descriptors is unsafe, and that applications should instead atomically create
file descriptors with FD_CLOEXEC already set. This is possible with open( )
and openat( ) using O_CLOEXEC, and with fcntl( ) using F_DUPFD_CLOEXEC, but
there are numerous other interfaces in the standard which also create new
file descriptors where atomic FD_CLOEXEC was not possible. Without support
for atomic FD_CLOEXEC on all possible file descriptors, then multi-threaded
applications have a data race where one thread can call fork( ) or
posix_spawn( ) in the window between when another thread has created a new
file descriptor and had a chance to use fcntl( ) to mark it FD_CLOEXEC, and
thus leak unintended file descriptors into a child process.
This bug goes hand-in-hand with 0000368 which guarantees that all other
file descriptors not exposed to the application are atomically marked
FD_CLOEXEC. As it adds new interfaces, it is necessarily targeted to
Issue 8.
Most of these changes are modeled after existing practice in at least Linux
and Cygwin. Also, the pipe2( ) function is emulated on many other platforms
in the gnulib library, although FD_CLOEXEC is not atomic without underlying
support.
In the process of adding this, I also folded in the changes in C1X to add
the 'x' modifier to the fopen( ) mode, although I have added some additional
requirements with <CX> shading to make it more useful. We may decide to
make further usability enhancements, such as allowing "wx+" to be a valid
mode (as is the case in glibc), rather than limiting ourselves to just the
C1X spelling of "w+x".
I noticed that the standard is silent on the behavior of accept( ) when
used on a socket marked O_NONBLOCK, and since BSD and Linux behaviors
differ (BSD inherits O_NONBLOCK while Linux always clears the flag in
the new descriptor), I documented that while mandating particular
behavior for the new function. That is, it is implementation-defined
whether accept(sock, addr, len) behaves like accept4(sock, addr, len, 0)
or accept4(sock, addr, len, O_NONBLOCK) when sock is O_NONBLOCK.
Also, I don't know of any platform where posix_spawn_file_actions_adddup2( )
can clear FD_CLOEXEC, but that seems like a useful change to make while
tightening the specification on how to properly use FD_CLOEXEC, since the
only standardized alternative for explicitly handing a file descriptor to
the child process while leaving it FD_CLOEXEC in the parent process involves
more complexity and risks EMFILE failure.
Not added here, but worth thinking about:
BSD and glibc both provide mkstemps (create a temporary file with a
specified suffix in the filename), and glibc added mkostemps as the
counterpart for mkstemps to match its mkostemp call added here.
Adding posix_spawn_file_actions_openat might be useful.
The documentation for SCM_RIGHTS as a means of creating new file
descriptors is very minimal (a one-line mention in <sys/socket.h>);
perhaps the sendmsg( ) and recvmsg( ) pages should document it further,
as well as adding pages to further document the various CMSG_ macros.
It may also be worth adding requirements from RFC 3542 on the CMSG_
macros, including the addition of CMSG_SPACE.
The fact that fcntl( ) has unspecified results on typed memory file
descriptors seems a bit worrying; perhaps we should improve things to
state that certain actions (F_GETFD, F_SETFD, F_DUPFD, F_DUPFD_CLOEXEC)
are still well-defined, while leaving other actions unspecified,
especially since dup( ) is well-defined but is identical to fcntl( )
with F_DUPFD.
int mkostemp(char *, int);
After line 12869 [XBD <sys/socket.h>], add the following:
The <sys/socket.h> header shall define the following symbolic
constants with values that are bitwise distinct from each other and
from the preceding SOCK_* constants:
SOCK_NONBLOCK Create a socket file descriptor with the O_NONBLOCK flag
atomically set on the new open file description.
SOCK_CLOEXEC Create a socket file descriptor with the FD_CLOEXEC flag
atomically set on that file descriptor.
After line 12897 [XBD <sys/socket.h>], add a line:
MSG_CMSG_CLOEXEC Atomically set the FD_CLOEXEC flag on any file
descriptors created via SCM_RIGHTS during recvmsg( ).
After line 12920 [XBD <sys/socket.h>], add a line:
int accept4(int, struct sockaddr *restrict, socklen_t *restrict, int);
After line 15050 [XBD <unistd.h>], add a line:
int dup3(int, int, int);
After line 15094 [XBD <unistd.h>], add a line:
int pipe2(int [2], int);
At line 16722 [XSH 2.4.3 Signal Actions], insert the following
functions in sorted order into the list of async-signal-safe
functions:
accept4( )
dup3( )
pipe2( )
At line 17860 [XSH 2.9.7 Thread Interactions with Regular File
Operations], insert "dup3( )" in sorted order into the list of atomic
file operation functions.
