David Kilroy [Wed, 12 Feb 2020 17:28:15 +0000 (14:28 -0300)]
elf: Allow dlopen of filter object to work [BZ #16272]
There are two fixes that are needed to be able to dlopen filter
objects. First _dl_map_object_deps cannot assume that map will be at
the beginning of l_searchlist.r_list[], as filtees are inserted before
map. Secondly dl_open_worker needs to ensure that filtees get
relocated.
In _dl_map_object_deps:
* avoiding removing relocation dependencies of map by setting
l_reserved to 0 and otherwise processing the rest of the search
list.
* ensure that map remains at the beginning of l_initfini - the list
of things that need initialisation (and destruction). Do this by
splitting the copy up. This may not be required, but matches the
initialization order without dlopen.
Modify dl_open_worker to relocate the objects in new->l_inifini.
new->l_initfini is constructed in _dl_map_object_deps, and lists the
objects that need initialization and destruction. Originally the list
of objects in new->l_next are relocated. All of these objects should
also be included in new->l_initfini (both lists are populated with
dependencies in _dl_map_object_deps). We can't use new->l_prev to pick
up filtees, as during a recursive dlopen from an interposed malloc
call, l->prev can contain objects that are not ready for relocation.
Add tests to verify that symbols resolve to the filtee implementation
when auxiliary and filter objects are used, both as a normal link and
when dlopen'd.
Joseph Myers [Wed, 12 Feb 2020 13:37:16 +0000 (13:37 +0000)]
Rename RWF_WRITE_LIFE_NOT_SET to RWH_WRITE_LIFE_NOT_SET following Linux 5.5.
Linux 5.5 renames RWF_WRITE_LIFE_NOT_SET to RWH_WRITE_LIFE_NOT_SET,
with the old name kept as an alias. This patch makes the
corresponding change in glibc.
Stefan Liebler [Wed, 12 Feb 2020 08:10:56 +0000 (09:10 +0100)]
S390: Fix non-ascii character in fenv.h.
The comment "isn't" contained a non-ascii character which leads to
an error if compiled with -finput-charset=ascii:
error: failure to convert ascii to UTF-8
This is observable in GCC testsuite:
FAIL: 17_intro/headers/c++1998/charset.cc (test for excess errors)
FAIL: 17_intro/headers/c++2011/charset.cc (test for excess errors)
FAIL: 17_intro/headers/c++2014/charset.cc (test for excess errors)
FAIL: 17_intro/headers/c++2017/charset.cc (test for excess errors)
FAIL: 17_intro/headers/c++2020/charset.cc (test for excess errors)
The code started out with bits form resolv/resolv_conf.c, but it
was enhanced to deal with directories and FIFOs in a more predictable
manner. A test case is included as well.
This will be used to implement the /etc/resolv.conf change detection.
This currently lives in a header file only. Once there are multiple
users, the implementations should be moved into C files.
Fangrui Song [Wed, 12 Feb 2020 06:10:19 +0000 (01:10 -0500)]
elf.h: Add R_RISCV_IRELATIVE
The number has been officially assigned by
https://github.com/riscv/riscv-elf-psabi-doc/pull/131
https://github.com/riscv/riscv-elf-psabi-doc/commit/d21ca40a7f56812a15e97450b7bc1599c0d35b82
Samuel Thibault [Mon, 10 Feb 2020 22:06:33 +0000 (23:06 +0100)]
hurd: Add __pthread_spin_wait and use it
900778283ac3 ("htl: make pthread_spin_lock really spin") made
pthread_spin_lock really spin and not block, but the current users of
__pthread_spin_lock were assuming that it blocks, i.e. they use it as a
lightweight mutex fitting in just one int.
Joseph Myers [Mon, 10 Feb 2020 22:17:59 +0000 (22:17 +0000)]
Use --disable-gdbserver in build-many-glibcs.py.
