Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

bpo-40423: Optimization: use close_range(2) if available #22651

Merged
merged 4 commits into from
Oct 11, 2020
Merged

bpo-40423: Optimization: use close_range(2) if available #22651

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
a2b6c30
configure: start checking for close_range
kevans91 Apr 27, 2020
fd8a008
_Py_closerange: add close_range(2) support
kevans91 Apr 28, 2020
3081c11
📜🤖 Added by blurb_it.
blurb-it[bot] Oct 11, 2020
e1dc819
_Py_closerange: last-minute braino, flip logic around
kevans91 Oct 11, 2020
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
3 changes: 3 additions & 0 deletions Misc/NEWS.d/next/C API/2020-10-11-19-17-44.bpo-40423.GsmgEj.rst
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,3 @@
The :mod:`subprocess` module and ``os.closerange`` will now use the
``close_range(low, high, flags)`` syscall when it is available for more
efficient closing of ranges of descriptors.
17 changes: 14 additions & 3 deletions Modules/posixmodule.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -8741,12 +8741,15 @@ os_close_impl(PyObject *module, int fd)
}

/* Our selection logic for which function to use is as follows:
* 1. If closefrom(2) is available, we'll attempt to use that next if we're
* 1. If close_range(2) is available, always prefer that; it's better for
* contiguous ranges like this than fdwalk(3) which entails iterating over
* the entire fd space and simply doing nothing for those outside the range.
* 2. If closefrom(2) is available, we'll attempt to use that next if we're
* closing up to sysconf(_SC_OPEN_MAX).
* 1a. Fallback to fdwalk(3) if we're not closing up to sysconf(_SC_OPEN_MAX),
* 2a. Fallback to fdwalk(3) if we're not closing up to sysconf(_SC_OPEN_MAX),
* as that will be more performant if the range happens to have any chunk of
* non-opened fd in the middle.
* 1b. If fdwalk(3) isn't available, just do a plain close(2) loop.
* 2b. If fdwalk(3) isn't available, just do a plain close(2) loop.
*/
#ifdef __FreeBSD__
#define USE_CLOSEFROM
Expand Down Expand Up @@ -8779,6 +8782,14 @@ void
_Py_closerange(int first, int last)
{
first = Py_MAX(first, 0);
#ifdef HAVE_CLOSE_RANGE
if (close_range(first, last, 0) == 0 || errno != ENOSYS) {
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It might be nicer to cache ENOSYS the first time we try close_range and avoid it in future _Py_closerange calls — assuming this is ever used more than once per process, which might not be true.

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I implemented such cache in the PEP 446 implementation: remind if a syscall supports atomic "O_CLOEXEC" flag. See for example _Py_open_cloexec_works in Python/fileutils.c.

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

In my opinion, I don't think this code path is sensitive enough to really care.

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

ah, nice, I hadn't thought of caching that. good idea. (one less failed syscall attempt before the real work likely involving many syscalls gets done in this case; so not a huge deal but still worthwhile)

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

For the subprocess use case, the cache will likely only be first initialized in child processes, each child process has to do the check, and so the cache is useless.

I would mostly benefit to os.closerange().

I'm fine with not adding a cache. I agree that os.closerange() is an uncommon function.

/* Any errors encountered while closing file descriptors are ignored;
* ENOSYS means no kernel support, though,
* so we'll fallback to the other methods. */
}
else
#endif /* HAVE_CLOSE_RANGE */
#ifdef USE_CLOSEFROM
if (last >= sysconf(_SC_OPEN_MAX)) {
/* Any errors encountered while closing file descriptors are ignored */
Expand Down
4 changes: 2 additions & 2 deletions configure
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -11672,8 +11672,8 @@ fi

# checks for library functions
for ac_func in alarm accept4 setitimer getitimer bind_textdomain_codeset chown \
clock confstr copy_file_range ctermid dup3 execv explicit_bzero explicit_memset \
faccessat fchmod fchmodat fchown fchownat \
clock confstr close_range copy_file_range ctermid dup3 execv explicit_bzero \
explicit_memset faccessat fchmod fchmodat fchown fchownat \
fdwalk fexecve fdopendir fork fpathconf fstatat ftime ftruncate futimesat \
futimens futimes gai_strerror getentropy \
getgrgid_r getgrnam_r \
Expand Down
4 changes: 2 additions & 2 deletions configure.ac
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3664,8 +3664,8 @@ fi

# checks for library functions
AC_CHECK_FUNCS(alarm accept4 setitimer getitimer bind_textdomain_codeset chown \
clock confstr copy_file_range ctermid dup3 execv explicit_bzero explicit_memset \
faccessat fchmod fchmodat fchown fchownat \
clock confstr close_range copy_file_range ctermid dup3 execv explicit_bzero \
explicit_memset faccessat fchmod fchmodat fchown fchownat \
fdwalk fexecve fdopendir fork fpathconf fstatat ftime ftruncate futimesat \
futimens futimes gai_strerror getentropy \
getgrgid_r getgrnam_r \
Expand Down
3 changes: 3 additions & 0 deletions pyconfig.h.in
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -139,6 +139,9 @@
/* Define to 1 if you have the `clock_settime' function. */
#undef HAVE_CLOCK_SETTIME

/* Define to 1 if you have the `close_range' function. */
#undef HAVE_CLOSE_RANGE

/* Define if the C compiler supports computed gotos. */
#undef HAVE_COMPUTED_GOTOS

Expand Down