Thread Sanitizer can't intercept syscall(SYS_close, fd) that's used
instead of close(fd); on Linux. That leads to false positives as Thread
Sanitizer thinks the descriptor is still being used by the thread.
clang defines pre- and post- syscall actions, so wrap the close
syscall() into the action macros. For gcc, use close() from glibc
instead of the syscall. This allows the thread sanitizer to intercept
closing of the file descriptor when libuv is compiled with Thread
Sanitizer.
PR-URL: https://github.com/libuv/libuv/pull/3112
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
`pthread_attr_init` is highly unlikely to fail on macOS. Removing the
fallback behavior here to be consistent with other parts of libuv (e.g.
`src/unix/thread.c`), which simply call `abort()`.
PR-URL: https://github.com/libuv/libuv/pull/3132
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Reviewed-By: Santiago Gimeno <santiago.gimeno@gmail.com>
This fixes `SIGBUS` crashes on macOS 10.15 due a new change in
`FSEvents.framework` that makes it allocate a large stack array for
event paths. (See the linked nodejs/node issue for details on the
`FSEvents.framework` memory requirements change itself.)
The existing size (`4 * PTHREAD_STACK_MIN` or 32KB) causes a stack
overflow when more than ~1000 events are received at once. Setting this
to `uv__thread_stack_size()` increases it to 8192KB (by default) on
64-bit machines. This value can be configured at runtime on macOS with
`ulimit -s <size-kb>`.
The 32KB limit was originally added to reduce virtual memory
fragmentation on 32-bit systems, which is not a concern on 64-bit
systems.
Fixes: nodejs/node#37697
Refs: joyent/libuv#964
PR-URL: https://github.com/libuv/libuv/pull/3132
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Reviewed-By: Santiago Gimeno <santiago.gimeno@gmail.com>
This function will be used by `uv__fsevents_loop_init` in a future
commit to determine the initial FSEvents pthread stack size.
PR-URL: https://github.com/libuv/libuv/pull/3132
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Reviewed-By: Santiago Gimeno <santiago.gimeno@gmail.com>
In both `test-tcp-open` and `test-udp-open`.
PR-URL: https://github.com/libuv/libuv/pull/3137
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Reviewed-By: Santiago Gimeno <santiago.gimeno@gmail.com>
There is no CMAKE_LINKER_FLAGS_DEBUG.
Signed-off-by: Zhao Zhili <zhilizhao@tencent.com>
PR-URL: https://github.com/libuv/libuv/pull/3137
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Reviewed-By: Santiago Gimeno <santiago.gimeno@gmail.com>
In the codebase we have used empty for loop for infinite conditions, so
to bring consistency replaced other occurrences of while in the codebase
with for loop.
PR-URL: https://github.com/libuv/libuv/pull/3128
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Reviewed-By: Colin Ihrig <cjihrig@gmail.com>
Reviewed-By: Santiago Gimeno <santiago.gimeno@gmail.com>
PR-URL: https://github.com/libuv/libuv/pull/3130
Reviewed-By: Colin Ihrig <cjihrig@gmail.com>
Reviewed-By: Richard Lau <rlau@redhat.com>
Reviewed-By: Santiago Gimeno <santiago.gimeno@gmail.com>
To properly handle sending UDP packet to unreachable address.
PR-URL: https://github.com/libuv/libuv/pull/2872
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Reviewed-By: Santiago Gimeno <santiago.gimeno@gmail.com>
The Linux kernel suppresses some ICMP error messages by default for UDP
sockets. This commit sets IP_RECVERR/IPV6_RECVERR on the socket to
enable full ICMP error reporting, hopefully resulting in faster failover
to working name servers.
PR-URL: https://github.com/libuv/libuv/pull/2872
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Reviewed-By: Santiago Gimeno <santiago.gimeno@gmail.com>
ZOSLIB uses `argv[0]` to determine the exepath. So it is necessary to
use an implementation of proctitle that does not modify `argv[0]`. Since
there is currently no support for process titles on z/OS, the custom
proctitle implementation simply stores the desired title in memory.
This resolves failure in tests `get_currentexe` and `process_title` on
z/OS.
PR-URL: https://github.com/libuv/libuv/pull/3060
Reviewed-By: Richard Lau <rlau@redhat.com>
On z/OS, `execvp()` does not set the environment for child process from
`environ` when ran in ASCII mode. Instead, `execvpe()` provided by
ZOSLIB must be used to set the environment explicitly.
PR-URL: https://github.com/libuv/libuv/pull/3060
Reviewed-By: Richard Lau <rlau@redhat.com>
`__rfim_utok` is treated as text when it should be treated as binary
while running in ASCII mode, resulting in an unwanted autoconversion. So
undo the conversion explicitly.
PR-URL: https://github.com/libuv/libuv/pull/3060
Reviewed-By: Richard Lau <rlau@redhat.com>
Since `nanosleep()` is implemented in ZOSLIB, we will remove the libuv
implementation to resolve conflict. The ZOSLIB implementation uses
BPX4CTW (cond_timed_wait).
PR-URL: https://github.com/libuv/libuv/pull/3060
Reviewed-By: Richard Lau <rlau@redhat.com>
Since `strnlen()` is implemented in ZOSLIB, we will remove the libuv
implementation to resolve conflict. The ZOSLIB implementation uses asm.