At line 17693 [XSH 2.9.5.2 Cancellation Points], insert "accept4( )"
in sorted order into the list of cancellation point functions.
At line 18326 [XSH 2.10.20.2 Compatibility with IPv4], change "in the
accept( )," to "in the accept( ), accept4( ),".
At line 19366 [XSH accept NAME], change "accept - accept a new
connection on a socket" to "accept, accept4 - accept a new connection
on a socket"
After line 19370 [XSH accept SYNOPSIS], add a line:
int accept4(int socket, struct sockaddr *restrict address, socklen_t
*restrict address_len, int flag);
After line 19395 [XSH accept DESCRIPTION], add the following:
If O_NONBLOCK is set on the file description for socket, it is
unspecified whether O_NONBLOCK will be set on the file description
created by accept( ).
The accept4( ) function shall be equivalent to the accept( ) function,
except that the O_NONBLOCK flag shall not be set on the new file
description if the flag argument is 0. Additionally, the flag
argument can be constructed from a bitwise-inclusive OR of flags from
the following list:
SOCK_CLOEXEC Atomically set the FD_CLOEXEC flag on the new file
descriptor.
SOCK_NONBLOCK Set the O_NONBLOCK file status flag on the new file
description.
At line 19397 [XSH accept RETURN VALUE], change "accept( )" to
"accept( ) and accept4( )"
At lines 19400 and 19417 [XSH accept ERRORS], change "accept( )
function" to "accept( ) and accept4( ) functions"
After line 19419 [XSH accept ERRORS], add the following:
The accept4( ) function may fail if:
[EINVAL] The value of the flag argument is invalid.
At line 19426 [XSH accept RATIONALE], change "None." to:
The SOCK_CLOEXEC flag of accept4( ) is necessary to avoid a data race
in multi-threaded applications. Without it, a file descriptor is
leaked into a child process created by one thread in the window
between another thread creating a file descriptor with accept( ) and
then using fcntl( ) to set the FD_CLOEXEC flag. The SOCK_NONBLOCK
flag is for convenience in avoiding additional fcntl( ) calls, as well
as providing specific control over the O_NONBLOCK flag, since
traditional implementations of accept( ) differ on whether O_NONBLOCK
is inherited from the socket argument.
After line 23749 [XSH creat RATIONALE], add the following:
In multi-threaded applications, the creat( ) function can leak file
descriptors into child processes. Applications should instead use
open( ) with the O_CLOEXEC flag to avoid the leak.
At line 24863 [XSH dup NAME], change "dup, dup2" to "dup, dup2, dup3"
After line 24867 [XSH dup SYNOPSIS], add a line:
int dup3(int fildes, int fildes2, int flag);
After line 24881 [XSH dup DESRIPTION], add the following:
The dup3( ) function shall be equivalent to the dup2( ) function if
the flag argument is 0, except that it shall be an error if fildes is
equal to fildes2. Additionally, the flag parameter can be set to
O_CLOEXEC (from <fcntl.h>) to cause FD_CLOEXEC flag to be set on the
new file descriptor.
At line 24882 [XSH dup DESCRIPTION], change "dup2( ) function" to
"dup2( ) or dup3( ) functions".
At lines 24890 and 24894 [XSH dup ERRORS], change "dup2( ) function"
to "dup2( ) and dup3( ) functions".
After line 24893 [XSH dup ERRORS], add the following:
The dup3( ) function shall fail if:
[EINVAL] The fildes and fildes2 arguments are equal.
After line 24895 [XSH dup ERRORS], add the following:
The dup3( ) function may fail if:
[EINVAL] The value of the flag argument is invalid.
After line 24926 [XSH dup RATIONALE], add the following:
The dup3( ) function with the O_CLOEXEC flag is necessary to avoid a
data race in multi-threaded applications. Without it, a file
descriptor is leaked into a child process created by one thread in the
window between another thread creating a file descriptor with dup2( )
and then using fcntl( ) to set the FD_CLOEXEC flag. The safe
counterpart for avoiding the same race in dup( ) is the use of the
F_DUP_CLOEXEC action of the fcntl( ) function.
