#include <linux/pipe_fs_i.h>
#include <linux/mm_inline.h>
#include <linux/swap.h>
+#include <linux/writeback.h>
+#include <linux/buffer_head.h>
#include <linux/module.h>
+#include <linux/syscalls.h>
/*
* Passed to the actors
loff_t pos; /* file position */
};
+/*
+ * Attempt to steal a page from a pipe buffer. This should perhaps go into
+ * a vm helper function, it's already simplified quite a bit by the
+ * addition of remove_mapping(). If success is returned, the caller may
+ * attempt to reuse this page for another destination.
+ */
static int page_cache_pipe_buf_steal(struct pipe_inode_info *info,
struct pipe_buffer *buf)
{
struct page *page = buf->page;
+ struct address_space *mapping = page_mapping(page);
WARN_ON(!PageLocked(page));
WARN_ON(!PageUptodate(page));
- if (!remove_mapping(page_mapping(page), page))
- return 1;
+ /*
+ * At least for ext2 with nobh option, we need to wait on writeback
+ * completing on this page, since we'll remove it from the pagecache.
+ * Otherwise truncate wont wait on the page, allowing the disk
+ * blocks to be reused by someone else before we actually wrote our
+ * data to them. fs corruption ensues.
+ */
+ wait_on_page_writeback(page);
- if (PageLRU(page)) {
- struct zone *zone = page_zone(page);
+ if (PagePrivate(page))
+ try_to_release_page(page, mapping_gfp_mask(mapping));
- spin_lock_irq(&zone->lru_lock);
- BUG_ON(!PageLRU(page));
- __ClearPageLRU(page);
- del_page_from_lru(zone, page);
- spin_unlock_irq(&zone->lru_lock);
- }
+ if (!remove_mapping(mapping, page))
+ return 1;
- buf->stolen = 1;
+ buf->flags |= PIPE_BUF_FLAG_STOLEN | PIPE_BUF_FLAG_LRU;
return 0;
}
{
page_cache_release(buf->page);
buf->page = NULL;
- buf->stolen = 0;
+ buf->flags &= ~(PIPE_BUF_FLAG_STOLEN | PIPE_BUF_FLAG_LRU);
}
static void *page_cache_pipe_buf_map(struct file *file,
struct pipe_buffer *buf)
{
struct page *page = buf->page;
-
- lock_page(page);
+ int err;
if (!PageUptodate(page)) {
- unlock_page(page);
- return ERR_PTR(-EIO);
- }
+ lock_page(page);
+
+ /*
+ * Page got truncated/unhashed. This will cause a 0-byte
+ * splice, if this is the first page
+ */
+ if (!page->mapping) {
+ err = -ENODATA;
+ goto error;
+ }
- if (!page->mapping) {
+ /*
+ * uh oh, read-error from disk
+ */
+ if (!PageUptodate(page)) {
+ err = -EIO;
+ goto error;
+ }
+
+ /*
+ * page is ok afterall, fall through to mapping
+ */
unlock_page(page);
- return ERR_PTR(-ENODATA);
}
- return kmap(buf->page);
+ return kmap(page);
+error:
+ unlock_page(page);
+ return ERR_PTR(err);
}
static void page_cache_pipe_buf_unmap(struct pipe_inode_info *info,
struct pipe_buffer *buf)
{
- if (!buf->stolen)
- unlock_page(buf->page);
kunmap(buf->page);
}
.steal = page_cache_pipe_buf_steal,
};
-static ssize_t move_to_pipe(struct inode *inode, struct page **pages,
+/*
+ * Pipe output worker. This sets up our pipe format with the page cache
+ * pipe buffer operations. Otherwise very similar to the regular pipe_writev().
