| Age | Commit message (Collapse) | Author |
|---|
|
- Cleanup dl-vdso.c code.
- Pass `void *` as first arg to `load_vdso()`, using 32-bit type
is completely wrong on 64bit architectures.
- Split libc code and vDSO-related code.
Move arch-specific implementations into separate files.
The performance improvement is for example 50-60 times on ARMv7
and about 4 times on x86_64.
Signed-off-by: Dmitry Chestnykh <dm.chestnykh@gmail.com>
|
|
The Linux kernels ELF-FDPIC binfmt program loader can support loading and
running conventional ELF format binaries on noMMU kernels when compiled
appropriately. That is when they are constant displacement binaries such
as generated using the -pie compile option.
Add a configure option to allow selecting ELF binary support in noMMU
mode configurations on architectures that support this. The main
requirement is to generate the ldso run-time loader to perform relocation
at load time. These configurations do not support shared libraries, so
there is no need to generate a full shared library, only the static
version is required.
The use of ELF format binaries does mean a slightly simpler toolchain
generation (does not require a -uclinux- for some architectures) and does
not require an extra tool like elf2flt.
This initial support targets M68K, ARM and RISC-V architectures. No kernel
changes are required, the required support for this is already in mainline
kernels (certainly as of linux-6.6).
Note that for the M68K and ARM architectures that the initialized
registers and stack layout at process startup is slightly different for
the flat format loader and the ELF/ELF-FDPIC loaders. So we need some
changes to the startup code (crt1.S) for them.
I have not done extensive testing outside of M68K, ARM and RISC-V.
I had to make changes to a couple of the dl-startup.h architecture files
to get them to build for this noMMU case. I did not dig down too deep on
the reasons, but they still seem ok for the MMU case as well.
Signed-off-by: Greg Ungerer <gerg@linux-m68k.org>
|
|
|
|
|
|
Sync with GNU C library and consolidate duplicate non
architecture specific defines.
MAP_UNINITIALIZED is only defined to 0x4000000 and used by
the Linux kernel when CONFIG_MMAP_ALLOW_UNINITIALIZED is enabled.
CONFIG_MMAP_ALLOW_UNINITIALIZED is only available for nommu.
See Documentation/nommu-mmap.txt.
|
|
|
|
Commit e3dec33ba (powerpc: Add TLS and NPTL support) introduced a duplicate of
_dl_add_to_slotinfo and _dl_initial_error_catch_tsd declarations.
Signed-off-by: Baruch Siach <baruch@tkos.co.il>
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>
|
|
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>
|
|
Building with NPTL enabled and shared library support disabled we hit:
In file included from libpthread/nptl/sysdeps/generic/dl-tls.c:30:0:
./ldso/include/dl-elf.h: In function '__dl_parse_dynamic_info':
./ldso/include/dl-elf.h:173:20: error: 'RTLD_NODELETE' undeclared (first use in this function)
./ldso/include/dl-elf.h:173:20: note: each undeclared identifier is reported only once for each function it appears in
make: *** [libpthread/nptl/sysdeps/generic/dl-tls.os] Error 1
A previous commit (f26c5f6952ce9bf8edec9c1571c47addb1bcc442) touched
on a similar issue, but added the include to the incorrect location.
Reported-by: Christophe Lyon <christophe.lyon@st.com> [arm nommu]
Reported-by: Daniel Beecham <daniel@lunix.se> [static x86_64]
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
|
|
Add support for %p to _dl_dprintf for later corrections in debug messages.
Disable _dl_debug_file if LD debugging is not enabled and change it's use to 2.
The use of dprintf in elfinterp.c will spit out warnings, will be fixed in the
reworked ldso.
Signed-off-by: Peter S. Mazinger <ps.m@gmx.net>
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>
|
|
Signed-off-by: Peter S. Mazinger <ps.m@gmx.net>
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>
|
|
The existing DSBT support relies on the kernel to provide DSBT info
as part of the load maps passed to user space. The problem with this
approach is that the DSBT info is in the dynamic section, so the
kernel must access a userspace mapping of the dynamic section (or
separately read a copy for the kernel) in order to retrieve the
information needed by userspace.
