[PATCH] Introduce sys_splice() system call
Changes
15 files changed · +669 −11
| File | Status | Add | Del | Content |
|---|---|---|---|---|
| arch/i386/kernel/syscall_table.S | modified | +1 | −0 | View file |
| arch/ia64/kernel/entry.S | modified | +1 | −0 | View file |
| fs/Makefile | modified | +1 | −1 | View file |
| fs/ext2/file.c | modified | +2 | −0 | View file |
| fs/ext3/file.c | modified | +2 | −0 | View file |
| fs/pipe.c | modified | +28 | −5 | View file |
| fs/reiserfs/file.c | modified | +2 | −0 | View file |
| fs/splice.c | added | +612 | −0 | View file |
| include/asm-i386/unistd.h | modified | +2 | −1 | View file |
| include/asm-ia64/unistd.h | modified | +2 | −1 | View file |
| include/asm-powerpc/unistd.h | modified | +2 | −1 | View file |
| include/asm-x86_64/unistd.h | modified | +3 | −1 | View file |
| include/linux/fs.h | modified | +4 | −0 | View file |
| include/linux/syscalls.h | modified | +2 | −0 | View file |
| net/socket.c | modified | +5 | −1 | View file |
Only the first 600 diff lines are shown. Download the complete patch.
arch/i386/kernel/syscall_table.S
View filediff --git a/arch/i386/kernel/syscall_table.S b/arch/i386/kernel/syscall_table.S
index 326595f..ce3ef4f 100644
--- a/arch/i386/kernel/syscall_table.S
+++ b/arch/i386/kernel/syscall_table.S
@@ -312,3 +312,4 @@ ENTRY(sys_call_table)
.long sys_unshare /* 310 */
.long sys_set_robust_list
.long sys_get_robust_list
+ .long sys_splice
arch/ia64/kernel/entry.S
View filediff --git a/arch/ia64/kernel/entry.S b/arch/ia64/kernel/entry.S
index 0e3eda9..750e8e7 100644
--- a/arch/ia64/kernel/entry.S
+++ b/arch/ia64/kernel/entry.S
@@ -1605,5 +1605,6 @@ sys_call_table:
data8 sys_ni_syscall // reserved for pselect
data8 sys_ni_syscall // 1295 reserved for ppoll
data8 sys_unshare
+ data8 sys_splice
.org sys_call_table + 8*NR_syscalls // guard against failures to increase NR_syscalls
fs/Makefile
View filediff --git a/fs/Makefile b/fs/Makefile
index 080b386..f3a4f70 100644
--- a/fs/Makefile
+++ b/fs/Makefile
@@ -10,7 +10,7 @@ obj-y := open.o read_write.o file_table.o buffer.o bio.o super.o \
ioctl.o readdir.o select.o fifo.o locks.o dcache.o inode.o \
attr.o bad_inode.o file.o filesystems.o namespace.o aio.o \
seq_file.o xattr.o libfs.o fs-writeback.o mpage.o direct-io.o \
- ioprio.o pnode.o drop_caches.o
+ ioprio.o pnode.o drop_caches.o splice.o
obj-$(CONFIG_INOTIFY) += inotify.o
obj-$(CONFIG_EPOLL) += eventpoll.o
fs/ext2/file.c
View filediff --git a/fs/ext2/file.c b/fs/ext2/file.c
index 509ccec..23e2c7c 100644
--- a/fs/ext2/file.c
+++ b/fs/ext2/file.c
@@ -53,6 +53,8 @@ const struct file_operations ext2_file_operations = {
.readv = generic_file_readv,
.writev = generic_file_writev,
.sendfile = generic_file_sendfile,
+ .splice_read = generic_file_splice_read,
+ .splice_write = generic_file_splice_write,
};
#ifdef CONFIG_EXT2_FS_XIP
fs/ext3/file.c
View filediff --git a/fs/ext3/file.c b/fs/ext3/file.c
index 783a796..1efefb6 100644
--- a/fs/ext3/file.c
+++ b/fs/ext3/file.c
@@ -119,6 +119,8 @@ const struct file_operations ext3_file_operations = {
.release = ext3_release_file,
.fsync = ext3_sync_file,
.sendfile = generic_file_sendfile,