+ */
+static ssize_t move_to_pipe(struct pipe_inode_info *pipe, struct page **pages,
int nr_pages, unsigned long offset,
unsigned long len, unsigned int flags)
{
- struct pipe_inode_info *info;
int ret, do_wakeup, i;
ret = 0;
do_wakeup = 0;
i = 0;
- mutex_lock(PIPE_MUTEX(*inode));
+ if (pipe->inode)
+ mutex_lock(&pipe->inode->i_mutex);
- info = inode->i_pipe;
for (;;) {
int bufs;
- if (!PIPE_READERS(*inode)) {
+ if (!pipe->readers) {
send_sig(SIGPIPE, current, 0);
if (!ret)
ret = -EPIPE;
break;
}
- bufs = info->nrbufs;
+ bufs = pipe->nrbufs;
if (bufs < PIPE_BUFFERS) {
- int newbuf = (info->curbuf + bufs) & (PIPE_BUFFERS - 1);
- struct pipe_buffer *buf = info->bufs + newbuf;
+ int newbuf = (pipe->curbuf + bufs) & (PIPE_BUFFERS - 1);
+ struct pipe_buffer *buf = pipe->bufs + newbuf;
struct page *page = pages[i++];
unsigned long this_len;
buf->offset = offset;
buf->len = this_len;
buf->ops = &page_cache_pipe_buf_ops;
- info->nrbufs = ++bufs;
+ pipe->nrbufs = ++bufs;
do_wakeup = 1;
ret += this_len;
}
if (do_wakeup) {
- wake_up_interruptible_sync(PIPE_WAIT(*inode));
- kill_fasync(PIPE_FASYNC_READERS(*inode), SIGIO,
- POLL_IN);
+ smp_mb();
+ if (waitqueue_active(&pipe->wait))
+ wake_up_interruptible_sync(&pipe->wait);
+ kill_fasync(&pipe->fasync_readers, SIGIO, POLL_IN);
do_wakeup = 0;
}
- PIPE_WAITING_WRITERS(*inode)++;
- pipe_wait(inode);
- PIPE_WAITING_WRITERS(*inode)--;
+ pipe->waiting_writers++;
+ pipe_wait(pipe);
+ pipe->waiting_writers--;
}
- mutex_unlock(PIPE_MUTEX(*inode));
+ if (pipe->inode)
+ mutex_unlock(&pipe->inode->i_mutex);
if (do_wakeup) {
- wake_up_interruptible(PIPE_WAIT(*inode));
- kill_fasync(PIPE_FASYNC_READERS(*inode), SIGIO, POLL_IN);
+ smp_mb();
+ if (waitqueue_active(&pipe->wait))
+ wake_up_interruptible(&pipe->wait);
+ kill_fasync(&pipe->fasync_readers, SIGIO, POLL_IN);
}
while (i < nr_pages)
return ret;
}
-static int __generic_file_splice_read(struct file *in, struct inode *pipe,
- size_t len, unsigned int flags)
+static int
+__generic_file_splice_read(struct file *in, struct pipe_inode_info *pipe,
+ size_t len, unsigned int flags)
{
struct address_space *mapping = in->f_mapping;
unsigned int offset, nr_pages;
- struct page *pages[PIPE_BUFFERS], *shadow[PIPE_BUFFERS];
+ struct page *pages[PIPE_BUFFERS];
struct page *page;
- pgoff_t index, pidx;
- int i, j;
+ pgoff_t index;
+ int i;
index = in->f_pos >> PAGE_CACHE_SHIFT;
offset = in->f_pos & ~PAGE_CACHE_MASK;
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
+ * initiate read-ahead on this page range. however, don't call into
+ * read-ahead if this is a non-zero offset (we are likely doing small
+ * chunk splice and the page is already there) for a single page.