This patch reworks the DSBT support to remove the reliance on DSBT
info coming from the kernel. Instead, ldso reads the info itself from
the dynamic section. One other benefit of this is that it allows the
existing kernel FDPIC loader to also load DSBT binaries.
Signed-off-by: Mark Salter <msalter@redhat.com>
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>
|
|
The setup logic is duplicated, so unify it in a local func.
Mark the variable const while we're doing this, and add missing
ifdef protection to the header that declares it availability.
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
|
|
The former is part of a standard (C99) while the latter is not.
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
|
|
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>
|
|
* origin/master: (32 commits)
libubacktrace: fix backtrace support on arm-eabi, which needs libgcc_eh linked too
getaddrinfo.c: fix incorrect check for ERANGE from gethostbyaddr_r
getaddrinfo.c: improve code readability. No functional changes
string: remove unused variable
x86_64: silence warning if !TLS
buildsys: prettify ssp.c handling
madvise is LINUX_SPECIFIC
test_nptl: fix expected result for tst-cputimer[123]
test_nptl: fix expected result for tst-clock2 test
buildsys: make $(LOCAL_INSTALL_PATH) phony
ether_aton: reject invalid input
tests: disable ether tests if !HAS_SOCKET
inet: add ether_aton testcase
sysconf: clock_getres depends on HAS_REALTIME
__rt_sigwaitinfo: depends on HAS_REALTIME
buildsys: minor fixes in Makefile.arch for C6X
buildsys: minor fixes in Makefile.arch for microblaze
libubacktrace: enabled for all archs indeed.
sparc: don't access fp registers when configured for no fpu
libubacktrace: generic implementation based dwarf
...
Conflicts:
ldso/ldso/dl-elf.c
ldso/ldso/mips/elfinterp.c
ldso/ldso/x86_64/elfinterp.c
Signed-off-by: Carmelo Amoroso <carmelo.amoroso@st.com>
|
|
We have a problem with DT_TEXTREL shared libraries on nommu machines.
The dynamic linker's strategy is to map the text segment read-only
first, then look for DT_TEXTREL, and use mprotect to change protections
if necessary. This fails on nommu, since a nommu kernel can decide to
share the memory for private read-only file mappings, and mprotect
doesn't (can't) do anything about this sharing. Existing nommu targets
apparently have no need for this, but on C6X, we may need to assign
library indices at run-time if no --dsbt-index option was passed to the
linker at build time.
Hence, the following patch, which instead of using mprotect, redoes the
mapping with PF_W set.
Signed-off-by: Bernd Schmidt <bernds@codesourcery.com>
|
|
Added support for the following tracing capabilities:
- LD_WARN to warn about undefined symbols during the lookup stage.
- LD_TRACE_PRELINKING to trace the needed libraries of the object
that we are prelinking.
Signed-off-by: Filippo Arcidiacono <filippo.arcidiacono@st.com>
Signed-off-by: Carmelo Amoroso <carmelo.amoroso@st.com>
|
|
Global symbol scope is implemented as a linked list of
local scope, that dynamically grows and shrinks when dlopen/
dlclose are called. Each local scope is implemented as an array
of pointer to struct elf_resolve.
This will help to detect conflict when LD_TRACE_PRELINKING option
will be implemented.
Signed-off-by: Filippo Arcidiacono <filippo.arcidiacono@st.com>
Signed-off-by: Carmelo Amoroso <carmelo.amoroso@st.com>
|
|
The dynamic linker can be run either indirectly through running some
dynamically linked program or library (in which case no command line options
to the dynamic linker can be passed and, in the ELF case, the dynamic linker
which is stored in the .interp section of the program is executed)
or directly by running:
/lib/ld-uClibc.so.* [OPTIONS] [PROGRAM [ARGUMENTS]]
Stand-alone execution is a prerequisite for adding prelink capabilities
to uClibc dynamic linker, as well useful for testing an updated version
of the dynamic linker without breaking the whole system.
