Hello hackers,
I want to be able to do synchronous vectored file I/O, so I made
wrapper macros for preadv() and pwritev() with fallbacks for systems
that don't have them. Following the precedent of the pg_pread() and
pg_pwrite() macros, the "pg_" prefix reflects a subtle contract
change: the fallback paths might have the side effect of changing the
file position.
They're non-standard system calls, but the BSDs and Linux have had
them for a long time, and for other systems we can use POSIX
readv()/writev() with an additional lseek(). The worst case is
Windows (and maybe our favourite antique Unix build farm animal?)
which has none of those things, so there is a further fallback to a
loop. Windows does have ReadFileScatter() and WriteFileGather(), but
those only work for overlapped (= asynchronous), unbuffered, page
aligned access. They'll very likely be useful for native AIO+DIO
support in the future, but don't fit the bill here.
This is part of a project to consolidate and offload I/O (about which
more soon), but seemed isolated enough to post separately and I guess
it could be independently useful.
From 4f4e26d0e2c023d62865365e3cb5ce1c89e63cb0 Mon Sep 17 00:00:00 2001
From: Thomas Munro <thomas.munro@gmail.com>
Date: Thu, 26 Nov 2020 15:48:31 +1300
Subject: [PATCH 07/11] Add pg_preadv() and pg_pwritev().
Provide synchronous scatter/gather I/O routines. These map to preadv(),
pwritev() with various fallbacks for systems that don't have them.
---
configure | 30 ++--------------------------
configure.ac | 9 +++++++--
src/include/pg_config.h.in | 15 ++++++++++++++
src/include/port.h | 37 ++++++++++++++++++++++++++++++++++
src/port/Makefile | 2 ++
src/port/pread.c | 41 ++++++++++++++++++++++++++++++++++++--
src/port/pwrite.c | 41 ++++++++++++++++++++++++++++++++++++--
src/tools/msvc/Solution.pm | 5 +++++
8 files changed, 146 insertions(+), 34 deletions(-)
diff --git a/configure b/configure
index cc2089e596..d25b8f2900 100755
--- a/configure
+++ b/configure
@@ -13199,7 +13199,7 @@ $as_echo "#define HAVE_STDBOOL_H 1" >>confdefs.h
fi
-for ac_header in atomic.h copyfile.h execinfo.h getopt.h ifaddrs.h langinfo.h mbarrier.h poll.h sys/epoll.h sys/event.h sys/ipc.h sys/prctl.h sys/procctl.h sys/pstat.h sys/resource.h sys/select.h sys/sem.h sys/shm.h sys/sockio.h sys/tas.h sys/un.h termios.h ucred.h wctype.h
+for ac_header in atomic.h copyfile.h execinfo.h getopt.h ifaddrs.h langinfo.h mbarrier.h poll.h sys/epoll.h sys/event.h sys/ipc.h sys/prctl.h sys/procctl.h sys/pstat.h sys/resource.h sys/select.h sys/sem.h sys/shm.h sys/sockio.h sys/tas.h sys/uio.h sys/un.h termios.h ucred.h wctype.h
do :
as_ac_Header=`$as_echo "ac_cv_header_$ac_header" | $as_tr_sh`
ac_fn_c_check_header_mongrel "$LINENO" "$ac_header" "$as_ac_Header" "$ac_includes_default"
@@ -15293,7 +15293,7 @@ fi
LIBS_including_readline="$LIBS"
LIBS=`echo "$LIBS" | sed -e 's/-ledit//g' -e 's/-lreadline//g'`
-for ac_func in backtrace_symbols clock_gettime copyfile fdatasync getifaddrs getpeerucred getrlimit kqueue mbstowcs_l memset_s poll posix_fallocate ppoll pstat pthread_is_threaded_np readlink setproctitle setproctitle_fast setsid shm_open strchrnul strsignal symlink sync_file_range uselocale wcstombs_l
+for ac_func in backtrace_symbols clock_gettime copyfile fdatasync getifaddrs getpeerucred getrlimit kqueue mbstowcs_l memset_s poll posix_fallocate ppoll pread preadv pstat pthread_is_threaded_np pwrite pwritev readlink readv setproctitle setproctitle_fast setsid shm_open strchrnul strsignal symlink sync_file_range uselocale wcstombs_l writev