*/
- memset(shadow, 0, i * sizeof(struct page *));
- for (j = 0, pidx = index; j < i; pidx++, j++)
- shadow[pages[j]->index - pidx] = pages[j];
+ if (!offset || nr_pages > 1)
+ do_page_cache_readahead(mapping, in, index, nr_pages);
/*
* now fill in the holes
*/
- for (i = 0, pidx = index; i < nr_pages; pidx++, i++) {
- int error;
-
- if (shadow[i])
- continue;
-
+ for (i = 0; i < nr_pages; i++, index++) {
/*
* no page there, look one up / create it
*/
- page = find_or_create_page(mapping, pidx,
+ page = find_or_create_page(mapping, index,
mapping_gfp_mask(mapping));
if (!page)
break;
if (PageUptodate(page))
unlock_page(page);
else {
- error = mapping->a_ops->readpage(in, page);
+ int error = mapping->a_ops->readpage(in, page);
if (unlikely(error)) {
page_cache_release(page);
break;
}
}
- shadow[i] = page;
+ pages[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 *));
+ if (i)
+ return move_to_pipe(pipe, pages, i, offset, len, flags);
- /*
- * Now we splice them into the pipe..
- */
-splice_them:
- return move_to_pipe(pipe, pages, i, offset, len, flags);
+ return 0;
}
-ssize_t generic_file_splice_read(struct file *in, struct inode *pipe,
+/**
+ * generic_file_splice_read - splice data from file to a pipe
+ * @in: file to splice from
+ * @pipe: pipe to splice to
+ * @len: number of bytes to splice
+ * @flags: splice modifier flags
+ *
+ * Will read pages from given file and fill them into a pipe.
+ *
+ */
+ssize_t generic_file_splice_read(struct file *in, struct pipe_inode_info *pipe,
size_t len, unsigned int flags)
{
ssize_t spliced;
ret = 0;
spliced = 0;
+
while (len) {
ret = __generic_file_splice_read(in, pipe, len, flags);
return ret;
}
+EXPORT_SYMBOL(generic_file_splice_read);
+
/*
- * Send 'len' bytes to socket from 'file' at position 'pos' using sendpage().
+ * Send 'sd->len' bytes to socket from 'sd->file' at position 'sd->pos'
+ * using sendpage().
*/
static int pipe_to_sendpage(struct pipe_inode_info *info,
struct pipe_buffer *buf, struct splice_desc *sd)
unsigned int offset;
ssize_t ret;
void *ptr;
+ int more;
/*
* sub-optimal, but we are limited by the pipe ->map. we don't
return PTR_ERR(ptr);
offset = pos & ~PAGE_CACHE_MASK;
+ more = (sd->flags & SPLICE_F_MORE) || sd->len < sd->total_len;
- ret = file->f_op->sendpage(file, buf->page, offset, sd->len, &pos,
- sd->len < sd->total_len);
+ ret = file->f_op->sendpage(file, buf->page, offset, sd->len, &pos,more);
buf->ops->unmap(info, buf);
if (ret == sd->len)
* - 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...
+ * If asked to move pages to the output file (SPLICE_F_MOVE is set in
+ * sd->flags), we attempt to migrate pages from the pipe to the output
+ * file address space page cache. This is possible if no one else has
+ * the pipe page referenced outside of the pipe and page cache. If
+ * SPLICE_F_MOVE isn't set, or we cannot move the page, we simply create
+ * a new page in the output file page cache and fill/dirty that.
*/
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;
+ gfp_t gfp_mask = mapping_gfp_mask(mapping);
unsigned int offset;
struct page *page;
pgoff_t index;
int ret;
/*
- * after this, page will be locked and unmapped
+ * make sure the data in this buffer is uptodate
*/
src = buf->ops->map(file, info, buf);
if (IS_ERR(src))
* reuse buf page, if SPLICE_F_MOVE is set
*/
if (sd->flags & SPLICE_F_MOVE) {
+ /*
+ * If steal succeeds, buf->page is now pruned from the vm
+ * side (LRU and page cache) and we can reuse it.
