The original code did not support dynamically loaded libraries and used
suboptimal access to TLS variables.
New implementation removes lazy resolving of TLS relocation - due to flaw
in TLSDESC design is impossible to switch resolver function at runtime
without expensive locking.
Due to this, 3 specialized resolvers are implemented:
- load time resolver for TLS relocation from libraries loaded with main
executable (thus with known TLS offset).
- resolver for undefined thread weak symbols.
- slower lazy resolver for dynamically loaded libraries with fast path for
already resolved symbols.
PR: 228892, 232149, 233204, 232311
MFC after: 2 weeks
Differential Revision: https://reviews.freebsd.org/D18417
The above commit fixed handling overaligned TLS segments in libc's
TLS Variant I implementation, but rtld provides its own implementation
for dynamically-linked executables which lacks these fixes. Thus,
port these changes to rtld.
This was previously commited as r337978 and reverted in r338149 due to
exposing a bug the ARM rtld. This bug was fixed in r338317 by mmel.
Submitted by: James Clarke
Approved by: re (kib)
Reviewed by: kbowling
Testing by: kbowling (powerpc64), br (riscv), kevans (armv7)
Obtained from: CheriBSD
Sponsored by: DARPA, AFRL
Differential Revision: https://reviews.freebsd.org/D16510
Michal Meloun reports that it breaks ctype (isspace()..) related
functions on armv7 so back out while we diagnose the issue.
Reported by: Michal Meloun <melounmichal@gmail.com>
The above commit fixed handling overaligned TLS segments in libc's
TLS Variant I implementation, but rtld provides its own implementation
for dynamically-linked executables which lacks these fixes. Thus,
port these changes to rtld.
Submitted by: James Clarke
Reviewed by: kbowling
Testing byL kbowling (powerpc64), br (riscv), kevans (armv7)
Obtained from: CheriBSD
Sponsored by: DARPA, AFRL
Differential Revision: https://reviews.freebsd.org/D16510
objects' init functions instead of doing the setup via a constructor
in libc as the init functions may already depend on these handlers
to be in place. This gets us rid of:
- the undefined order in which libc constructors as __guard_setup()
and jemalloc_constructor() are executed WRT __sparc_utrap_setup(),
- the requirement to link libc last so __sparc_utrap_setup() gets
called prior to constructors in other libraries (see r122883).
For static binaries, crt1.o still sets up the user trap handlers.
o Move misplaced prototypes for MD functions in to the MD prototype
section of rtld.h.
o Sprinkle nitems().
Mainly focus on files that use BSD 2-Clause license, however the tool I
was using misidentified many licenses so this was mostly a manual - error
prone - task.
The Software Package Data Exchange (SPDX) group provides a specification
to make it easier for automated tools to detect and summarize well known
opensource licenses. We are gradually adopting the specification, noting
that the tags are considered only advisory and do not, in any way,
superceed or replace the license texts.
No functional change intended.
The directives I added in r323466 and r323501 did not define a valid
CFA until several instructions into the associated functions. This
triggers an assertion in GDB when generating a stack trace while
stopped at the first instruction of PLT stub entry point since there
is no valid CFA rule for the first instruction.
This is probably just wrong on my part as the non-simple .cfi_startproc
would have defined a valid CFA. Instead, define a valid CFA as sp + 0
at the start of the functions and then use .cfa_def_offset to change the
offset when sp is adjusted later in the function.
Sponsored by: DARPA / AFRL
This is only cosmetic, but the entry point for rtld is not a leaf function,
and this avoids two .frame directives for rtld_start.
Sponsored by: DARPA / AFRL
Newer binutils supports extensions to the MIPS ABI for non-PIC code
that is used when compiling O32 binaries with clang 5 (but not used
for N64 oddly enough). These extensions require support for
R_MIPS_COPY relocations as well as a second PLT GOT using
R_MIPS_JUMP_SLOT relocations.
For R_MIPS_COPY, use the same approach as on other architectures where
fixups are deferred to the MD do_copy_relocations.
The additional PLT GOT for jump slots is located in a .got.plt section
which is identified by a DT_MIPS_PLTGOT dynamic entry. This GOT also
requires fixups for the first two GOT entries just as the normal GOT.
However, the entry point for this second GOT uses a different calling
convention. Rather than passing an offset into the GOT, it passes an
offset into the .rel.plt section. This requires a second entry point
(_rtld_pltbind_start) which calls the normal _rtld_bind() rather than
_mips_rtld_bind(). This also means providing a real version of
reloc_jmpslot() which is used by _rtld_bind().
In addition, add real implementions of reloc_plt() and
reloc_jmpslots() which walk .rel.plt handling R_MIPS_JUMP_SLOT
relocations.