+ .splice_read = generic_file_splice_read,
+ .splice_write = generic_file_splice_write,
};
struct inode_operations ext3_file_inode_operations = {
fs/pipe.c
View filediff --git a/fs/pipe.c b/fs/pipe.c
index e2f4f1d..2414bf2 100644
--- a/fs/pipe.c
+++ b/fs/pipe.c
@@ -15,6 +15,7 @@
#include <linux/pipe_fs_i.h>
#include <linux/uio.h>
#include <linux/highmem.h>
+#include <linux/pagemap.h>
#include <asm/uaccess.h>
#include <asm/ioctls.h>
@@ -94,11 +95,20 @@ static void anon_pipe_buf_release(struct pipe_inode_info *info, struct pipe_buff
{
struct page *page = buf->page;
- if (info->tmp_page) {
- __free_page(page);
+ /*
+ * If nobody else uses this page, and we don't already have a
+ * temporary page, let's keep track of it as a one-deep
+ * allocation cache
+ */
+ if (page_count(page) == 1 && !info->tmp_page) {
+ info->tmp_page = page;
return;
}
- info->tmp_page = page;
+
+ /*
+ * Otherwise just release our reference to it
+ */
+ page_cache_release(page);
}
static void *anon_pipe_buf_map(struct file *file, struct pipe_inode_info *info, struct pipe_buffer *buf)
@@ -152,6 +162,11 @@ pipe_readv(struct file *filp, const struct iovec *_iov,
chars = total_len;
addr = ops->map(filp, info, buf);
+ if (IS_ERR(addr)) {
+ if (!ret)
+ ret = PTR_ERR(addr);
+ break;
+ }
error = pipe_iov_copy_to_user(iov, addr + buf->offset, chars);
ops->unmap(info, buf);
if (unlikely(error)) {
@@ -254,8 +269,16 @@ pipe_writev(struct file *filp, const struct iovec *_iov,
struct pipe_buf_operations *ops = buf->ops;
int offset = buf->offset + buf->len;
if (ops->can_merge && offset + chars <= PAGE_SIZE) {
- void *addr = ops->map(filp, info, buf);
- int error = pipe_iov_copy_from_user(offset + addr, iov, chars);
+ void *addr;
+ int error;
+
+ addr = ops->map(filp, info, buf);
+ if (IS_ERR(addr)) {
+ error = PTR_ERR(addr);
+ goto out;
+ }
+ error = pipe_iov_copy_from_user(offset + addr, iov,
+ chars);
ops->unmap(info, buf);
ret = error;
do_wakeup = 1;
fs/reiserfs/file.c
View filediff --git a/fs/reiserfs/file.c b/fs/reiserfs/file.c
index 010094d..cf6e1cf 100644
--- a/fs/reiserfs/file.c
+++ b/fs/reiserfs/file.c
@@ -1576,6 +1576,8 @@ const struct file_operations reiserfs_file_operations = {
.sendfile = generic_file_sendfile,
.aio_read = generic_file_aio_read,
.aio_write = reiserfs_aio_write,
+ .splice_read = generic_file_splice_read,
+ .splice_write = generic_file_splice_write,
};
struct inode_operations reiserfs_file_inode_operations = {
fs/splice.c
View filediff --git a/fs/splice.c b/fs/splice.c
new file mode 100644
index 0000000..efa47c1
--- /dev/null
+++ b/fs/splice.c
@@ -0,0 +1,612 @@
+/*
+ * "splice": joining two ropes together by interweaving their strands.
+ *
+ * This is the "extended pipe" functionality, where a pipe is used as
+ * an arbitrary in-memory buffer. Think of a pipe as a small kernel
+ * buffer that you can use to transfer data from one end to the other.
+ *
+ * The traditional unix read/write is extended with a "splice()" operation
+ * that transfers data buffers to or from a pipe buffer.
+ *
+ * Named by Larry McVoy, original implementation from Linus, extended by
+ * Jens to support splicing to files and fixing the initial implementation
+ * bugs.