At line 27380 [XSH fdopen DESCRIPTION], change:
The mode argument is a character string having one of the following
values:
to:
The mode argument points to a character string, which shall also be
valid for fopen( ). The string prefix has the following effects:
After line 27386 [XSH fdopen DESCRIPTION], change:
The meaning of these flags is exactly as specified in fopen( ), except
that modes beginning with w shall not cause truncation of the file.
to:
The meaning of these flags is exactly as specified in fopen( ), except
that modes beginning with w shall not cause truncation of the file,
and the use of x shall have no effect. The FD_CLOEXEC flag of fildes
shall be unchanged if e was not present, and shall be set if e is
present.
At line 27424 [XSH fdopen RATIONALE], add the following to the same
paragraph:
Since fdopen( ) does not create a file, the x mode modifier is
silently ignored. The e mode modifier is not strictly necessary for
fdopen( ), since FD_CLOEXEC must not be changed when it is absent;
however, since it is necessary to avoid a data race in multi-threaded
applications using freopen( ), and consistency dictates that all mode
arguments should allow the same strings.
At line 27471 [XSH fdopendir DESCRIPTION], change:
It is unspecified whether the FD_CLOEXEC flag will be set on the file
descriptor by a successful call to fdopendir( ).
to:
It is unspecified whether the FD_CLOEXEC flag will be set on the file
descriptor by a successful call to fdopendir( ) if it was not
previously set. However, the flag shall not be cleared if it was
previously set.
At line 29115 [XSH fopen DESCRIPTION], change:
The mode argument points to a string. If the string is one of the
following, the file shall be opened in the indicated mode. Otherwise,
the behavior is undefined.
to:
The mode argument points to a character string. If the string begins
with one of these six prefixes, followed by a (possibly empty) suffix
consisting of the additional characters documented below, then the
file shall be opened in the mode indicated by the prefix. Otherwise,
the behavior is undefined.
After line 29125 [XSH fopen DESCRIPTION], add the following:
Additionally, the following characters may appear anywhere in the
suffix of the mode string, to further affect how the file is opened.
Behavior is unspecified if a character occurs more than once.
<CX>'e' The underlying file descriptor shall have the FD_CLOEXEC flag
atomically set, as if by the O_CLOEXEC flag to open( ).</CX>
'x' If specified with a prefix beginning with 'w' <CX>or 'a'</CX>,
then the function shall fail if the file already exists, <CX>as if by
the O_EXCL flag to open( ). If specified with a prefix beginning with
'r', this modifier shall have no effect.</CX>
After line 29168 [XSH fopen ERRORS], add a line:
<CX>[EEXIST] The mode argument begins with 'w' or 'a' and includes
'x', but the file already exists.</CX>
At line 29223 [XSH fopen RATIONALE], change "None." to:
The e mode suffix character is provided as a convenience to avoid a
data race in multi-threaded applications. Without it, a file
descriptor is leaked into a child process created by one thread in the
window between another thread creating a file descriptor with fopen( )
and then using fileno( ) and fcntl( ) to set the FD_CLOEXEC flag. It
is also possible to avoid the race by using open( ) with O_CLOEXEC
followed by fdopen( ), however, there is no safe alternative for the
freopen( ) function, and consistency dictates that the e modifier
should be standardized for all mode strings.
The x mode suffix character was added by C1x only for files opened
with a mode string beginning with w. However, this standard requires
that it also work for mode strings beginning with a, as well as being
silently ignored rather than being an error for mode strings beginning
with r. Therefore, while open( ) has undefined behavior if O_EXCL is
specified without O_CREAT, the same is not true of fopen( ).
This standard follows the lead of C1x in requiring that b and + appear
in the mode prefix and x in the mode suffix; however, implementations
are encouraged to treat both b and + as part of the mode suffix
(requiring only w, a, or r as the first character and letting all
other characters appear in any order).
At line 30897 [XSH freopen DESCRIPTION], change "freopen( ) Mode" to
"freopen( ) Mode Prefix".
After line 30903 [XSH freopen DESCRIPTION], add the following:
<CX>Additionally, the presence of e in the mode suffix shall behave as
if the O_CLOEXEC flag to open( ) were in use. The presence of x in
the mode suffix where the prefix begins with a or w shall behave as if
the O_EXCL flag to open( ) were specified, and shall be ignored if the
prefix begins with r.</CX>
After line 30914 [XSH freopen ERRORS], add a line:
<CX>[EEXIST] The mode argument begins with 'w' or 'a' and includes
'x', but the file already exists.</CX>
At line 30979 [XSH freopen RATIONALE], change "None." to:
The e mode suffix character is necessary to avoid a data race in
multi-threaded applications. Without it, a file descriptor is leaked
into a child process created by one thread in the window between
another thread creating a file descriptor with freopen( ) and then
using fileno( ) and fcntl( ) to set the FD_CLOEXEC flag.
