Add SWAP support

src/Makefile.am

@@ -76,6 +76,7 @@ libzmq_la_SOURCES = app_thread.hpp \
     lb.hpp \
     likely.hpp \
     msg_content.hpp \
+    msg_store.hpp \
     mutex.hpp \
     object.hpp \
     options.hpp \
@@ -134,6 +135,7 @@ libzmq_la_SOURCES = app_thread.hpp \
     ip.cpp \
     kqueue.cpp \
     lb.cpp \
+    msg_store.cpp \
     object.cpp \
     options.cpp \
     owned.cpp \

src/msg_store.cpp

@@ -0,0 +1,313 @@
+/*
+    Copyright (c) 2007-2010 iMatix Corporation
+
+    This file is part of 0MQ.
+
+    0MQ is free software; you can redistribute it and/or modify it under
+    the terms of the Lesser GNU General Public License as published by
+    the Free Software Foundation; either version 3 of the License, or
+    (at your option) any later version.
+
+    0MQ is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    Lesser GNU General Public License for more details.
+
+    You should have received a copy of the Lesser GNU General Public License
+    along with this program.  If not, see <http://www.gnu.org/licenses/>.
+*/
+
+#include "../include/zmq.h"
+
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <fcntl.h>
+#include <string.h>
+#include <sstream>
+#include <algorithm>
+
+#ifdef ZMQ_HAVE_WINDOWS
+#include <io.h>
+#else
+#include <unistd.h>
+#endif
+
+#include "atomic_counter.hpp"
+#include "msg_store.hpp"
+#include "err.hpp"
+
+zmq::msg_store_t::msg_store_t (int64_t filesize_, size_t block_size_) :
+    fd (-1),
+    filesize (filesize_),
+    file_pos (0),
+    write_pos (0),
+    read_pos (0),
+    block_size (block_size_),
+    write_buf_start_addr (0)
+{
+    zmq_assert (filesize > 0);
+    zmq_assert (block_size > 0);
+
+    buf1 = new (std::nothrow) char [block_size];
+    zmq_assert (buf1);
+
+    buf2 = new (std::nothrow) char [block_size];
+    zmq_assert (buf2);
+
+    read_buf = write_buf = buf1;
+}
+
+zmq::msg_store_t::~msg_store_t ()
+{
+    delete [] buf1;
+    delete [] buf2;
+
+    if (fd == -1)
+        return;
+
+#ifdef ZMQ_HAVE_WINDOWS
+    int rc = _close (fd);
+#else
+    int rc = close (fd);
+#endif
+    errno_assert (rc == 0);
+
+#ifdef ZMQ_HAVE_WINDOWS
+    rc = _unlink (filename.c_str ());
+#else
+    rc = unlink (filename.c_str ());
+#endif
+    errno_assert (rc == 0);
+}
+
+int zmq::msg_store_t::init ()
+{
+    static zmq::atomic_counter_t seqnum (0);
+
+    //  Get process ID.
+#ifdef ZMQ_HAVE_WINDOWS
+    int pid = GetCurrentThreadId ();
+#else
+    pid_t pid = getpid ();
+#endif
+
+    std::ostringstream outs;
+    outs << "zmq_" << pid << '_' << seqnum.get () << ".swap";
+    filename = outs.str ();
+
+    seqnum.add (1);
+
+    //  Open the backing file.
+#ifdef ZMQ_HAVE_WINDOWS
+    fd = _open (filename.c_str (), _O_RDWR | _O_CREAT, 0600);
+#else
+    fd = open (filename.c_str (), O_RDWR | O_CREAT, 0600);
+#endif
+    if (fd == -1)
+        return -1;
+
+#ifdef ZMQ_HAVE_LINUX
+    //  Enable more aggresive read-ahead optimization.
+    posix_fadvise (fd, 0, filesize, POSIX_FADV_SEQUENTIAL);
+#endif
+    return 0;
+}
+
+bool zmq::msg_store_t::store (zmq_msg_t *msg_)
+{
+    size_t msg_size = zmq_msg_size (msg_);
+
+    //  Check buffer space availability.
+    //  NOTE: We always keep one byte open.
+    if (buffer_space () <= (int64_t) (sizeof msg_size + 1 + msg_size))
+        return false;
+
+    //  Don't store the ZMQ_MSG_SHARED flag.
+    uint8_t msg_flags = msg_->flags & ~ZMQ_MSG_SHARED;
+
+    //  Write message length, flags, and message body.
+    copy_to_file (&msg_size, sizeof msg_size);
+    copy_to_file (&msg_flags, sizeof msg_flags);
+    copy_to_file (zmq_msg_data (msg_), msg_size);
+
+    zmq_msg_close (msg_);
+
+    return true;
+}
+
+void zmq::msg_store_t::fetch (zmq_msg_t *msg_)
+{
+    //  There must be at least one message available.
+    zmq_assert (read_pos != write_pos);
+
+    //  Retrieve the message size.
+    size_t msg_size;
+    copy_from_file (&msg_size, sizeof msg_size);
+
+    //  Initialize the message.
+    zmq_msg_init_size (msg_, msg_size);
+
+    //  Retrieve the message flags.
+    copy_from_file (&msg_->flags, sizeof msg_->flags);