Reviewed by: kib
Sponsored by: DARPA / AFRL
Differential Revision: https://reviews.freebsd.org/D12326
Fix warnings about redundant declarations in rtld
when libthr in increased to WARNS=6.
Reviewed by: kib
MFC after: 3 days
Sponsored by: Dell EMC
Differential Revision: https://reviews.freebsd.org/D10934
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
The MIPS ABI does not require the second GOT entry to be reserved for use
by the runtime linker as on other architectures. Instead, static linkers
use a special value in the second GOT entry to indicate if the entry is
reserved. This value is supposed to consist of an address with the MSB
set and the rest of the bits all zero which is an invalid user address.
However, the old binutils currently in the tree uses the 32-bit mask value
(2^31) on 64-bit MIPS instead of 2^63. This was fixed in upstream
binutils in 2008 to use 2^63 on 64-bit MIPS.
The first part of this change changes the runtime check in init_pltgot()
to check for both values (2^31 and 2^63) when deciding whether to store
the current object pointer in GOT[1] which fixes dynamic N64 binaries
compiled with modern binutils.
However, the initial version of this fix exposed another related bug in
that _rtld_relocate_nonplt_self() was only checking for the new value
(2^63) in GOT[1] and incorrectly treated GOT[1] as a local GOT entry
(and did not relocate the final local GOT entry). To handle this, fix
all of the places that check for GOT[1]'s status to use the same macro
that checks for both values on N64.
Reviewed by: kan, imp
Sponsored by: DARPA / AFRL
Differential Revision: https://reviews.freebsd.org/D9708
The duplicate call to store_ptr() was added in r204687, but it should
have no effect as it only stores an Elf_Sword and the later store_ptr()
does a write that is at least as large if not larger.
Reviewed by: jmallett
Obtained from: CheriBSD (sort of)
Sponsored by: DARPA / AFRL
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
(hopefully) stock gcc 4.2.1 on i386 and other arches.
In particular:
- Do not use %ebx in the asm constraints on i386, since rtld is
compiled with -fPIC and gcc cannot handle GOT-base register reload
(clang and newer gcc can).
- Avoid direct use of [static N] construct in the function
declaration/definion. In-tree gcc was patched to support this, but
stock 4.2.1 cannot handle the feature.
Requested by: bde
Sponsored by: The FreeBSD Foundation
MFC after: 1 week
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
=v is some ye olde gcc "use this specific register as the temp register"
thing that they've deprecated and clang/llvm doesn't implement.
Poked again and again by: sbruno
This work, originally from Stacey Son, uses the MIPS UserReg for
reading the TLS data, and will fall back to the normal syscall path
when it isn't supported.
This code dynamically patches cpu_switch() to bypass the UserReg
instruction so to avoid generating a machine exception.
Thanks to sson for the original work, and to Dan Nelson for
bringing it to date and testing it on MIPS32 with me.
Tested:
* mips64 (sson)
* mips74k (dnelson_1901@yahoo.com) - AR9344 SoC, UserReg support
* mips24k (adrian) - AR9331 SoC, no UserReg support
Obtained from: sson, dnelson_1901@yahoo.com
This is only an interim fix; MIPS should be using the MI code instead,
which does not have this issue.
Reviewed by: imp
Differential Revision: https://reviews.freebsd.org/D661
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
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
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
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
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)
r211130 in favor of this more general fix.
This fixes a compilation error for mips 64-bit little endian build.
libexec/rtld-elf/mips/reloc.c:196: warning: right shift count >= width of type
Suggested by: stefanf, jchandra, bde
libexec/rtld-elf/mips/reloc.c:196: warning: right shift count >= width of type
When the expression '(r_info) >> 32' was passed to bswap32() it was promptly
changed to '(uint32_t)(r_info) >> 32' which is not what we intended.
derivitive of NetBSD's mips_reloc.c, so pull in the copyright notice
from there.
Also, a minor tweak to load/store pointers. Other changes from NetBSD
likely would be useful too...
Obtained from: NetBSD
from working.
From p4 filelog of the upstream file in p4
//depot/projects/mips2-jnpr/src/libexec/rtld-elf/mips/reloc.c
... #6 change 140737 edit on 2008/04/27 by gonzo@gonzo_jeeves (text+ko)
o Looks like handler for R_MIPS_REL32 brought by CS 137942
is broken for tradmips. Code from NetBSD's
libexec/ld.elf_so/arch/mips/mips_reloc.c works just fine.
... #3 change 137942 edit on 2008/03/17 by rrs@rrs-mips2-jnpr (text+ko)
Any relocation symbol lookup if its 0. It looks like
this is the way the compiler indicates you need to
look in another shared library. When we hit these
as we relocate a object we will do the symbol
lookups and setup the relocation table with the
right value.
Submitted by: rrs@, gonzo@
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.