+ */
if (buf->ops->steal(info, buf))
goto find_page;
+ /*
+ * this will also set the page locked
+ */
page = buf->page;
- if (add_to_page_cache_lru(page, mapping, index,
- mapping_gfp_mask(mapping)))
+ if (add_to_page_cache(page, mapping, index, gfp_mask))
goto find_page;
+
+ if (!(buf->flags & PIPE_BUF_FLAG_LRU))
+ lru_cache_add(page);
} else {
find_page:
ret = -ENOMEM;
- page = find_or_create_page(mapping, index,
- mapping_gfp_mask(mapping));
+ page = find_or_create_page(mapping, index, gfp_mask);
if (!page)
- goto out;
+ goto out_nomem;
/*
* If the page is uptodate, it is also locked. If it isn't
}
ret = mapping->a_ops->prepare_write(file, page, 0, sd->len);
- if (ret)
+ if (ret == AOP_TRUNCATED_PAGE) {
+ page_cache_release(page);
+ goto find_page;
+ } else if (ret)
goto out;
- if (!buf->stolen) {
+ if (!(buf->flags & PIPE_BUF_FLAG_STOLEN)) {
char *dst = kmap_atomic(page, KM_USER0);
memcpy(dst + offset, src + buf->offset, sd->len);
}
ret = mapping->a_ops->commit_write(file, page, 0, sd->len);
- if (ret < 0)
+ if (ret == AOP_TRUNCATED_PAGE) {
+ page_cache_release(page);
+ goto find_page;
+ } else if (ret)
goto out;
- set_page_dirty(page);
- ret = write_one_page(page, 0);
+ mark_page_accessed(page);
+ balance_dirty_pages_ratelimited(mapping);
out:
- if (ret < 0)
- unlock_page(page);
- if (!buf->stolen)
+ if (!(buf->flags & PIPE_BUF_FLAG_STOLEN)) {
page_cache_release(page);
+ unlock_page(page);
+ }
+out_nomem:
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,
+/*
+ * Pipe input worker. Most of this logic works like a regular pipe, the
+ * key here is the 'actor' worker passed in that actually moves the data
+ * to the wanted destination. See pipe_to_file/pipe_to_sendpage above.
+ */
+static ssize_t move_from_pipe(struct pipe_inode_info *pipe, 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;
sd.file = out;
sd.pos = out->f_pos;
- mutex_lock(PIPE_MUTEX(*inode));
+ if (pipe->inode)
+ mutex_lock(&pipe->inode->i_mutex);
- info = inode->i_pipe;
for (;;) {
- int bufs = info->nrbufs;
+ int bufs = pipe->nrbufs;
if (bufs) {
- int curbuf = info->curbuf;
- struct pipe_buffer *buf = info->bufs + curbuf;
+ int curbuf = pipe->curbuf;
+ struct pipe_buffer *buf = pipe->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);
+ err = actor(pipe, buf, &sd);
if (err) {
if (!ret && err != -ENODATA)
ret = err;
buf->len -= sd.len;
if (!buf->len) {
buf->ops = NULL;
- ops->release(info, buf);
+ ops->release(pipe, buf);
curbuf = (curbuf + 1) & (PIPE_BUFFERS - 1);
- info->curbuf = curbuf;
- info->nrbufs = --bufs;
+ pipe->curbuf = curbuf;
+ pipe->nrbufs = --bufs;
do_wakeup = 1;
}
if (bufs)
continue;
- if (!PIPE_WRITERS(*inode))
+ if (!pipe->writers)
break;
- if (!PIPE_WAITING_WRITERS(*inode)) {
+ if (!pipe->waiting_writers) {
if (ret)
break;
}
}
if (do_wakeup) {
- wake_up_interruptible_sync(PIPE_WAIT(*inode));
- kill_fasync(PIPE_FASYNC_WRITERS(*inode),SIGIO,POLL_OUT);
+ smp_mb();
+ if (waitqueue_active(&pipe->wait))
+ wake_up_interruptible_sync(&pipe->wait);
+ kill_fasync(&pipe->fasync_writers, SIGIO, POLL_OUT);
do_wakeup = 0;
}
- pipe_wait(inode);
+ pipe_wait(pipe);
}
- mutex_unlock(PIPE_MUTEX(*inode));
+ if (pipe->inode)
+ mutex_unlock(&pipe->inode->i_mutex);
if (do_wakeup) {
- wake_up_interruptible(PIPE_WAIT(*inode));
- kill_fasync(PIPE_FASYNC_WRITERS(*inode), SIGIO, POLL_OUT);
+ smp_mb();
+ if (waitqueue_active(&pipe->wait))
+ wake_up_interruptible(&pipe->wait);
+ kill_fasync(&pipe->fasync_writers, SIGIO, POLL_OUT);
}
mutex_lock(&out->f_mapping->host->i_mutex);
}
-ssize_t generic_file_splice_write(struct inode *inode, struct file *out,
- size_t len, unsigned int flags)
+/**
+ * generic_file_splice_write - splice data from a pipe to a file
+ * @pipe: pipe info
+ * @out: file to write to
+ * @len: number of bytes to splice
+ * @flags: splice modifier flags
+ *
+ * Will either move or copy pages (determined by @flags options) from
+ * the given pipe inode to the given file.