The x mode suffix character was added by C1x only for files opened
with a mode string beginning with w. However, this standard requires
that it also work for mode strings beginning with a, as well as being
silently ignored rather than being an error for mode strings beginning
with r. Therefore, while open( ) has undefined behavior if O_EXCL is
specified without O_CREAT, the same is not true of freopen( ).
This standard follows the lead of C1x in requiring that b and + appear
in the mode prefix and x in the mode suffix; however, implementations
are encouraged to treat both b and + as part of the mode suffix
(requiring only w, a, or r as the first character and letting all
other characters appear in any order).
At line 42413 [XSH mkdtemp NAME], change "mkdtemp, mkstemp" to
"mkdtemp, mkostemp, mkstemp".
After line 42416 [XSH mkdtemp SYNOPSIS], add a line:
int mkostemp (char *template, int flag);
After line 42433 [XSH mkdtemp DESCRIPTION], add the following:
The mkostemp( ) function shall be equivalent to the mkstemp( )
function, except that the flag argument may contain additional flags
(from <fcntl.h>) to be used as if by open( ). Behavior is unspecified
if the flag argument contains more than the following flags:
O_APPEND Set append mode.
O_CLOEXEC Set the FD_CLOEXEC file descriptor flag.
<SIO>O_DSYNC Write according to the synchronized I/O data integrity
completion.</SIO>
<SIO>O_RSYNC Synchronized read I/O operations.</SIO>
<XSI|SIO>O_SYNC Write according to synchronized I/O file integrity
completion.</XSI|SIO>
At line 42463 [XSH mkdtemp ERRORS], change "mkstemp( ) function" to
"mkstemp( ) and mkostemp( ) functions".
After line 42463 [XSH mkdtemp ERRORS], add the following:
The mkostemp( ) function may fail if:
[EINVAL] The value of the flag argument is invalid.
At line 42479 [XSH mkdtemp RATIONALE], replace "None." with:
The function mkostemp( ) with the O_CLOEXEC flag is necessary to avoid
a data race in multi-threaded applications. Without it, a file
descriptor is leaked into a child process created by one thread in the
window between another thread creating a temporary file descriptor
with mkstemp( ) and then using fcntl( ) to set the FD_CLOEXEC flag.
At line 45830 [XSH pipe NAME], change "pipe" to "pipe, pipe2".
After line 45833 [XSH pipe SYNOPSIS], add a line:
int pipe2(int fildes[2], int flag);
After line 45849 [XSH pipe DESCRIPTION], add the following:
The pipe2( ) function shall be equivalent to the pipe( ) function if
the flag argument is 0. Additionally, the flag argument can be
constructed from a bitwise-inclusive OR of flags from the following
list (provided by <fcntl.h>):
O_CLOEXEC Atomically set the FD_CLOEXEC flag on both new file
descriptors.
O_NONBLOCK Set the O_NONBLOCK file status flag on both new file
descriptions.
At line 45854 [XSH pipe ERRORS], change "pipe( ) function" to "pipe( )
and pipe2( ) functions"
After line 45858 [XSH pipe ERRORS], add the following:
The pipe2( ) function may fail if:
[EINVAL] The value of the flag argument is invalid.
At line 45897 [XSH pipe RATIONALE], add the following:
The O_CLOEXEC flag of pipe2( ) is necessary to avoid a data race in
multi-threaded applications. Without it, a file descriptor is leaked
into a child process created by one thread in the window between
another thread creating a file descriptor with pipe( ) and then using
fcntl( ) to set the FD_CLOEXEC flag. The O_NONBLOCK flag is for
convenience in avoiding additional fcntl( ) calls.
At lines 46095 and 46099 [XSH popen DESCRIPTION], change "If mode is"
to "If mode starts with".
After line 46102 [XSH popen DESCRIPTION], add the following:
3. If mode includes a second character of e, then the file descriptor
underlying the stream returned to the calling process by popen( )
shall have the FD_CLOEXEC flag atomically set, and the implementation
shall ensure that the file descriptor given to the child process does
not clear the FD_CLOEXEC flag until after the child process is created
(as if after the fork( ) but before the execl( )).
At line 46103 [XSH popen DESCRIPTION], change "3." to "4.".
At line 46148 [XSH popen APPLICATION USAGE], change "r and w" to "r,
w, re, and we".