do :
as_ac_var=`$as_echo "ac_cv_func_$ac_func" | $as_tr_sh`
ac_fn_c_check_func "$LINENO" "$ac_func" "$as_ac_var"
@@ -15970,32 +15970,6 @@ esac
fi
-ac_fn_c_check_func "$LINENO" "pread" "ac_cv_func_pread"
-if test "x$ac_cv_func_pread" = xyes; then :
- $as_echo "#define HAVE_PREAD 1" >>confdefs.h
-
-else
- case " $LIBOBJS " in
- *" pread.$ac_objext "* ) ;;
- *) LIBOBJS="$LIBOBJS pread.$ac_objext"
- ;;
-esac
-
-fi
-
-ac_fn_c_check_func "$LINENO" "pwrite" "ac_cv_func_pwrite"
-if test "x$ac_cv_func_pwrite" = xyes; then :
- $as_echo "#define HAVE_PWRITE 1" >>confdefs.h
-
-else
- case " $LIBOBJS " in
- *" pwrite.$ac_objext "* ) ;;
- *) LIBOBJS="$LIBOBJS pwrite.$ac_objext"
- ;;
-esac
-
-fi
-
ac_fn_c_check_func "$LINENO" "random" "ac_cv_func_random"
if test "x$ac_cv_func_random" = xyes; then :
$as_echo "#define HAVE_RANDOM 1" >>confdefs.h
diff --git a/configure.ac b/configure.ac
index c819aeab26..9c21e46ec8 100644
--- a/configure.ac
+++ b/configure.ac
@@ -1342,6 +1342,7 @@ AC_CHECK_HEADERS(m4_normalize([
sys/shm.h
sys/sockio.h
sys/tas.h
+ sys/uio.h
sys/un.h
termios.h
ucred.h
@@ -1671,9 +1672,14 @@ AC_CHECK_FUNCS(m4_normalize([
poll
posix_fallocate
ppoll
+ pread
+ preadv
pstat
pthread_is_threaded_np
+ pwrite
+ pwritev
readlink
+ readv
setproctitle
setproctitle_fast
setsid
@@ -1684,6 +1690,7 @@ AC_CHECK_FUNCS(m4_normalize([
sync_file_range
uselocale
wcstombs_l
+ writev
]))
# These typically are compiler builtins, for which AC_CHECK_FUNCS fails.
@@ -1744,8 +1751,6 @@ AC_REPLACE_FUNCS(m4_normalize([
inet_aton
link
mkdtemp
- pread
- pwrite
random
srandom
strlcat
diff --git a/src/include/pg_config.h.in b/src/include/pg_config.h.in
index 465df421b0..18a68c21ac 100644
--- a/src/include/pg_config.h.in
+++ b/src/include/pg_config.h.in
@@ -412,6 +412,9 @@
/* Define to 1 if you have the `pread' function. */
#undef HAVE_PREAD
+/* Define to 1 if you have the `preadv' function. */
+#undef HAVE_PREADV
+
/* Define to 1 if you have the `pstat' function. */
#undef HAVE_PSTAT
@@ -430,6 +433,9 @@
/* Define to 1 if you have the `pwrite' function. */
#undef HAVE_PWRITE
+/* Define to 1 if you have the `pwritev' function. */
+#undef HAVE_PWRITEV
+
/* Define to 1 if you have the `random' function. */
#undef HAVE_RANDOM
@@ -445,6 +451,9 @@
/* Define to 1 if you have the `readlink' function. */
#undef HAVE_READLINK
+/* Define to 1 if you have the `readv' function. */
+#undef HAVE_READV
+
/* Define to 1 if you have the global variable
'rl_completion_append_character'. */
#undef HAVE_RL_COMPLETION_APPEND_CHARACTER
@@ -626,6 +635,9 @@
/* Define to 1 if you have the <sys/ucred.h> header file. */
#undef HAVE_SYS_UCRED_H
+/* Define to 1 if you have the <sys/uio.h> header file. */
+#undef HAVE_SYS_UIO_H
+
/* Define to 1 if you have the <sys/un.h> header file. */
#undef HAVE_SYS_UN_H
@@ -680,6 +692,9 @@
/* Define to 1 if you have the <winldap.h> header file. */
#undef HAVE_WINLDAP_H
+/* Define to 1 if you have the `writev' function. */
+#undef HAVE_WRITEV
+
/* Define to 1 if you have the `X509_get_signature_nid' function. */
#undef HAVE_X509_GET_SIGNATURE_NID
diff --git a/src/include/port.h b/src/include/port.h
index 5dfb00b07c..70cd1d5287 100644
--- a/src/include/port.h
+++ b/src/include/port.h
@@ -17,6 +17,10 @@
#include <netdb.h>
#include <pwd.h>
+#ifdef HAVE_SYS_UIO_H
+#include <sys/uio.h>
+#endif
+
/*
* Windows has enough specialized port stuff that we push most of it off
* into another file.