Now that binutils-gdb has gdbserver at top level, an extra
--disable-gdbserver configure option is needed when configuring
binutils from a git checkout to avoid it also building gdbserver
unnecessarily (although fairly harmlessly). This patch updates the
options used in build-many-glibcs.py accordingly (although this might
end up not being needed depending on what happens regarding whether
gdbserver gets built for host != target).
Tested with a build-many-glibcs.py compilers build for
aarch64-linux-gnu using binutils-gdb master.
Wilco Dijkstra [Mon, 10 Feb 2020 16:08:40 +0000 (16:08 +0000)]
Improve random memcpy benchmark
Improve the random memcpy benchmark. Double the number of copies and
increase the memory sizes tested to 512KB. Add a more detailed
distribution of memcpy alignment and sizes up to 4096 based on SPEC2017
traces.
Samuel Thibault [Sun, 9 Feb 2020 16:34:26 +0000 (16:34 +0000)]
htl: Do not put spin_lock inlines in public headers
They were not getting used anyway.
Also do not make libsupport use them, it would make tests using it have
to be made to link against libmachuser for gsync_wait.
Florian Weimer [Sun, 9 Feb 2020 10:01:39 +0000 (11:01 +0100)]
x86: Remove <bits/select.h> and use the generic version
Particularly on CPUs without ERMS, the string instructions are slow,
so it is unclear whether this architecture-specific implementation is
in fact an optimization.
Florian Weimer [Sun, 9 Feb 2020 10:51:08 +0000 (11:51 +0100)]
Linux: Add io/tst-o_path-locks test
The O_PATH-based fchmodat emulation will rely on the fact that closing
an O_PATH descriptor never releases POSIX advisory locks, so this
commit adds a test case for this behavior.
Florian Weimer [Fri, 7 Feb 2020 17:48:04 +0000 (18:48 +0100)]
resolv: Fix CNAME chaining in resolv/tst-resolv-ai_idn-common.c
The second CNAME record optionally generated by the response function
used the question name, not the redirected name from the first CNAME.
This breaks the chain and results in failures of these IDNA tests if
CNAME owner names are checked as expected (which the current
implementation does not do).
Lukasz Majewski [Thu, 16 Jan 2020 22:14:27 +0000 (23:14 +0100)]
y2038: linux: Provide __settimeofday64 implementation
This patch provides new __settimeofday64 explicit 64 bit function for setting
64 bit time in the kernel (by internally calling __clock_settime64).
Moreover, a 32 bit version - __settimeofday has been refactored to internally
use __settimeofday64.
The __settimeofday is now supposed to be used on systems still supporting 32
bit time (__TIMESIZE != 64) - hence the necessary conversion of struct
timeval to 64 bit struct __timespec64.
Internally the settimeofday uses __settimeofday64. This patch is necessary
for having architectures with __WORDSIZE == 32 Y2038 safe.
Run-time tests:
- Run specific tests on ARM/x86 32bit systems (qemu):
https://github.com/lmajewski/meta-y2038 and run tests:
https://github.com/lmajewski/y2038-tests/commits/master
Above tests were performed with Y2038 redirection applied as well as without
to test proper usage of both __settimeofday64 and __settimeofday.
Lukasz Majewski [Thu, 16 Jan 2020 22:44:59 +0000 (23:44 +0100)]
y2038: Provide conversion helpers for struct __timeval64
Those functions allow easy conversion between Y2038 safe, glibc internal
struct __timeval64 and other time related data structures (like struct timeval
or struct __timespec64).
Lukasz Majewski [Fri, 17 Jan 2020 14:29:56 +0000 (15:29 +0100)]
y2038: alpha: Rename valid_timeval_to_timeval64 to valid_timeval32_to_timeval
Without this patch the naming convention for functions to convert
struct timeval32 to struct timeval (which supports 64 bit time on Alpha) was
a bit misleading. The name 'valid_timeval_to_timeval64' suggest conversion
of struct timeval to struct __timeval64 (as in ./include/time.h).
As on alpha the struct timeval supports 64 bit time it seems more readable
to emphasis struct timeval32 in the conversion function name.
