This patch adds a way to close a range of file descriptors on
posix_spawn as a new file action. The API is similar to the one
provided by Solaris 11 [1], where the file action causes the all open
file descriptors greater than or equal to input on to be closed when
the new process is spawned.
The function posix_spawn_file_actions_addclosefrom_np is safe to be
implemented by iterating over /proc/self/fd, since the Linux spawni.c
helper process does not use CLONE_FILES, so its has own file descriptor
table and any failure (in /proc operation) aborts the process creation
and returns an error to the caller.
I am aware that this file action might be redundant to the current
approach of POSIX in promoting O_CLOEXEC in more interfaces. However
O_CLOEXEC is still not the default and for some specific usages, the
caller needs to close all possible file descriptors to avoid them
leaking. Some examples are CPython (discussed in BZ#10353) and OpenJDK
jspawnhelper [2] (where OpenJDK spawns a helper process to exactly
closes all file descriptors). Most likely any environment which calls
functions that might open file descriptor under the hood and aim to use
posix_spawn might face the same requirement.
Checked on x86_64-linux-gnu and i686-linux-gnu on kernel 5.11 and 4.15.
[1] https://docs.oracle.com/cd/E36784_01/html/E36874/posix-spawn-file-actions-addclosefrom-np-3c.html
[2] https://github.com/openjdk/jdk/blob/master/src/java.base/unix/native/libjava/childproc.c#L82
The function closes all open file descriptors greater than or equal to
input argument. Negative values are clamped to 0, i.e, it will close
all file descriptors.
As indicated by the bug report, this is a common symbol provided by
different systems (Solaris, OpenBSD, NetBSD, FreeBSD) and, although
its has inherent issues with not taking in consideration internal libc
file descriptors (such as syslog), this is also a common feature used
in multiple projects [1][2][3][4][5].
The Linux fallback implementation iterates over /proc and close all
file descriptors sequentially. Although it was raised the questioning
whether getdents on /proc/self/fd might return disjointed entries
when file descriptor are closed; it does not seems the case on my
testing on multiple kernel (v4.18, v5.4, v5.9) and the same strategy
is used on different projects [1][2][3][5].
Also, the interface is set a fail-safe meaning that a failure in the
fallback results in a process abort.
Checked on x86_64-linux-gnu and i686-linux-gnu on kernel 5.11 and 4.15.
[1] 5238e95759/src/basic/fd-util.c (L217)
[2] ddf4b77e11/src/lxc/start.c (L236)
[3] 9e4f2f3a6b/Modules/_posixsubprocess.c (L220)
[4] 5f47c0613e/src/libstd/sys/unix/process2.rs (L303-L308)
[5] https://github.com/openjdk/jdk/blob/master/src/java.base/unix/native/libjava/childproc.c#L82
It was added on Linux 5.9 (278a5fbaed89) with CLOSE_RANGE_CLOEXEC
added on 5.11 (582f1fb6b721f). Although FreeBSD has added the same
syscall, this only adds the symbol on Linux ports. This syscall is
required to provided a fail-safe way to implement the closefrom
symbol (BZ #10353).
Checked on x86_64-linux-gnu and i686-linux-gnu on kernel 5.11 and 4.15.
The code to allocate a stack from xsigstack is refactored so it can
be more generic. The new support_stack_alloc() also set PROT_EXEC
if DEFAULT_STACK_PERMS has PF_X. This is required on some
architectures (hppa for instance) and trying to access the rtld
global from testsuite will require more intrusive refactoring
in the ldsodefs.h header.
Checked on x86_64-linux-gnu and i686-linux-gnu. I also ran
tst-xsigstack on both hppa and ia64.
_int_realloc is correctly declared at the top to be static, but
incorrectly defined without the static keyword. Fix that. The
generated binaries have identical code.
Both tests try to dlopen libm.so at runtime, so make them depend on it
so that they're executed if libm.so has been updated.
