summaryrefslogtreecommitdiffstats
path: root/src/util/poll_fd.c
diff options
context:
space:
mode:
authorDaniel Baumann <daniel.baumann@progress-linux.org>2024-04-07 16:18:56 +0000
committerDaniel Baumann <daniel.baumann@progress-linux.org>2024-04-07 16:18:56 +0000
commitb7c15c31519dc44c1f691e0466badd556ffe9423 (patch)
treef944572f288bab482a615e09af627d9a2b6727d8 /src/util/poll_fd.c
parentInitial commit. (diff)
downloadpostfix-b7c15c31519dc44c1f691e0466badd556ffe9423.tar.xz
postfix-b7c15c31519dc44c1f691e0466badd556ffe9423.zip
Adding upstream version 3.7.10.upstream/3.7.10
Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>
Diffstat (limited to 'src/util/poll_fd.c')
-rw-r--r--src/util/poll_fd.c269
1 files changed, 269 insertions, 0 deletions
diff --git a/src/util/poll_fd.c b/src/util/poll_fd.c
new file mode 100644
index 0000000..80cd0f6
--- /dev/null
+++ b/src/util/poll_fd.c
@@ -0,0 +1,269 @@
+/*++
+/* NAME
+/* poll_fd 3
+/* SUMMARY
+/* wait until file descriptor becomes readable or writable
+/* SYNOPSIS
+/* #include <iostuff.h>
+/*
+/* int readable(fd)
+/* int fd;
+/*
+/* int writable(fd)
+/* int fd;
+/*
+/* int read_wait(fd, time_limit)
+/* int fd;
+/* int time_limit;
+/*
+/* int write_wait(fd, time_limit)
+/* int fd;
+/* int time_limit;
+/*
+/* int poll_fd(fd, request, time_limit, true_res, false_res)
+/* int fd;
+/* int request;
+/* int time_limit;
+/* int true_res;
+/* int false_res;
+/* DESCRIPTION
+/* The read*() and write*() functions in this module are macros
+/* that provide a convenient interface to poll_fd().
+/*
+/* readable() asks the kernel if the specified file descriptor
+/* is readable, i.e. a read operation would not block.
+/*
+/* writable() asks the kernel if the specified file descriptor
+/* is writable, i.e. a write operation would not block.
+/*
+/* read_wait() waits until the specified file descriptor becomes
+/* readable, or until the time limit is reached.
+/*
+/* write_wait() waits until the specified file descriptor
+/* becomes writable, or until the time limit is reached.
+/*
+/* poll_fd() waits until the specified file descriptor becomes
+/* readable or writable, or until the time limit is reached.
+/*
+/* Arguments:
+/* .IP fd
+/* File descriptor. With implementations based on select(), a
+/* best effort is made to handle descriptors >=FD_SETSIZE.
+/* .IP request
+/* POLL_FD_READ (wait until readable) or POLL_FD_WRITE (wait
+/* until writable).
+/* .IP time_limit
+/* A positive value specifies a time limit in seconds. A zero
+/* value effects a poll (return immediately). A negative value
+/* means wait until the requested POLL_FD_READ or POLL_FD_WRITE
+/* condition becomes true.
+/* .IP true_res
+/* Result value when the requested POLL_FD_READ or POLL_FD_WRITE
+/* condition is true.
+/* .IP false_res
+/* Result value when the requested POLL_FD_READ or POLL_FD_WRITE
+/* condition is false.
+/* DIAGNOSTICS
+/* Panic: interface violation. All system call errors are fatal
+/* unless specified otherwise.
+/*
+/* readable() and writable() return 1 when the requested
+/* POLL_FD_READ or POLL_FD_WRITE condition is true, zero when
+/* it is false. They never return an error indication.
+/*
+/* read_wait() and write_wait() return zero when the requested
+/* POLL_FD_READ or POLL_FD_WRITE condition is true, -1 (with
+/* errno set to ETIMEDOUT) when it is false.
+/*
+/* poll_fd() returns true_res when the requested POLL_FD_READ
+/* or POLL_FD_WRITE condition is true, false_res when it is
+/* false. When poll_fd() returns a false_res value < 0, it
+/* also sets errno to ETIMEDOUT.
+/* LICENSE
+/* .ad
+/* .fi
+/* The Secure Mailer license must be distributed with this software.
+/* AUTHOR(S)
+/* Wietse Venema
+/* IBM T.J. Watson Research
+/* P.O. Box 704
+/* Yorktown Heights, NY 10598, USA
+/*--*/
+
+/* System library. */
+
+#include <sys_defs.h>
+#include <sys/time.h>
+#include <signal.h>
+#include <errno.h>
+#include <unistd.h>
+#include <string.h>
+
+ /*
+ * Use poll() with fall-back to select(). MacOSX needs this for devices.
+ */
+#if defined(USE_SYSV_POLL_THEN_SELECT)
+#define poll_fd_sysv poll_fd
+#define USE_SYSV_POLL
+#define USE_BSD_SELECT
+int poll_fd_bsd(int, int, int, int, int);
+
+ /*
+ * Use select() only.