Hence the helper function naming change to 'valid_timeval32_to_timeval'.
Lukasz Majewski [Thu, 16 Jan 2020 22:40:51 +0000 (23:40 +0100)]
y2038: Introduce struct __timeval64 - new internal glibc type
This type is a glibc's "internal" type similar to struct timeval but
whose tv_sec field is a __time64_t rather than a time_t, which makes it
Y2038-proof. This struct is NOT supposed to be passed to the kernel -
instead it shall be converted to struct __timespec64 and clock_[sg]ettime
syscalls shall be used (which are now Y2038 safe).
Lukasz Majewski [Sat, 25 Jan 2020 21:33:41 +0000 (22:33 +0100)]
y2038: Define __suseconds64_t type to be used with struct __timeval64
The __suseconds64_t type is supposed to be the 64 bit type across all
architectures.
It would be mostly used internally in the glibc - however, when passed to
Linux kernel (very unlikely), if necessary, it shall be converted to 32
bit type (i.e. __suseconds_t)
Joseph Myers [Fri, 7 Feb 2020 13:55:29 +0000 (13:55 +0000)]
Update kernel version to 5.5 in tst-mman-consts.py.
This patch updates the kernel version in the test tst-mman-consts.py
to 5.5. (There are no new constants covered by this test in 5.5 that
need any other header changes.)
Joseph Myers [Fri, 7 Feb 2020 13:54:58 +0000 (13:54 +0000)]
Update syscall lists for Linux 5.5.
Linux 5.5 has no new syscalls to add to syscall-names.list, but it
does newly enable the clone3 syscall for AArch64. This patch updates
the kernel version listed in syscall-names.list and regenerates the
AArch64 arch-syscall.h.
Lukasz Majewski [Wed, 15 Jan 2020 09:14:23 +0000 (10:14 +0100)]
y2038: linux: Provide __timespec_get64 implementation
This patch provides new instance of Linux specific timespec_get.c file placed
in ./sysdeps/unix/sysv/linux/.
When compared to this file version from ./time directory, it provides
__timespec_get64 explicit 64 bit function for getting 64 bit time in the
struct __timespec64 (for compilation using C11 standard).
Moreover, a 32 bit version - __timespec_get internally uses
__timespec_get64.
The __timespec_get is now supposed to be used on systems still supporting 32
bit time (__TIMESIZE != 64) - hence the necessary conversion to 32 bit struct
timespec.
Internally the timespec_get uses __clock_gettime64. This patch is necessary
for having architectures with __WORDSIZE == 32 Y2038 safe.
Run-time tests:
- Run specific tests on ARM/x86 32bit systems (qemu):
https://github.com/lmajewski/meta-y2038 and run tests:
https://github.com/lmajewski/y2038-tests/commits/master
Above tests were performed with Y2038 redirection applied as well as without
to test proper usage of both __timespec_get64 and __timespec_get.
DJ Delorie [Mon, 3 Feb 2020 19:57:23 +0000 (14:57 -0500)]
Run nptl/tst-pthread-getattr in a container
See https://bugzilla.redhat.com/show_bug.cgi?id=1653942
This test depends on the kernel's assignment of memory regions, but
running under ld.so explicitly changes those assignments, sometimes
sufficiently to cause the test to fail (esp with address space
randomization).
The easiest way to "fix" the test, is to run it the way the user would
- without ld.so. Running it in a container does that.
DJ Delorie [Mon, 3 Feb 2020 19:49:25 +0000 (14:49 -0500)]
test-container: add exec, cwd
exec <path_to_test_binary> [optional_argv_0]
copies test binary to specified location and runs it from
there. If the second argument is provided, that will
be used for argv[0]
cwd <directory>
attempts to chdir(directory) before running test
Note: "cwd" not "cd" as it takes effect just before the
test binary runs, not when it's encountered in the script,
so it can't be used as a path shortcut like "cd" would imply.
