Hello, developers.
Here is simple no-way-zero-copy support for sendfile
for socket <-> socket connections.
It is quite convenient for programmer to not use
recv/sendfile, but sendfile/sendfile instead.
If this way is usefull and should be cleaned up,
I will finish it and probably implement sendpage()
for fs, so it could be really usefull, else just FYI.
Thanks.
Here is quite simple and stupid implemetation:
diff --git a/fs/read_write.c b/fs/read_write.c
--- a/fs/read_write.c
+++ b/fs/read_write.c
@@ -685,7 +685,7 @@ static ssize_t do_sendfile(int out_fd, i
retval = -EINVAL;
if (unlikely(pos < 0))
goto fput_out;
- if (unlikely(pos + count > max)) {
+ if (unlikely((unsigned long long)(pos + count) > (unsigned long
long)max)) {
retval = -EOVERFLOW;
if (pos >= max)
goto fput_out;
diff --git a/net/socket.c b/net/socket.c
--- a/net/socket.c
+++ b/net/socket.c
@@ -44,6 +44,8 @@
* Tigran Aivazian : sys_send(args) calls sys_sendto(args,
NULL, 0)
* Tigran Aivazian : Made listen(2) backlog sanity checks
* protocol-independent
+ * Evgeniy Polyakov: Added recv sendfile path for
socket<->socket
+ * communications, i.e. sock_sendfile().
*
*
* This program is free software; you can redistribute it and/or
@@ -114,6 +116,7 @@ static ssize_t sock_writev(struct file *
unsigned long count, loff_t *ppos);
static ssize_t sock_sendpage(struct file *file, struct page *page,
int offset, size_t size, loff_t *ppos, int more);
+ssize_t sock_sendfile(struct file *file, loff_t *ppos, size_t count,
read_actor_t actor, void *target);
/*
@@ -134,7 +137,8 @@ static struct file_operations socket_fil
.fasync = sock_fasync,
.readv = sock_readv,
.writev = sock_writev,
- .sendpage = sock_sendpage
+ .sendpage = sock_sendpage,
+ .sendfile = sock_sendfile,
};
/*
@@ -720,6 +724,81 @@ static ssize_t sock_aio_write(struct kio
return __sock_sendmsg(iocb, sock, &x->async_msg, size);
}
+ssize_t sock_sendfile(struct file *file, loff_t *ppos, size_t count,
read_actor_t actor, void *target)
+{
+ struct socket *sock;
+ struct page *page;
+ int err = 0;
+ struct msghdr msg;
+ struct kvec iov;
+ size_t size;
+ read_descriptor_t desc;
+ unsigned long offset = 0;
+ size_t orig_count = count;
+
+ printk("%s: ppos=%Lx, count=%lx.\n", __func__, *ppos, count);
+
+ if (!count)
+ return 0;
+
+ if (!ppos != 0)
+ return -ERANGE;
+
+ desc.written = 0;
+ desc.count = count;
+ desc.arg.data = target;
+ desc.error = 0;
+
+ page = alloc_page(GFP_KERNEL);
+ if (!page)
+ return -ENOMEM;
+
+ sock = SOCKET_I(file->f_dentry->d_inode);
+
+ while (count) {
+
+ size = min(count, PAGE_SIZE);
+
+ sock->sk->sk_allocation |= GFP_NOIO;
+ iov.iov_base = page_address(page);
+ iov.iov_len = size;
+ msg.msg_name = NULL;
+ msg.msg_namelen = 0;
+ msg.msg_control = NULL;
+ msg.msg_controllen = 0;
+ msg.msg_namelen = 0;
+ msg.msg_flags = MSG_NOSIGNAL;
+
+ err = kernel_recvmsg(sock, &msg, &iov, 1, size, 0);
+
+ if (signal_pending(current)) {
+ flush_signals(current);
+ printk("Interrupted by signal\n");
+ return -ERESTARTSYS;
+ }
+
+ if (err <= 0) {
+ printk("Failed to receive message: size=%lu,
err=%d.\n", iov.iov_len, err);
+
+ if (!err)
+ err = -EPIPE;
+ break;
+ }
+
+ count -= err;
+
+ err = actor(&desc, page, offset, size);
+ *ppos += err;
+ }
+
+ __free_pages(page, 0);
+
+ if (err)
+ return err;
+
+ return orig_count;
+}
+
ssize_t sock_sendpage(struct file *file, struct page *page,
int offset, size_t size, loff_t *ppos, int more)
{
--
Evgeniy Polyakov
-
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
