries before the normal path-based mechanisms are used. This allows us to execute unprivileged binaries from within a Capsicum sandbox even if they require shared libraries. Approved by: rwatson (mentor) Reviewed by: kib MFC after: 3 weeks Sponsored by: DARPA/AFRL
2014-06-20 17:08:32 +00:00
return (-1);
Rename rtld's parse_libdir to parse_integer. This is a more accurate name, as the integer doesn't have to be a library directory descriptor. It is also a prerequisite for more argument parsing coming in the near future (e.g., parsing explicit binary descriptors). Reviewed by: kib MFC after: 2 weeks Sponsored by: NSERC
2017-05-16 13:27:44 +00:00
return (n);
Add the LD_LIBRARY_PATH_FDS environmental variable. This variable allows the loading of shared libraries via directory descriptors rather than via library paths. If LD_LIBRARY_PATH_FDS=3:4:12, the directories represented by file descriptors 3, 4 and 12 will searched for shared libraries before the normal path-based mechanisms are used. This allows us to execute unprivileged binaries from within a Capsicum sandbox even if they require shared libraries. Approved by: rwatson (mentor) Reviewed by: kib MFC after: 3 weeks Sponsored by: DARPA/AFRL
2014-06-20 17:08:32 +00:00
}
Fix style [1], add static keyword before static function definition. Noted by: bapt [1] Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2017-05-18 09:31:30 +00:00
static void
print_usage(const char *argv0)
Allow rtld direct-exec to take a file descriptor. When executing rtld directly, allow a file descriptor to be explicitly specified rather than opened from the given path. This, together with the LD_LIBRARY_PATH_FDS environment variable, allows dynamically-linked applications to be executed from within capability mode. Also add some rudimentary argument parsing (without pulling in getopt or the like) to accept this file descriptor, a help (-h) option and a basic usage string. Reviewed by: kib Sponsored by: NSERC, RDC Differential Revision: https://reviews.freebsd.org/D10751
2017-05-17 22:51:28 +00:00
{
rtld_printf("Usage: %s [-h] [-f <FD>] [--] <binary> [<args>]\n"
"\n"
"Options:\n"
" -h Display this help message\n"
For ld.so direct execution mode, implement -p option: search for the binary in $PATH. Sponsored by: The FreeBSD Foundation MFC after: 2 weeks Differential revision: https://reviews.freebsd.org/D10790
2017-05-23 10:00:52 +00:00
" -p Search in PATH for named binary\n"
Allow rtld direct-exec to take a file descriptor. When executing rtld directly, allow a file descriptor to be explicitly specified rather than opened from the given path. This, together with the LD_LIBRARY_PATH_FDS environment variable, allows dynamically-linked applications to be executed from within capability mode. Also add some rudimentary argument parsing (without pulling in getopt or the like) to accept this file descriptor, a help (-h) option and a basic usage string. Reviewed by: kib Sponsored by: NSERC, RDC Differential Revision: https://reviews.freebsd.org/D10751
2017-05-17 22:51:28 +00:00
" -f <FD> Execute <FD> instead of searching for <binary>\n"
" -- End of RTLD options\n"
" <binary> Name of process to execute\n"
" <args> Arguments to the executed process\n", argv0);
}
On shared object unload, in __cxa_finalize, call and clear all installed atexit and __cxa_atexit handlers that are either installed by unloaded dso, or points to the functions provided by the dso. Use _rtld_addr_phdr to locate segment information from the address of private variable belonging to the dso, supplied by crtstuff.c. Provide utility function __elf_phdr_match_addr to do the match of address against dso executable segment. Call back into libthr from __cxa_finalize using weak __pthread_cxa_finalize symbol to remove any atfork handler which function points into unloaded object. The rtld needs private __pthread_cxa_finalize symbol to not require resolution of the weak undefined symbol at initialization time. This cannot work, since rtld is relocated before sym_zero is set up. Idea by: kan Reviewed by: kan (previous version) MFC after: 3 weeks
2010-08-23 15:38:02 +00:00
/*
* Overrides for libc_pic-provided functions.