+ */
+#elif defined(USE_BSD_SELECT)
+#define poll_fd_bsd poll_fd
+#undef USE_SYSV_POLL
+
+ /*
+ * Use poll() only.
+ */
+#elif defined(USE_SYSV_POLL)
+#define poll_fd_sysv poll_fd
+
+ /*
+ * Sanity check.
+ */
+#else
+#error "specify USE_SYSV_POLL, USE_BSD_SELECT or USE_SYSV_POLL_THEN_SELECT"
+#endif
+
+#ifdef USE_SYSV_POLL
+#include <poll.h>
+#endif
+
+#ifdef USE_SYS_SELECT_H
+#include <sys/select.h>
+#endif
+
+/* Utility library. */
+
+#include <msg.h>
+#include <iostuff.h>
+
+#ifdef USE_BSD_SELECT
+
+/* poll_fd_bsd - block with time_limit until file descriptor is ready */
+
+int poll_fd_bsd(int fd, int request, int time_limit,
+ int true_res, int false_res)
+{
+ fd_set req_fds;
+ fd_set *read_fds;
+ fd_set *write_fds;
+ fd_set except_fds;
+ struct timeval tv;
+ struct timeval *tp;
+ int temp_fd = -1;
+
+ /*
+ * Sanity checks.
+ */
+ if (FD_SETSIZE <= fd) {
+ if ((temp_fd = dup(fd)) < 0 || temp_fd >= FD_SETSIZE)
+ msg_fatal("descriptor %d does not fit FD_SETSIZE %d", fd, FD_SETSIZE);
+ fd = temp_fd;
+ }
+
+ /*
+ * Use select() so we do not depend on alarm() and on signal() handlers.
+ * Restart select() when interrupted by some signal. Some select()
+ * implementations reduce the time to wait when interrupted, which is
+ * exactly what we want.
+ */
+ FD_ZERO(&req_fds);
+ FD_SET(fd, &req_fds);
+ except_fds = req_fds;
+ if (request == POLL_FD_READ) {
+ read_fds = &req_fds;
+ write_fds = 0;
+ } else if (request == POLL_FD_WRITE) {
+ read_fds = 0;
+ write_fds = &req_fds;
+ } else {
+ msg_panic("poll_fd: bad request %d", request);
+ }
+
+ if (time_limit >= 0) {
+ tv.tv_usec = 0;
+ tv.tv_sec = time_limit;
+ tp = &tv;
+ } else {
+ tp = 0;
+ }
+
+ for (;;) {
+ switch (select(fd + 1, read_fds, write_fds, &except_fds, tp)) {
+ case -1:
+ if (errno != EINTR)
+ msg_fatal("select: %m");
+ continue;
+ case 0:
+ if (temp_fd != -1)
+ (void) close(temp_fd);
+ if (false_res < 0)
+ errno = ETIMEDOUT;
+ return (false_res);
+ default:
+ if (temp_fd != -1)
+ (void) close(temp_fd);
+ return (true_res);
+ }
+ }
+}
+
+#endif
+
+#ifdef USE_SYSV_POLL
+
+#ifdef USE_SYSV_POLL_THEN_SELECT
+#define HANDLE_SYSV_POLL_ERROR(fd, req, time_limit, true_res, false_res) \
+ return (poll_fd_bsd((fd), (req), (time_limit), (true_res), (false_res)))
+#else
+#define HANDLE_SYSV_POLL_ERROR(fd, req, time_limit, true_res, false_res) \
+ msg_fatal("poll: %m")
+#endif
+
+/* poll_fd_sysv - block with time_limit until file descriptor is ready */
+
+int poll_fd_sysv(int fd, int request, int time_limit,
+ int true_res, int false_res)
+{
+ struct pollfd pollfd;
+
+ /*
+ * System-V poll() is optimal for polling a few descriptors.
+ */
+#define WAIT_FOR_EVENT (-1)
+
+ pollfd.fd = fd;
+ if (request == POLL_FD_READ) {
+ pollfd.events = POLLIN;
+ } else if (request == POLL_FD_WRITE) {
+ pollfd.events = POLLOUT;
+ } else {
+ msg_panic("poll_fd: bad request %d", request);
+ }
+
+ for (;;) {
+ switch (poll(&pollfd, 1, time_limit < 0 ?
+ WAIT_FOR_EVENT : time_limit * 1000)) {
+ case -1:
+ if (errno != EINTR)
+ HANDLE_SYSV_POLL_ERROR(fd, request, time_limit,
+ true_res, false_res);
+ continue;
+ case 0:
+ if (false_res < 0)
+ errno = ETIMEDOUT;
+ return (false_res);
+ default:
+ if (pollfd.revents & POLLNVAL)
+ HANDLE_SYSV_POLL_ERROR(fd, request, time_limit,
+ true_res, false_res);
+ return (true_res);
+ }
+ }
+}
+
+#endif