Reviewed-by: Florian Weimer <fweimer@redhat.com>
The tcache allocator layer uses the tcache pointer as a key to
identify a block that may be freed twice. Since this is in the
application data area, an attacker exploiting a use-after-free could
potentially get access to the entire tcache structure through this
key. A detailed write-up was provided by Awarau here:
https://awaraucom.wordpress.com/2020/07/19/house-of-io-remastered/
Replace this static pointer use for key checking with one that is
generated at malloc initialization. The first attempt is through
getrandom with a fallback to random_bits(), which is a simple
pseudo-random number generator based on the clock. The fallback ought
to be sufficient since the goal of the randomness is only to make the
key arbitrary enough that it is very unlikely to collide with user
data.
Co-authored-by: Eyal Itkin <eyalit@checkpoint.com>
This partially fixes static-only NSS support (bug 27959): The files
module no longer needs dlopen. Support for the dns module remains
to be added, and also support for disabling dlopen altogether.
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
This is only needed if nss_files is loaded by nscd.
Before:
text data bss dec hex filename
767 0 24952 25719 6477 nss/files-init.os
After:
text data bss dec hex filename
666 0 0 666 29a nss/files-init.os
Using PATH_MAX bytes unconditionally for the directory name
is wasteful, but fixing that would constitute another break
of this semi-public ABI. (The other issue is that with
symbolic links, an arbitrary set of parent directories may need
watching, not just a single one.)
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
This patch updates the kernel version in the test tst-mman-consts.py
to 5.13. (There are no new MAP_* constants covered by this test in
5.13 that need any other header changes.)
Tested with build-many-glibcs.py.
It's tst-realloc, not tst-posix-realloc. Verified this time to ensure
that the total number of tests reduced by 1.
Reported-by: Stefan Liebler <stli@linux.ibm.com>
They are no longer needed after everything has been moved into
libc. The _dl_vsym test has to be removed because the symbol
cannot be used outside libc anymore.
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
The remaining symbols are mostly used by libthread_db.
__pthread_get_minstack has to remain exported even though unused.
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
Now that there are no internal users anymore, these new symbol
versions can be removed from the public ABI. The compatibility
symbols remain.
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
This avoids an ABI hazard (types changing between different modules
of glibc) without introducing linknamespace issues. In particular,
NSS modules now call __lll_lock_wait_private@@GLIBC_PRIVATE to wait
on internal locks (the unlock path is inlined and performs a direct
system call).
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
The tunable will not work with *any* non-zero tunable value since its
list of allowed values is 0-3. Fix the documentation to reflect that.
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
The realloc (NULL, 0) test in tst-realloc fails with gcc 7.x but
passes with newer gcc. This is because a newer gcc transforms the
realloc call to malloc (0), thus masking the bug in mcheck.
Disable the test with mcheck for now. The malloc removal patchset
will fix this and then remove this test from the exclusion list.
Reported-by: Stefan Liebler <stli@linux.ibm.com>
Reviewed-by: Carlos O'Donell <carlos@redhat.com>
This patch consolidates the setsockopt implementation on
sysdeps/unix/sysv/linux/getsockopt.c. The changes are:
1. Remove it from auto-generation syscalls.list on all architectures.
2. Add __ASSUME_SETSOCKOPT_SYSCALL as default and undef if for
specific kernel versions on some architectures.
This also fix a potential issue where 32-bit time_t ABI should use the
linux setsockopt which overrides the underlying SO_* constants used for
socket timestamping for _TIME_BITS=64.
Checked on x86_64-linux-gnu and i686-linux-gnu.
This patch consolidates the getsockopt Linux syscall implementation on
sysdeps/unix/sysv/linux/getsockopt.c. The changes are:
1. Remove it from auto-generation syscalls.list on all architectures.
2. Add __ASSUME_GETSOCKOPT_SYSCALL as default and undef if for
specific kernel versions on some architectures.
This also fix a potential issue where 32-bit time_t ABI should use the
linux getsockopt which overrides the underlying SO_* constants used for
socket timestamping for _TIME_BITS=64.
Checked on x86_64-linux-gnu and i686-linux-gnu.