PR-URL: https://github.com/libuv/libuv/pull/3060
Reviewed-By: Richard Lau <rlau@redhat.com>
This commit introduces ZOSLIB for z/OS, which is a C/C++ library that
implements additional POSIX APIs not available in the LE C Runtime
Library, and provides API for EBCDIC <-> ASCII conversion. This library
requires the linker to be set to CXX when building for z/OS. ZOSLIB is
designed to be installed separately, and then linked to libuv with the
`-DZOSLIB_DIR` option.
PR-URL: https://github.com/libuv/libuv/pull/3060
Reviewed-By: Richard Lau <rlau@redhat.com>
Nanosecond resolution for the timestamp fields `st_atim`, `st_mtim`, and
`st_ctim` are not supported on z/OS.
PR-URL: https://github.com/libuv/libuv/pull/3060
Reviewed-By: Richard Lau <rlau@redhat.com>
Moving to new style test macros will make debugging easier in case
of test failure and improve redability. This commit will replace all
ASSERT macros matching the statement:
`ASSERT(identifier (== or !=) value);`
to:
`ASSERT_(NOT_)NULL(identifier);`
Refs: https://github.com/libuv/libuv/issues/2974
PR-URL: https://github.com/libuv/libuv/pull/3081
Reviewed-By: Santiago Gimeno <santiago.gimeno@gmail.com>
Applications running on hardened runtime based on Chromium/Electron
create mmap regions using MAP_JIT flag. With macOS Big Sur the fork()
calls done by uv_spawn have become slow. This is because fork() seems
to physically copy all JIT memory regions (no-copy-on-write). On
previous OS, these regions weren't accessible at all in the forked
process, explaining the regression.
The fix is to use posix_spawn() on macOS. This spawns a new process
directly, without copying any memory mappings.
Note that fork() is still used on earlier versions of macOS if the
necessary posix_spawn() platform-specific extensions are not available.
Fixes: https://github.com/libuv/libuv/issues/3050
PR-URL: https://github.com/libuv/libuv/pull/3064
Reviewed-By: Jameson Nash <vtjnash@gmail.com>
Reviewed-By: Santiago Gimeno <santiago.gimeno@gmail.com>
Consider the following scenario:
uv_poll_init(loop, poll, fd);
uv_poll_start(poll, UV_READABLE, cb);
// the cb gets invoked etc.
uv_poll_stop(poll);
close(fd);
fd = allocate_new_socket(); // allocate_new_socket() is assigned the same fd by "bad luck" from the OS
// some time later:
uv_poll_init(loop, otherpoll, fd);
uv_poll_start(otherpoll, UV_READABLE, cb);
uv_close(poll); // uv__io_stop: Assertion `loop->watchers[w->fd] == w' failed.
According to documentation, "however the fd can be safely closed
immediately after a call to uv_poll_stop() or uv_close()."
Though, in this scenario, we close()'d our file descriptor, and by
bad luck we got the same file descriptor again and register a new
handle for it and start polling.
Previously that would lead to an assertion failure, if we were to
properly free the original handle via uv_close().
This commit fixes that by moving the check whether a only a single
poll handle is active to uv_poll_start() instead of the stopping
routines.
Fixes: https://github.com/libuv/libuv/issues/1172
Fixes: https://github.com/bwoebi/php-uv/issues/81
Fixes: https://github.com/b2wdigital/aiologger/issues/82
Fixes: https://github.com/invenia/LibPQ.jl/issues/140
PR-URL: https://github.com/libuv/libuv/pull/2686
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Reviewed-By: Jameson Nash <vtjnash@gmail.com>
* Windows 7 went out of support earlier this year.
* As did Python 2.7. We no longer have to worry about MSVC 2008.
Python 3.5 and up use VS 2015.
PR-URL: https://github.com/libuv/libuv/pull/2821
Reviewed-By: Bartosz Sosnowski <bartosz@janeasystems.com>
Reviewed-By: Jameson Nash <vtjnash@gmail.com>
Reviewed-By: Joao Reis <reis@janeasystems.com>
RFC 2606 reserves the .invalid top-level domain for this purpose.
PR-URL: https://github.com/libuv/libuv/pull/3063
Reviewed-By: Colin Ihrig <cjihrig@gmail.com>
Reviewed-By: Santiago Gimeno <santiago.gimeno@gmail.com>
Reviewed-By: Richard Lau <riclau@uk.ibm.com>
PR-URL: https://github.com/libuv/libuv/pull/3061
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Reviewed-By: Colin Ihrig <cjihrig@gmail.com>
Reviewed-By: Santiago Gimeno <santiago.gimeno@gmail.com>
`uv_fs_utime()` and `uv_fs_futime()` receive the timestamp as
a `double` and then convert it to `struct timeval` or `struct timespec`
where necessary but the calculation for the sub-second part exhibited
rounding errors for dates in the deep past or the far-flung future,
causing the timestamps to be off by sometimes over half a second on
unix, or to be reinterpreted as unsigned and end up off by more than
just sign but many also decades.
Fixes: https://github.com/nodejs/node/issues/32369 (partially)
PR-URL: https://github.com/libuv/libuv/pull/2747
Co-authored-by: Jameson Nash <vtjnash@gmail.com>
Reviewed-By: Colin Ihrig <cjihrig@gmail.com>