Currently supported option:
--library-path PATH use given PATH instead of content of the environment
variable LD_LIBRARY_PATH
(Mandatory for prelinking)
Not supported options:
--list list all dependencies and how they are resolved
--verify verify that given object really is a dynamically linked
object we can handle
--inhibit-rpath LIST ignore RUNPATH and RPATH information in object names
in LIST
This feature can be enabled by setting LDSO_STANDALONE_SUPPORT=y
Signed-off-by: Filippo Arcidiacono <filippo.arcidiacono@st.com>
Signed-off-by: Carmelo Amoroso <carmelo.amoroso@st.com>
|
|
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>
|
|
Signed-off-by: Austin Foxley <austinf@cetoncorp.com>
|
|
this was causing tls symbols to leak into non-tls builds
Signed-off-by: Austin Foxley <austinf@cetoncorp.com>
|
|
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>
|
|
The TLS merge 534661b91c9849 introduced multiple style problems as well as
random breakage:
- missing _dl_free
- incomplete parametrization of _dl_lookup_hash
- restore FDPIC handling in _dl_lookup_hash
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
|
|
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
|
|
Signed-off-by: Austin Foxley <austinf@cetoncorp.com>
|
|
If an arch (like alpha and others) have no ldso support at all, then there
is no point in attempting to walk loaded modules in the dl-* helper funcs.
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
|
|
SIgned-off-by: Carmelo Amoroso <carmelo.amoroso@st.com>
|
|
text data bss dec hex filename
- 16709 240 92 17041 4291 lib/ld-uClibc.so
+ 16634 236 92 16962 4242 lib/ld-uClibc.so
- 4602 344 4 4950 1356 lib/libdl-0.9.30-svn.so
+ 4571 328 4 4903 1327 lib/libdl-0.9.30-svn.so
- 4602 344 4 4950 1356 lib/libdl.so
+ 4571 328 4 4903 1327 lib/libdl.so
|
|
Some powerpc machines can support 64k pages, enabled by the
CONFIG_64K_PAGES option in linux.
However, the uClibc dynamic loader won't currently work on these
machines, as it uses hard-coded values (PAGE_ALIGN, ADDR_ALIGN and
OFFS_ALIGN) in the ldso architecture-specific headers. When running on
a kernel with 64k pages, ld.so tries to mmap with 4k-aligned addresses,
rather than 64k, so mmap fails with -EINVAL.
When booting a 64k machine with a uClibc dynamic linker, init fails
with:
/init:500: can't map '/lib/libc.so.0'
/init:500: can't map '/lib/libc.so.0'
/init:500: can't map '/lib/libc.so.0'
/init: can't load library 'libc.so.0'
Kernel panic - not syncing: Attempted to kill init!
This change allows ld.so determine these alignment masks at runtime,
rather than compile-time. Since we have the _dl_pagesize variable
available, we can use that to generate the appropriate masks.
Since almost all of the architectures can use the common definitions for
the _ALIGN macros, we can consolidate them all in ldso.h, and override
in the sysdep headers where necessary (ie, mips).
This allows me to start a uClibc-based root fs on a 64k machine.
Signed-off-by: Jeremy Kerr <jk at ozlabs org>
|
|
|
|
|
|
exists, and move some definitions to their proper place.
|
|
A couple more target macros for ld.so to deal with FD-PIC support. We need
special code to compute the initial got and dpnt, and we need to pass extra
arguments to _dl_get_ready_to_run.
|
|
Add a new function _dl_free. In _dl_malloc, ensure we always get back a full
page from mmap.
Reset _dl_malloc_function and _dl_free_function when libdl is initialized.
|
|
Mazinger
|
|
|
|
be 0 in this case. Avoids some warnings).
|
|
|
|
|
|
|
|
|
|
need the defines from the former.
|
|
|
|
|
|
|
|
|
|
|
|
bitsize
|