In dlerror_run, free corresponds to the local malloc in the
namespace, but GLRO (dl_catch_error) uses the malloc from the base
namespace. elf/tst-dlmopen-gethostbyname triggers this mismatch,
but it does not crash, presumably because of a fastbin deallocation.
Fixes commit c2059edce2 ("elf: Use
_dl_catch_error from base namespace in dl-libc.c [BZ #27646]") and
commit b2964eb1d9 ("dlfcn: Failures
after dlmopen should not terminate process [BZ #24772]").
This avoids crashes in libc when cmsg is null and refrencing msg
structure when it is null
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
Separate the malloc check implementation from the malloc hooks. They
still use the hooks but are now maintained in a separate file.
Reviewed-by: Carlos O'Donell <carlos@redhat.com>
The variable and function pair appear to provide a way for users to
set conditional breakpoints in mtrace when a specific address is
returned by the allocator. This can be achieved by using conditional
breakpoints in gdb so it is redundant. There is no documentation of
this interface in the manual either, so it appears to have been a hack
that got added to debug malloc. Deprecate these symbols and do not
call tr_break anymore.
Reviewed-by: DJ Delorie <dj@redhat.com>
Reviewed-by: Carlos O'Donell <carlos@redhat.com>
Dependencies on hooks.c and arena.c get auto-computed when generating
malloc.o{,s}.d so there is no need to add them manually.
Reviewed-by: Carlos O'Donell <carlos@redhat.com>
Reviewed-by: Andreas Schwab <schwab@linux-m68k.org>
After commit 1e26d35193 ("malloc: Fix
tcache leak after thread destruction [BZ #22111]"),
tcache_shutting_down is still not early enough. When we detach a
thread with no tcache allocated, tcache_shutting_down would still be
false.
Reviewed-by: DJ Delorie <dj@redhat.com>
Like malloc-check, add generic rules to run all tests in malloc by
linking with libmcheck.a so as to provide coverage for mcheck().
Currently the following 12 tests fail:
FAIL: malloc/tst-malloc-backtrace-mcheck
FAIL: malloc/tst-malloc-fork-deadlock-mcheck
FAIL: malloc/tst-malloc-stats-cancellation-mcheck
FAIL: malloc/tst-malloc-tcache-leak-mcheck
FAIL: malloc/tst-malloc-thread-exit-mcheck
FAIL: malloc/tst-malloc-thread-fail-mcheck
FAIL: malloc/tst-malloc-usable-static-mcheck
FAIL: malloc/tst-malloc-usable-static-tunables-mcheck
FAIL: malloc/tst-malloc-usable-tunables-mcheck
FAIL: malloc/tst-malloc_info-mcheck
FAIL: malloc/tst-memalign-mcheck
FAIL: malloc/tst-posix_memalign-mcheck
and they have been added to tests-exclude-mcheck for now to keep
status quo. At least the last two can be attributed to bugs in
mcheck() but I haven't fixed them here since they should be fixed by
removing malloc hooks. Others need to be triaged to check if they're
due to mcheck bugs or due to actual bugs.
Reviewed-by: DJ Delorie <dj@redhat.com>
Build of iconvconfig failed with CFLAGS=-Os since __feof_unlocked is
not a public symbol. Replace with feof_unlocked (defined to
__feof_unlocked when IS_IN (libc)) to fix this.
Reported-by: Szabolcs Nagy <szabolcs.nagy@arm.com>
Reviewed-by: Szabolcs Nagy <szabolcs.nagy@arm.com>
The symbols gai_cancel, gai_error, gai_suspend, getaddrinfo_a,
__gai_suspend_time64 were moved using scripts/move-symbol-to-libc.py.
For Hurd (which remains !PTHREAD_IN_LIBC), a few #define redirects
had to be added because several pthread functions are not available
under __. (Linux uses __ prefixes for most hidden aliases, and has
to in some cases to avoid linknamespace issues.)
This patch modifies the current POWER9 implementation of strcpy and
stpcpy to optimize it for POWER9/10.
Since no new POWER10 instructions are used, the original POWER9 strcpy is
modified instead of creating a new implementation for POWER10. This
implementation is based on both the original POWER9 implementation of
strcpy and the preamble of the new POWER10 implementation of strlen.