+
+    //  Retrieve the message payload.
+    copy_from_file (zmq_msg_data (msg_), msg_size);
+}
+
+void zmq::msg_store_t::commit ()
+{
+    commit_pos = write_pos;
+}
+
+void zmq::msg_store_t::rollback ()
+{
+    if (commit_pos == write_pos || read_pos == write_pos)
+        return;
+
+
+    if (write_pos > read_pos)
+        zmq_assert (read_pos <= commit_pos && commit_pos <= write_pos);
+    else
+        zmq_assert (read_pos <= commit_pos || commit_pos <= write_pos);
+
+    if (commit_pos / block_size == read_pos / block_size) {
+        write_buf_start_addr = commit_pos % block_size;
+        write_buf = read_buf;
+    }
+    else if (commit_pos / block_size != write_pos / block_size) {
+        write_buf_start_addr = commit_pos % block_size;
+        fill_buf (write_buf, write_buf_start_addr);
+    }
+
+    write_pos = commit_pos;
+}
+
+bool zmq::msg_store_t::empty ()
+{
+    return read_pos == write_pos;
+}
+
+bool zmq::msg_store_t::full ()
+{
+    return buffer_space () == 1;
+}
+
+void zmq::msg_store_t::copy_from_file (void *buffer_, size_t count_)
+{
+    char *ptr = (char*) buffer_;
+    size_t n, n_left = count_;
+
+    while (n_left > 0) {
+
+        n = std::min (n_left, std::min ((size_t) (filesize - read_pos),
+            (size_t) (block_size - read_pos % block_size)));
+
+        memcpy (ptr, &read_buf [read_pos % block_size], n);
+        ptr += n;
+
+        read_pos = (read_pos + n) % filesize;
+        if (read_pos % block_size == 0) {
+            if (read_pos / block_size == write_pos / block_size)
+                read_buf = write_buf;
+            else
+                fill_buf (read_buf, read_pos);
+        }
+
+        n_left -= n;
+    }
+}
+
+void zmq::msg_store_t::copy_to_file (const void *buffer_, size_t count_)
+{
+    char *ptr = (char*) buffer_;
+    size_t n, n_left = count_;
+
+    while (n_left > 0) {
+
+        n = std::min (n_left, std::min ((size_t) (filesize - write_pos),
+            (size_t) (block_size - write_pos % block_size)));
+
+        memcpy (&write_buf [write_pos % block_size], ptr, n);
+        ptr += n;
+
+        write_pos = (write_pos + n) % filesize;
+        if (write_pos % block_size == 0) {
+
+            save_write_buf ();
+            write_buf_start_addr = write_pos;
+
+            if (write_buf == read_buf) {
+                if (read_buf == buf2)
+                    write_buf = buf1;
+                else
+                    write_buf = buf2;
+            }
+        }
+
+        n_left -= n;
+    }
+}
+
+void zmq::msg_store_t::fill_buf (char *buf, int64_t pos)
+{
+    if (file_pos != pos) {
+#ifdef ZMQ_HAVE_WINDOWS
+        __int64 offset = _lseeki64 (fd, pos, SEEK_SET);
+#else
+        off_t offset = lseek (fd, (off_t) pos, SEEK_SET);
+#endif
+        errno_assert (offset == pos);
+        file_pos = pos;
+    }
+
+    size_t i = 0;
+    size_t n = std::min (block_size, (size_t) (filesize - file_pos));
+
+    while (i < n) {
+#ifdef ZMQ_HAVE_WINDOWS
+        int rc = _read (fd, &buf [i], n - i);
+#else
+        ssize_t rc = read (fd, &buf [i], n - i);
+#endif
+        errno_assert (rc > 0);
+        i += rc;
+    }
+
+    file_pos += n;
+}
+
+void zmq::msg_store_t::save_write_buf ()
+{
+    if (file_pos != write_buf_start_addr) {
+#ifdef ZMQ_HAVE_WINDOWS
+        __int64 offset = _lseeki64 (fd, write_buf_start_addr, SEEK_SET);
+#else
+        off_t offset = lseek (fd, (off_t) write_buf_start_addr, SEEK_SET);
+#endif
+        errno_assert (offset == write_buf_start_addr);
+        file_pos = write_buf_start_addr;
+    }
+
+    size_t i = 0;
+    size_t n = std::min (block_size, (size_t) (filesize - file_pos));
+
+    while (i < n) {
+#ifdef ZMQ_HAVE_WINDOWS
+        int rc = _write (fd, &write_buf [i], n - i);
+#else
+        ssize_t rc = write (fd, &write_buf [i], n - i);
+#endif
+        errno_assert (rc > 0);
+        i += rc;
+    }
+
+    file_pos += n;
+}
+
+int64_t zmq::msg_store_t::buffer_space ()
+{
+    if (write_pos < read_pos)
+        return read_pos - write_pos;
+
+    return filesize - (write_pos - read_pos);
+}