After line 46167 [XSH popen RATIONALE], add the following:
The e mode modifier to popen( ) is necessary to avoid a data race in
multi-threaded applications. Without it, a file descriptor is leaked
into a child process created by one thread in the window between
another thread creating the pipe as if by pipe( ) and then using
fileno( ) and fcntl( ) on the result of popen( ). Implementations may
implement the e mode modifier by using pipe2( ) with the O_CLOEXEC
flag, then relying on dup2( ) in the child to clear the FD_CLOEXEC
flag on the child descriptor. Although the standard is clear that a
conforming application should not call popen( ) when file descriptor 0
or 1 is closed, implementations should take care to ensure that the
FD_CLOEXEC flag is properly cleared even when the result of pipe2( )
is one of the standard descriptors, since dup2( ) does not clear the
FD_CLOEXEC flag if its two arguments are identical. It is also
recommended, but not required, that implementations avoid leaking the
child's file descriptor even when the e mode modifier is not present,
by always using pipe2( ) with the O_CLOEXEC flag, followed by use of
fcntl( ) in the parent to clear FD_CLOEXEC of just the parent's file
descriptor.
After line 46469 [XSH posix_openpt DESCRIPTION], add a line:
O_CLOEXEC Atomically set the FD_CLOEXEC flag on the file descriptor.
After line 46515 [XSH posix_openpt RATIONALE], add the following:
The function posix_openpt( ) with the O_CLOEXEC flag is necessary to
avoid a data race in multi-threaded applications. Without it, a file
descriptor is leaked into a child process created by one thread in the
window between another thread creating a file descriptor with
posix_openpt( ) and then using fcntl( ) to set the FD_CLOEXEC flag.
At line 46976 [XSH posix_spawn_file_actions_adddup2], add a sentence:
If fildes and newfildes are equal, then the action shall ensure that
the FD_CLOEXEC flag of fildes is cleared (even though dup2( ) would
leave it unchanged).
After line 46999 [XSH posix_spawn_file_actions_adddup2], add the
following:
Although dup2( ) is required to do nothing when fildes and newfildes
are equal and fildes is an open descriptor, the use of
posix_spawn_file_actions_adddup2( ) is required to clear the
FD_CLOEXEC flag of fildes. This is because there is no counterpart of
posix_spawn_file_actions_fcntl( ) that could be used for clearing the
flag; it would also be possible to achieve this effect by using two
calls to posix_spawn_file_actions_adddup2( ) and a temporary fildes
value known to not conflict with any other file descriptors, coupled
with a posix_spawn_file_actions_close( ) to avoid leaking the
temporary, but this approach is complex, and risks EMFILE or ENFILE
failure that can be avoided with the in-place removal of FD_CLOEXEC.
There is no need for posix_spawn_file_actions_adddup3( ), since it
makes no sense to create a file descriptor with FD_CLOEXEC set before
spawning the child process, where that file descriptor would
immediately be closed again.
At line 48927 [XSH posix_typed_mem_open DESCRIPTION], change:
The FD_CLOEXEC file descriptor flag associated with the new file
descriptor shall be cleared.
to:
The FD_CLOEXEC file descriptor flag associated with the new file
descriptor shall be cleared unless oflag includes O_CLOEXEC.
At line 48930 [XSH posix_typed_mem_open DESCRIPTION], change "dup( ),
dup2( )," to "dup( ), dup2( ), dup3( ),".
After line 48937 [XSH posix_typed_mem_open DESCRIPTION], add the
following:
Additionally, the value of oflag may include the following flag:
O_CLOEXEC Set the FD_CLOEXEC file descriptor flag.
At line 48967 [XSH posix_typed_mem_open RATIONALE], replace "None."
with:
The use of the O_CLOEXEC flag to posix_typed_mem_open( ) is necessary
to avoid leaking typed memory file descriptors to child processes,
since fcntl( ) has unspecified results on typed memory objects and
therefore cannot be used to set FD_CLOEXEC after the fact. The exec
family of functions already unmaps all memory associated with a typed
memory object, but does not close the file descriptor unless
FD_CLOEXEC is also set.
At line 49184 [XSH pselect DESCRIPTION], change "accept( ) function"
to "accept( ) or accept4( ) function"
After line 56369 [XSH recvmsg DESCRIPTION], add the following:
MSG_CMSG_CLOEXEC On sockets that permit a cmsg_type of SCM_RIGHTS in
the msg_control ancillary data as a means of copying file
descriptors into the process, the file descriptors shall be created
with the FD_CLOEXEC flag atomically set.