The changes also affect stpcpy, which uses the same implementation with
some additional code before returning.
On POWER9, averaging improvements across the benchmark
inputs (length/source alignment/destination alignment), for an
experiment that ran the benchmark five times, bench-strcpy showed an
improvement of 5.23%, and bench-stpcpy showed an improvement of 6.59%.
On POWER10, bench-strcpy showed 13.16%, and bench-stpcpy showed 13.59%.
The changes are:
1. Removed the null string optimization.
Although this results in a few extra cycles for the null string, in
combination with the second change, this resulted in improvements for
for other cases.
2. Adapted the preamble from strlen for POWER10.
This is the part of the function that handles up to the first 16 bytes
of the string.
3. Increased number of unrolled iterations in the main loop to 6.
Reviewed-by: Matheus Castanho <msc@linux.ibm.com>
Tested-by: Matheus Castanho <msc@linux.ibm.com>
From
https://www.intel.com/content/www/us/en/support/articles/000059422/processors.html
* Intel TSX will be disabled by default.
* The processor will force abort all Restricted Transactional Memory (RTM)
transactions by default.
* A new CPUID bit CPUID.07H.0H.EDX[11](RTM_ALWAYS_ABORT) will be enumerated,
which is set to indicate to updated software that the loaded microcode is
forcing RTM abort.
* On processors that enumerate support for RTM, the CPUID enumeration bits
for Intel TSX (CPUID.07H.0H.EBX[11] and CPUID.07H.0H.EBX[4]) continue to
be set by default after microcode update.
* Workloads that were benefited from Intel TSX might experience a change
in performance.
* System software may use a new bit in Model-Specific Register (MSR) 0x10F
TSX_FORCE_ABORT[TSX_CPUID_CLEAR] functionality to clear the Hardware Lock
Elision (HLE) and RTM bits to indicate to software that Intel TSX is
disabled.
1. Add RTM_ALWAYS_ABORT to CPUID features.
2. Set RTM usable only if RTM_ALWAYS_ABORT isn't set. This skips the
string/tst-memchr-rtm etc. testcases on the affected processors, which
always fail after a microcde update.
3. Check RTM feature, instead of usability, against /proc/cpuinfo.
This fixes BZ #28033.
Linux 5.13 has three new syscalls (landlock_create_ruleset,
landlock_add_rule, landlock_restrict_self). Update syscall-names.list
and regenerate the arch-syscall.h headers with build-many-glibcs.py
update-syscalls.
Tested with build-many-glibcs.py.
On s390 (31bit), the pointer to the first byte after s always wraps
around with n >= 0x80000000 and can lead to stop searching before
end of s.
Thus this patch just use NULL as byte after s in this case and
the srst instruction stops searching with "not found" when wrapping
around from top address to zero.
This is observable with testcase string/test-memchr
starting with commit "String: Add overflow tests for strnlen, memchr,
and strncat [BZ #27974]"
https://sourceware.org/git/?p=glibc.git;a=commit;h=da5a6fba0febbfc90896ce1b2eb75c6d8a88a72d
Starting with recent commit 84f7ce8447
"posix: Add glob64 with 64-bit time_t support", elf/check-localplt
fails due to extra PLT reference __glob64_time64 in __glob64_time64
itself.
This is observable with gcc 7.5 on x86_64 with -m32 or s390x with
-m31. E.g. if build with gcc 10, gcc is generating a call to
__glob64_time64.localalias.
This patch is adding a hidden version of __glob64_time64 in the
same way as for __globfree64_time64.
Add hp-timing.h using the cntvct_el0 counter. Return timing in nanoseconds
so it is fully compatible with generic hp-timing. Don't set HP_TIMING_INLINE
in the dynamic linker since it adds unnecessary overheads and some ancient
kernels may not handle emulating cntcvt correctly. Currently cntvct_el0 is
only used for timing in the benchtests.
Reviewed-by: Szabolcs Nagy <szabolcs.nagy@arm.com>