@@ -431,6 +435,39 @@ extern ssize_t pg_pread(int fd, void *buf, size_t nbyte, off_t offset);
extern ssize_t pg_pwrite(int fd, const void *buf, size_t nbyte, off_t offset);
#endif
+/*
+ * Likewise for the non-standard but common extensions preadv(2) and
+ * pwritev(2). If sys/uio.h is also missing, define a POSIX-compatible iovec
+ * struct here.
+ */
+#ifndef HAVE_SYS_UIO_H
+struct iovec
+{
+ void *iov_base;
+ size_t iov_len;
+};
+#endif
+
+/*
+ * If <limits.h> didn't define IOV_MAX, we'll define our own for the purposes
+ * of our replacement implementations.
+ */
+#ifndef IOV_MAX
+#define IOV_MAX 16
+#endif
+
+#ifdef HAVE_PREADV
+#define pg_preadv preadv
+#else
+extern ssize_t pg_preadv(int fd, const struct iovec *iov, int iovcnt, off_t offset);
+#endif
+
+#ifdef HAVE_PWRITEV
+#define pg_pwritev pwritev
+#else
+extern ssize_t pg_pwritev(int fd, const struct iovec *iov, int iovcnt, off_t offset);
+#endif
+
#if !HAVE_DECL_STRLCAT
extern size_t strlcat(char *dst, const char *src, size_t siz);
#endif
diff --git a/src/port/Makefile b/src/port/Makefile
index e41b005c4f..bc4923ce84 100644
--- a/src/port/Makefile
+++ b/src/port/Makefile
@@ -53,6 +53,8 @@ OBJS = \
pgstrcasecmp.o \
pgstrsignal.o \
pqsignal.o \
+ pread.o \
+ pwrite.o \
qsort.o \
qsort_arg.o \
quotes.o \
diff --git a/src/port/pread.c b/src/port/pread.c
index e7fecebe5f..4704ebbe7f 100644
--- a/src/port/pread.c
+++ b/src/port/pread.c
@@ -1,7 +1,7 @@
/*-------------------------------------------------------------------------
*
* pread.c
- * Implementation of pread(2) for platforms that lack one.
+ * Implementation of pread[v](2) for platforms that lack one.
*
* Portions Copyright (c) 1996-2020, PostgreSQL Global Development Group
*
@@ -9,7 +9,8 @@
* src/port/pread.c
*
* Note that this implementation changes the current file position, unlike
- * the POSIX function, so we use the name pg_pread().
+ * the POSIX function, so we use the name pg_pread(). Likewise for the
+ * iovec version.
*
*-------------------------------------------------------------------------
*/
@@ -23,6 +24,7 @@
#include <unistd.h>
#endif
+#ifndef HAVE_PREAD
ssize_t
pg_pread(int fd, void *buf, size_t size, off_t offset)
{
@@ -56,3 +58,38 @@ pg_pread(int fd, void *buf, size_t size, off_t offset)
return read(fd, buf, size);
#endif
}
+#endif
+
+#ifndef HAVE_PREADV
+ssize_t
+pg_preadv(int fd, const struct iovec *iov, int iovcnt, off_t offset)
+{
+#ifdef HAVE_READV
+ if (iovcnt == 1)
+ return pg_pread(fd, iov[0].iov_base, iov[0].iov_len, offset);
+ if (lseek(fd, offset, SEEK_SET) < 0)
+ return -1;
+ return readv(fd, iov, iovcnt);
+#else
+ ssize_t sum = 0;
+ ssize_t part;
+
+ for (int i = 0; i < iovcnt; ++i)
+ {
+ part = pg_pread(fd, iov[i].iov_base, iov[i].iov_len, offset);
+ if (part < 0)
+ {
+ if (i == 0)
+ return -1;
+ else
+ return sum;
+ }
+ sum += part;
+ offset += part;
+ if (part < iov[i].iov_len)
+ return sum;
+ }
+ return sum;
+#endif
+}
+#endif
diff --git a/src/port/pwrite.c b/src/port/pwrite.c
index 2e0c154462..ed9897e882 100644
--- a/src/port/pwrite.c
+++ b/src/port/pwrite.c
@@ -1,7 +1,7 @@
/*-------------------------------------------------------------------------
*
* pwrite.c
- * Implementation of pwrite(2) for platforms that lack one.