+ *
+ * Copyright (C) 2005 Jens Axboe <[email protected]>
+ * Copyright (C) 2005 Linus Torvalds <[email protected]>
+ *
+ */
+#include <linux/fs.h>
+#include <linux/file.h>
+#include <linux/pagemap.h>
+#include <linux/pipe_fs_i.h>
+#include <linux/mm_inline.h>
+
+/*
+ * Passed to the actors
+ */
+struct splice_desc {
+ unsigned int len, total_len; /* current and remaining length */
+ unsigned int flags; /* splice flags */
+ struct file *file; /* file to read/write */
+ loff_t pos; /* file position */
+};
+
+static void page_cache_pipe_buf_release(struct pipe_inode_info *info,
+ struct pipe_buffer *buf)
+{
+ page_cache_release(buf->page);
+ buf->page = NULL;
+}
+
+static void *page_cache_pipe_buf_map(struct file *file,
+ struct pipe_inode_info *info,
+ struct pipe_buffer *buf)
+{
+ struct page *page = buf->page;
+
+ lock_page(page);
+
+ if (!PageUptodate(page)) {
+ unlock_page(page);
+ return ERR_PTR(-EIO);
+ }
+
+ if (!page->mapping) {
+ unlock_page(page);
+ return ERR_PTR(-ENODATA);
+ }
+
+ return kmap(buf->page);
+}
+
+static void page_cache_pipe_buf_unmap(struct pipe_inode_info *info,
+ struct pipe_buffer *buf)
+{
+ unlock_page(buf->page);
+ kunmap(buf->page);
+}
+
+static struct pipe_buf_operations page_cache_pipe_buf_ops = {
+ .can_merge = 0,
+ .map = page_cache_pipe_buf_map,
+ .unmap = page_cache_pipe_buf_unmap,
+ .release = page_cache_pipe_buf_release,
+};
+
+static ssize_t move_to_pipe(struct inode *inode, struct page **pages,
+ int nr_pages, unsigned long offset,
+ unsigned long len)
+{
+ struct pipe_inode_info *info;
+ int ret, do_wakeup, i;
+
+ ret = 0;
+ do_wakeup = 0;
+ i = 0;
+
+ mutex_lock(PIPE_MUTEX(*inode));
+
+ info = inode->i_pipe;
+ for (;;) {
+ int bufs;
+
+ if (!PIPE_READERS(*inode)) {
+ send_sig(SIGPIPE, current, 0);
+ if (!ret)
+ ret = -EPIPE;
+ break;
+ }
+
+ bufs = info->nrbufs;
+ if (bufs < PIPE_BUFFERS) {
+ int newbuf = (info->curbuf + bufs) & (PIPE_BUFFERS - 1);
+ struct pipe_buffer *buf = info->bufs + newbuf;
+ struct page *page = pages[i++];
+ unsigned long this_len;
+
+ this_len = PAGE_CACHE_SIZE - offset;
+ if (this_len > len)
+ this_len = len;
+
+ buf->page = page;
+ buf->offset = offset;
+ buf->len = this_len;
+ buf->ops = &page_cache_pipe_buf_ops;
+ info->nrbufs = ++bufs;
+ do_wakeup = 1;
+
+ ret += this_len;
+ len -= this_len;
+ offset = 0;
+ if (!--nr_pages)
+ break;
+ if (!len)
+ break;
+ if (bufs < PIPE_BUFFERS)
+ continue;
+
+ break;
+ }
+
+ if (signal_pending(current)) {
+ if (!ret)
+ ret = -ERESTARTSYS;
+ break;
+ }
+
+ if (do_wakeup) {
+ wake_up_interruptible_sync(PIPE_WAIT(*inode));
+ kill_fasync(PIPE_FASYNC_READERS(*inode), SIGIO,
+ POLL_IN);
+ do_wakeup = 0;
+ }
+
+ PIPE_WAITING_WRITERS(*inode)++;
+ pipe_wait(inode);
+ PIPE_WAITING_WRITERS(*inode)--;
+ }
+
+ mutex_unlock(PIPE_MUTEX(*inode));
+
+ if (do_wakeup) {
+ wake_up_interruptible(PIPE_WAIT(*inode));
+ kill_fasync(PIPE_FASYNC_READERS(*inode), SIGIO, POLL_IN);
+ }
+
+ while (i < nr_pages)
+ page_cache_release(pages[i++]);
+
+ return ret;
+}
+
+static int __generic_file_splice_read(struct file *in, struct inode *pipe,
+ size_t len)
+{
+ struct address_space *mapping = in->f_mapping;
+ unsigned int offset, nr_pages;
+ struct page *pages[PIPE_BUFFERS], *shadow[PIPE_BUFFERS];
+ struct page *page;
+ pgoff_t index, pidx;
+ int i, j;
+
+ index = in->f_pos >> PAGE_CACHE_SHIFT;
+ offset = in->f_pos & ~PAGE_CACHE_MASK;
+ nr_pages = (len + offset + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
+
+ if (nr_pages > PIPE_BUFFERS)
+ nr_pages = PIPE_BUFFERS;
+
+ /*
+ * initiate read-ahead on this page range
+ */
+ do_page_cache_readahead(mapping, in, index, nr_pages);
+
+ /*
+ * Get as many pages from the page cache as possible..