At line 56432 [XSH recvmsg RATIONALE], replace "None." with:
The use of the MSG_CMSG_CLOEXEC flag to recvmsg( ) when using
SCM_RIGHTS to receive file descriptors via ancillary data is necessary
to avoid a data race in multi-threaded applications. Without it, a
file descriptor is leaked into a child process created by one thread
in the window between another thread calling recvmsg( ) and using
fcntl( ) to set the FD_CLOEXEC flag.
At line 60458 [XSH shm_open RATIONALE], add a sentence:
The O_CLOEXEC open flag is not listed, because all shared memory
objects are created with the FD_CLOEXEC flag already set; an
application may later use fcntl( ) to clear that flag to allow the
shared memory file descriptor to be preserved across the exec family
of functions.
After line 62515 [XSH socket DESCRIPTION], add the following:
Additionally, the type argument may contain the bitwise-inclusive OR
of flags from the following list:
SOCK_CLOEXEC Atomically set the FD_CLOEXEC flag on the new file
descriptor.
SOCK_NONBLOCK Set the O_NONBLOCK file status flag on the new file
description.
At line 62547 [XSH socket RATIONALE], replace "None." with:
The use of the SOCK_CLOEXEC flag in the type argument of socket( ) is
necessary to avoid a data race in multi-threaded applications.
Without it, a file descriptor is leaked into a child process created
by one thread in the window between another thread calling socket( )
and using fcntl( ) to set the FD_CLOEXEC flag. The SOCK_NONBLOCK
flag is for convenience in avoiding additional fcntl( ) calls.
After line 62586 [XSH socketpair DESCRIPTION], add the following:
Additionally, the type argument may contain the bitwise-inclusive OR
of flags from the following list:
SOCK_CLOEXEC Atomically set the FD_CLOEXEC flag on the new file
descriptors.
SOCK_NONBLOCK Set the O_NONBLOCK file status flag on the new file
descriptions.
At line 62620 [XSH socketpair RATIONALE], replace "None." with:
The use of the SOCK_CLOEXEC flag in the type argument of socketpair( )
is necessary to avoid a data race in multi-threaded applications.
Without it, a file descriptor is leaked into a child process created
by one thread in the window between another thread using socketpair( )
and using fcntl( ) to set the FD_CLOEXEC flag. The SOCK_NONBLOCK
flag is for convenience in avoiding additional fcntl( ) calls.
After line 67169 [XSH tmpfile APPLICATION USAGE], add the following:
In multi-threaded applications, the tmpfile( ) function can leak file
descriptors into child processes. Applications should instead use
mkostemp( ) with the O_CLOEXEC flag, followed by fdopen( ), to avoid
the leak.
At line 67239 [XSH tmpnam APPLICATION USAGE], change "mkstemp( )" to
"mkostemp( ), mkstemp( )".
At line 119194 [XRAT B.2.8.3 Memory Management], change "dup2( )," to
"dup2( ), dup3( ),".
At line 123297 [XRAT B.3.3 Examples for Spawn], change:
if (dup2(fd, newfd) == −1)
to:
if (fd == newfd)
{
int flags = fcntl(fd, F_GETFD);
if (flags == -1)
exit(127);
flags &= ~FD_CLOEXEC;
if (fcntl(fd, F_SETFD, flags) == -1)
exit(127);
}
else if (dup2(fd, newfd) == -1)
At line 125602 [XRAT D.2.3 Access to Data], change "fopen( ), and
pipe( )" to "fopen( ), freopen( ), pipe( ), and pipe2( )".
At line 125605 [XRAT D.2.3 Access to Data], change "dup2( )," to
"dup2( ), dup3( ),".
At line 126332 [XRAT E.1 Subprofiling Option Groups], add "dup3( )" in
order to POSIX_FD_MGMT.
At line 126381 [XRAT E.1 Subprofiling Option Groups], add "accept4( )" in
order to POSIX_NETWORKING.
At line 126391 [XRAT E.1 Subprofiling Option Groups], add "pipe2( )" in
order to POSIX_PIPE.
bug411_atomic_CLOEXEC.odt [^] (50,219 bytes) 2014-05-03 16:45
bug411_atomic_CLOEXEC.pdf [^] (220,549 bytes) 2014-05-03 16:45 bug411_atomic_CLOEXEC_v2.odt [^] (32,338 bytes) 2020-03-12 18:42 bug411_atomic_CLOEXEC_v2.pdf [^] (145,796 bytes) 2020-03-12 18:42 