*/
Use the newly provided aux vectors to get pagesize and osreldate information. Use local version of getpagesize(), rtld_getpagesize() in private allocator. Override the __getosreldate() previously fetched from libc_pic.a with local version that uses aux value if present. Note that __getosreldate() is used by rtld indirectly, by mmap(2) libc wrapper. To be able to utilize aux, split digest_dynamic() for use by init_rtld() into two parts, where the first one does not call malloc(), and the second part uses it. init_rtld() is able to initialize global variables before digest_dynamic2() calls. In particular, pagesize and osreldate are set up from the aux values. Now, rtld avoids (two) sysctl calls in startup. Tested by: marius (sparc64) MFC after: 1 month
2010-08-17 09:05:39 +00:00
int
__getosreldate(void)
{
size_t len;
int oid[2];
int error, osrel;
if (osreldate != 0)
return (osreldate);
oid[0] = CTL_KERN;
oid[1] = KERN_OSRELDATE;
osrel = 0;
len = sizeof(osrel);
error = sysctl(oid, 2, &osrel, &len, NULL, 0);
if (error == 0 && osrel > 0 && len == sizeof(osrel))
osreldate = osrel;
return (osreldate);
}
On shared object unload, in __cxa_finalize, call and clear all installed atexit and __cxa_atexit handlers that are either installed by unloaded dso, or points to the functions provided by the dso. Use _rtld_addr_phdr to locate segment information from the address of private variable belonging to the dso, supplied by crtstuff.c. Provide utility function __elf_phdr_match_addr to do the match of address against dso executable segment. Call back into libthr from __cxa_finalize using weak __pthread_cxa_finalize symbol to remove any atfork handler which function points into unloaded object. The rtld needs private __pthread_cxa_finalize symbol to not require resolution of the weak undefined symbol at initialization time. This cannot work, since rtld is relocated before sym_zero is set up. Idea by: kan Reviewed by: kan (previous version) MFC after: 3 weeks
2010-08-23 15:38:02 +00:00
Provide short-circuit exit(3) implementation for rtld. There is no atexit finalizers in rtld to call on rtld exiting (due to errors). Submitted by: bde No objections from: kan MFC after: 2 weeks
2012-03-24 19:13:10 +00:00
void
exit(int status)
{
_exit(status);
}
void (*__cleanup)(void);
int __isthreaded = 0;
int _thread_autoinit_dummy_decl = 1;
On shared object unload, in __cxa_finalize, call and clear all installed atexit and __cxa_atexit handlers that are either installed by unloaded dso, or points to the functions provided by the dso. Use _rtld_addr_phdr to locate segment information from the address of private variable belonging to the dso, supplied by crtstuff.c. Provide utility function __elf_phdr_match_addr to do the match of address against dso executable segment. Call back into libthr from __cxa_finalize using weak __pthread_cxa_finalize symbol to remove any atfork handler which function points into unloaded object. The rtld needs private __pthread_cxa_finalize symbol to not require resolution of the weak undefined symbol at initialization time. This cannot work, since rtld is relocated before sym_zero is set up. Idea by: kan Reviewed by: kan (previous version) MFC after: 3 weeks
2010-08-23 15:38:02 +00:00
/*
* No unresolved symbols for rtld.
*/
void
__pthread_cxa_finalize(struct dl_phdr_info *a)
{
}
Rtld on diet part 1: Provide rtld-private implementations of __stack_chk_guard, __stack_chk_fail() and __chk_fail() symbols, to be used by functions linked from libc_pic.a. This avoids use of libc stack_protector.c, which pulls in syslog(3) and stdio as dependency. Also, do initialize rtld-private copy __stack_chk_guard, previously libc-provided one was not initialized, since we do not call rtld object _init() methods. Reviewed by: kan MFC after: 3 weeks
2012-03-12 12:15:47 +00:00
void
__stack_chk_fail(void)
{
_rtld_error("stack overflow detected; terminated");
Make die available as rtld_die for use by MD relocation code Reviewed by: kib Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D2213
2015-04-02 21:35:36 +00:00
rtld_die();
Rtld on diet part 1: Provide rtld-private implementations of __stack_chk_guard, __stack_chk_fail() and __chk_fail() symbols, to be used by functions linked from libc_pic.a. This avoids use of libc stack_protector.c, which pulls in syslog(3) and stdio as dependency. Also, do initialize rtld-private copy __stack_chk_guard, previously libc-provided one was not initialized, since we do not call rtld object _init() methods. Reviewed by: kan MFC after: 3 weeks
2012-03-12 12:15:47 +00:00
}
Remove libssp_nonshared from the rtld linking set. The only use for the library was definition for the weak alias of __stack_chk_fail. No objections from: kan MFC after: 2 weeks
2012-03-24 19:14:44 +00:00
__weak_reference(__stack_chk_fail, __stack_chk_fail_local);
Rtld on diet part 1: Provide rtld-private implementations of __stack_chk_guard, __stack_chk_fail() and __chk_fail() symbols, to be used by functions linked from libc_pic.a. This avoids use of libc stack_protector.c, which pulls in syslog(3) and stdio as dependency. Also, do initialize rtld-private copy __stack_chk_guard, previously libc-provided one was not initialized, since we do not call rtld object _init() methods. Reviewed by: kan MFC after: 3 weeks
2012-03-12 12:15:47 +00:00
void
__chk_fail(void)
{
_rtld_error("buffer overflow detected; terminated");
Make die available as rtld_die for use by MD relocation code Reviewed by: kib Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D2213
2015-04-02 21:35:36 +00:00
rtld_die();
Rtld on diet part 1: Provide rtld-private implementations of __stack_chk_guard, __stack_chk_fail() and __chk_fail() symbols, to be used by functions linked from libc_pic.a. This avoids use of libc stack_protector.c, which pulls in syslog(3) and stdio as dependency. Also, do initialize rtld-private copy __stack_chk_guard, previously libc-provided one was not initialized, since we do not call rtld object _init() methods. Reviewed by: kan MFC after: 3 weeks
2012-03-12 12:15:47 +00:00
}
Rtld on diet 3. Stop using strerror(3) in rtld, which brings in msgcat and stdio. Directly access sys_errlist array of errno messages with private rtld_strerror() function. Now, $ size /libexec/ld-elf.so.1 text data bss dec hex filename 96983 2480 8744 108207 1a6af /libexec/ld-elf.so.1 Reviewed by: dim, kan MFC after: 2 weeks
2012-03-14 15:39:59 +00:00
const char *
rtld_strerror(int errnum)
{
if (errnum < 0 || errnum >= sys_nerr)
return ("Unknown error");
return (sys_errlist[errnum]);
}
Reference in New Issue Copy Permalink