+ * Start IO on the page cache entries we create (we
+ * can assume that any pre-existing ones we find have
+ * already had IO started on them).
+ */
+ i = find_get_pages(mapping, index, nr_pages, pages);
+
+ /*
+ * common case - we found all pages and they are contiguous,
+ * kick them off
+ */
+ if (i && (pages[i - 1]->index == index + i - 1))
+ goto splice_them;
+
+ /*
+ * fill shadow[] with pages at the right locations, so we only
+ * have to fill holes
+ */
+ memset(shadow, 0, i * sizeof(struct page *));
+ for (j = 0, pidx = index; j < i; pidx++, j++)
+ shadow[pages[j]->index - pidx] = pages[j];
+
+ /*
+ * now fill in the holes
+ */
+ for (i = 0, pidx = index; i < nr_pages; pidx++, i++) {
+ int error;
+
+ if (shadow[i])
+ continue;
+
+ /*
+ * no page there, look one up / create it
+ */
+ page = find_or_create_page(mapping, pidx,
+ mapping_gfp_mask(mapping));
+ if (!page)
+ break;
+
+ if (PageUptodate(page))
+ unlock_page(page);
+ else {
+ error = mapping->a_ops->readpage(in, page);
+
+ if (unlikely(error)) {
+ page_cache_release(page);
+ break;
+ }
+ }
+ shadow[i] = page;
+ }
+
+ if (!i) {
+ for (i = 0; i < nr_pages; i++) {
+ if (shadow[i])
+ page_cache_release(shadow[i]);
+ }
+ return 0;
+ }
+
+ memcpy(pages, shadow, i * sizeof(struct page *));
+
+ /*
+ * Now we splice them into the pipe..
+ */
+splice_them:
+ return move_to_pipe(pipe, pages, i, offset, len);
+}
+
+ssize_t generic_file_splice_read(struct file *in, struct inode *pipe,
+ size_t len, unsigned int flags)
+{
+ ssize_t spliced;
+ int ret;
+
+ ret = 0;
+ spliced = 0;
+ while (len) {
+ ret = __generic_file_splice_read(in, pipe, len);
+
+ if (ret <= 0)
+ break;
+
+ in->f_pos += ret;
+ len -= ret;
+ spliced += ret;
+ }
+
+ if (spliced)
+ return spliced;
+
+ return ret;
+}
+
+/*
+ * Send 'len' bytes to socket from 'file' at position 'pos' using sendpage().
+ */
+static int pipe_to_sendpage(struct pipe_inode_info *info,
+ struct pipe_buffer *buf, struct splice_desc *sd)
+{
+ struct file *file = sd->file;
+ loff_t pos = sd->pos;
+ unsigned int offset;
+ ssize_t ret;
+ void *ptr;
+
+ /*
+ * sub-optimal, but we are limited by the pipe ->map. we don't
+ * need a kmap'ed buffer here, we just want to make sure we
+ * have the page pinned if the pipe page originates from the
+ * page cache
+ */
+ ptr = buf->ops->map(file, info, buf);
+ if (IS_ERR(ptr))
+ return PTR_ERR(ptr);
+
+ offset = pos & ~PAGE_CACHE_MASK;
+
+ ret = file->f_op->sendpage(file, buf->page, offset, sd->len, &pos,
+ sd->len < sd->total_len);
+
+ buf->ops->unmap(info, buf);
+ if (ret == sd->len)
+ return 0;
+
+ return -EIO;
+}
+
+/*
+ * This is a little more tricky than the file -> pipe splicing. There are
+ * basically three cases:
+ *
+ * - Destination page already exists in the address space and there
+ * are users of it. For that case we have no other option that
+ * copying the data. Tough luck.
+ * - Destination page already exists in the address space, but there
+ * are no users of it. Make sure it's uptodate, then drop it. Fall
+ * through to last case.
+ * - Destination page does not exist, we can add the pipe page to
+ * the page cache and avoid the copy.
+ *
+ * For now we just do the slower thing and always copy pages over, it's
+ * easier than migrating pages from the pipe to the target file. For the
+ * case of doing file | file splicing, the migrate approach had some LRU
+ * nastiness...