+ * Implementation of pwrite[v](2) for platforms that lack one.
*
* Portions Copyright (c) 1996-2020, PostgreSQL Global Development Group
*
@@ -9,7 +9,8 @@
* src/port/pwrite.c
*
* Note that this implementation changes the current file position, unlike
- * the POSIX function, so we use the name pg_pwrite().
+ * the POSIX function, so we use the name pg_pwrite(). Likewise for the
+ * iovec version.
*
*-------------------------------------------------------------------------
*/
@@ -23,6 +24,7 @@
#include <unistd.h>
#endif
+#ifndef HAVE_PWRITE
ssize_t
pg_pwrite(int fd, const void *buf, size_t size, off_t offset)
{
@@ -53,3 +55,38 @@ pg_pwrite(int fd, const void *buf, size_t size, off_t offset)
return write(fd, buf, size);
#endif
}
+#endif
+
+#ifndef HAVE_PWRITEV
+ssize_t
+pg_pwritev(int fd, const struct iovec *iov, int iovcnt, off_t offset)
+{
+#ifdef HAVE_WRITEV
+ if (iovcnt == 1)
+ return pg_pwrite(fd, iov[0].iov_base, iov[0].iov_len, offset);
+ if (lseek(fd, offset, SEEK_SET) < 0)
+ return -1;
+ return writev(fd, iov, iovcnt);
+#else
+ ssize_t sum = 0;
+ ssize_t part;
+
+ for (int i = 0; i < iovcnt; ++i)
+ {
+ part = pg_pwrite(fd, iov[i].iov_base, iov[i].iov_len, offset);
+ if (part < 0)
+ {
+ if (i == 0)
+ return -1;
+ else
+ return sum;
+ }
+ sum += part;
+ offset += part;
+ if (part < iov[i].iov_len)
+ return sum;
+ }
+ return sum;
+#endif
+}
+#endif
diff --git a/src/tools/msvc/Solution.pm b/src/tools/msvc/Solution.pm
index 22d6abd367..832a363282 100644
--- a/src/tools/msvc/Solution.pm
+++ b/src/tools/msvc/Solution.pm
@@ -329,17 +329,20 @@ sub GenerateFiles
HAVE_PPC_LWARX_MUTEX_HINT => undef,
HAVE_PPOLL => undef,
HAVE_PREAD => undef,
+ HAVE_PREADV => undef,
HAVE_PSTAT => undef,
HAVE_PS_STRINGS => undef,
HAVE_PTHREAD => undef,
HAVE_PTHREAD_IS_THREADED_NP => undef,
HAVE_PTHREAD_PRIO_INHERIT => undef,
HAVE_PWRITE => undef,
+ HAVE_PWRITEV => undef,
HAVE_RANDOM => undef,
HAVE_READLINE_H => undef,
HAVE_READLINE_HISTORY_H => undef,
HAVE_READLINE_READLINE_H => undef,
HAVE_READLINK => undef,
+ HAVE_READV => undef,
HAVE_RL_COMPLETION_APPEND_CHARACTER => undef,
HAVE_RL_COMPLETION_MATCHES => undef,
HAVE_RL_COMPLETION_SUPPRESS_QUOTE => undef,
@@ -400,6 +403,7 @@ sub GenerateFiles
HAVE_SYS_TYPES_H => 1,
HAVE_SYS_UCRED_H => undef,
HAVE_SYS_UN_H => undef,
+ HAVE_SYS_UIO_H => undef,
HAVE_TERMIOS_H => undef,
HAVE_TYPEOF => undef,
HAVE_UCRED_H => undef,
@@ -417,6 +421,7 @@ sub GenerateFiles
HAVE_WINLDAP_H => undef,
HAVE_WCSTOMBS_L => 1,
HAVE_WCTYPE_H => 1,
+ HAVE_WRITEV => undef,
HAVE_X509_GET_SIGNATURE_NID => 1,
HAVE_X86_64_POPCNTQ => undef,
HAVE__BOOL => undef,
--
2.20.1