Andreas Schwab [Mon, 27 Jan 2020 09:05:53 +0000 (10:05 +0100)]
rt: avoid PLT setup in timer_[sg]ettime
The functions __timer_gettime64 and __timer_settime64 live in librt, not
libc. Use proper hidden aliases so that the callers do not need to set up
the PLT register.
Fixes commits cae1635a70 ("y2038: linux: Provide __timer_settime64
implementation") and 562cdc19c7 ("y2038: linux: Provide __timer_gettime64
implementation").
Lukasz Majewski [Mon, 6 Jan 2020 12:52:00 +0000 (13:52 +0100)]
y2038: linux: Provide __sched_rr_get_interval64 implementation
This patch replaces auto generated wrapper (as described in
sysdeps/unix/sysv/linux/syscalls.list) for sched_rr_get_interval with one which
adds extra support for reading 64 bit time values on machines with
__TIMESIZE != 64.
There is no functional change for architectures already supporting 64 bit
time ABI.
The sched_rr_get_interval declaration in ./include/sched.h is not followed by
corresponding libc_hidden_proto(), so it has been assumed that newly introduced
syscall wrapper doesn't require libc_hidden_def() (which has been added by
template used with auto generation script).
Moreover, the code for building sched_rr_gi.c file is already placed in
./posix/Makefiles, so there was no need to add it elsewhere.
Performed tests and validation are the same as for timer_gettime() conversion
(sysdeps/unix/sysv/linux/timer_gettime.c).
Lukasz Majewski [Fri, 6 Dec 2019 15:20:33 +0000 (16:20 +0100)]
y2038: linux: Provide __timerfd_settime64 implementation
This patch replaces auto generated wrapper (as described in
sysdeps/unix/sysv/linux/syscalls.list) for timerfd_settime with one which
adds extra support for reading and writing from Linux kernel 64 bit time
values on machines with __TIMESIZE != 64.
There is no functional change for archs already supporting 64 bit time ABI.
This patch is conceptually identical to timer_settime conversion already
done in sysdeps/unix/sysv/linux/timer_settime.c.
Please refer to corresponding commit message for detailed description of
introduced functions and the testing procedure.
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
---
Changes for v4:
- Update date from 2019 to 2020
Changes for v3:
- Add missing libc_hidden_def()
Changes for v2:
- Remove "Contributed by" from the file header
- Remove early check for (fd < 0) in __timerfd_settime64 as the fd
correctness check is already done in Linux kernel
- Add single descriptive comment line to provide concise explanation
of the code
Lukasz Majewski [Fri, 6 Dec 2019 09:41:50 +0000 (10:41 +0100)]
y2038: linux: Provide __timerfd_gettime64 implementation
This patch replaces auto generated wrapper (as described in
sysdeps/unix/sysv/linux/syscalls.list) for timerfd_gettime with one which
adds extra support for reading 64 bit time values on machines with
__TIMESIZE != 64.
There is no functional change for architectures already supporting 64 bit
time ABI.
This patch is conceptually identical to timer_gettime conversion already
done in sysdeps/unix/sysv/linux/timer_gettime.c.
Please refer to corresponding commit message for detailed description of
introduced functions and the testing procedure.
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
---
Changes for v4:
- Update date from 2019 to 2020
Changes for v3:
- Add missing libc_hidden_def()
Changes for v2:
- Remove "Contributed by" from the file header
- Remove early check for (fd < 0) in __timerfd_gettime64 as the fd
correctness check is already done in Linux kernel
- Add single descriptive comment line to provide concise explanation
of the code
H.J. Lu [Sat, 1 Feb 2020 13:44:55 +0000 (05:44 -0800)]
i386: Don't unnecessarily save and restore EAX, ECX and EDX [BZ# 25262]
On i386, since EAX, ECX and EDX are caller-saved, there are no need
to save and restore EAX, ECX and EDX in getcontext, setcontext and
swapcontext. They just need to clear EAX on success. The extra
scratch registers are needed to enable CET.