+ */
+static int pipe_to_file(struct pipe_inode_info *info, struct pipe_buffer *buf,
+ struct splice_desc *sd)
+{
+ struct file *file = sd->file;
+ struct address_space *mapping = file->f_mapping;
+ unsigned int offset;
+ struct page *page;
+ char *src, *dst;
+ pgoff_t index;
+ int ret;
+
+ /*
+ * after this, page will be locked and unmapped
+ */
+ src = buf->ops->map(file, info, buf);
+ if (IS_ERR(src))
+ return PTR_ERR(src);
+
+ index = sd->pos >> PAGE_CACHE_SHIFT;
+ offset = sd->pos & ~PAGE_CACHE_MASK;
+
+find_page:
+ ret = -ENOMEM;
+ page = find_or_create_page(mapping, index, mapping_gfp_mask(mapping));
+ if (!page)
+ goto out;
+
+ /*
+ * If the page is uptodate, it is also locked. If it isn't
+ * uptodate, we can mark it uptodate if we are filling the
+ * full page. Otherwise we need to read it in first...
+ */
+ if (!PageUptodate(page)) {
+ if (sd->len < PAGE_CACHE_SIZE) {
+ ret = mapping->a_ops->readpage(file, page);
+ if (unlikely(ret))
+ goto out;
+
+ lock_page(page);
+
+ if (!PageUptodate(page)) {
+ /*
+ * page got invalidated, repeat
+ */
+ if (!page->mapping) {
+ unlock_page(page);
+ page_cache_release(page);
+ goto find_page;
+ }
+ ret = -EIO;
+ goto out;
+ }
+ } else {
+ WARN_ON(!PageLocked(page));
+ SetPageUptodate(page);
+ }
+ }
+
+ ret = mapping->a_ops->prepare_write(file, page, 0, sd->len);
+ if (ret)
+ goto out;
+
+ dst = kmap_atomic(page, KM_USER0);
+ memcpy(dst + offset, src + buf->offset, sd->len);
+ flush_dcache_page(page);
+ kunmap_atomic(dst, KM_USER0);
+
+ ret = mapping->a_ops->commit_write(file, page, 0, sd->len);
+ if (ret < 0)
+ goto out;
+
+ set_page_dirty(page);
+ ret = write_one_page(page, 0);
+out:
+ if (ret < 0)
+ unlock_page(page);
+ page_cache_release(page);
+ buf->ops->unmap(info, buf);
+ return ret;
+}
+
+typedef int (splice_actor)(struct pipe_inode_info *, struct pipe_buffer *,
+ struct splice_desc *);
+
+static ssize_t move_from_pipe(struct inode *inode, struct file *out,
+ size_t len, unsigned int flags,
+ splice_actor *actor)
+{
+ struct pipe_inode_info *info;
+ int ret, do_wakeup, err;
+ struct splice_desc sd;
+
+ ret = 0;
+ do_wakeup = 0;
+
+ sd.total_len = len;
+ sd.flags = flags;
+ sd.file = out;
+ sd.pos = out->f_pos;
+
+ mutex_lock(PIPE_MUTEX(*inode));
+
+ info = inode->i_pipe;
+ for (;;) {
+ int bufs = info->nrbufs;
+
+ if (bufs) {
+ int curbuf = info->curbuf;
+ struct pipe_buffer *buf = info->bufs + curbuf;
+ struct pipe_buf_operations *ops = buf->ops;
+
+ sd.len = buf->len;
+ if (sd.len > sd.total_len)
+ sd.len = sd.total_len;
+
+ err = actor(info, buf, &sd);
+ if (err) {
+ if (!ret && err != -ENODATA)
+ ret = err;
+
+ break;
+ }
+
+ ret += sd.len;
This diff is incomplete. Download the complete patch.
include/asm-i386/unistd.h
View fileThis diff is incomplete. Download the complete patch.
include/asm-ia64/unistd.h
View fileThis diff is incomplete. Download the complete patch.
include/asm-powerpc/unistd.h
View fileThis diff is incomplete. Download the complete patch.
include/asm-x86_64/unistd.h
View fileThis diff is incomplete. Download the complete patch.
include/linux/fs.h
View fileThis diff is incomplete. Download the complete patch.
include/linux/syscalls.h
View fileThis diff is incomplete. Download the complete patch.
net/socket.c
View fileThis diff is incomplete. Download the complete patch.