+ *
+ */
+ssize_t
+generic_file_splice_write(struct pipe_inode_info *pipe, struct file *out,
+ size_t len, unsigned int flags)
{
- return move_from_pipe(inode, out, len, flags, pipe_to_file);
+ struct address_space *mapping = out->f_mapping;
+ ssize_t ret;
+
+ ret = move_from_pipe(pipe, out, len, flags, pipe_to_file);
+
+ /*
+ * if file or inode is SYNC and we actually wrote some data, sync it
+ */
+ if (unlikely((out->f_flags & O_SYNC) || IS_SYNC(mapping->host))
+ && ret > 0) {
+ struct inode *inode = mapping->host;
+ int err;
+
+ mutex_lock(&inode->i_mutex);
+ err = generic_osync_inode(mapping->host, mapping,
+ OSYNC_METADATA|OSYNC_DATA);
+ mutex_unlock(&inode->i_mutex);
+
+ if (err)
+ ret = err;
+ }
+
+ return ret;
}
-ssize_t generic_splice_sendpage(struct inode *inode, struct file *out,
+EXPORT_SYMBOL(generic_file_splice_write);
+
+/**
+ * generic_splice_sendpage - splice data from a pipe to a socket
+ * @inode: pipe inode
+ * @out: socket to write to
+ * @len: number of bytes to splice
+ * @flags: splice modifier flags
+ *
+ * Will send @len bytes from the pipe to a network socket. No data copying
+ * is involved.
+ *
+ */
+ssize_t generic_splice_sendpage(struct pipe_inode_info *pipe, struct file *out,
size_t len, unsigned int flags)
{
- return move_from_pipe(inode, out, len, flags, pipe_to_sendpage);
+ return move_from_pipe(pipe, out, len, flags, pipe_to_sendpage);
}
-EXPORT_SYMBOL(generic_file_splice_write);
-EXPORT_SYMBOL(generic_file_splice_read);
+EXPORT_SYMBOL(generic_splice_sendpage);
-static long do_splice_from(struct inode *pipe, struct file *out, size_t len,
- unsigned int flags)
+/*
+ * Attempt to initiate a splice from pipe to file.
+ */
+static long do_splice_from(struct pipe_inode_info *pipe, struct file *out,
+ size_t len, unsigned int flags)
{
loff_t pos;
int ret;
return out->f_op->splice_write(pipe, out, len, flags);
}
-static long do_splice_to(struct file *in, struct inode *pipe, size_t len,
- unsigned int flags)
+/*
+ * Attempt to initiate a splice from a file to a pipe.
+ */
+static long do_splice_to(struct file *in, struct pipe_inode_info *pipe,
+ size_t len, unsigned int flags)
{
loff_t pos, isize, left;
int ret;
return in->f_op->splice_read(in, pipe, len, flags);
}
+/*
+ * Determine where to splice to/from.
+ */
static long do_splice(struct file *in, struct file *out, size_t len,
unsigned int flags)
{
- struct inode *pipe;
+ struct pipe_inode_info *pipe;
- pipe = in->f_dentry->d_inode;
- if (pipe->i_pipe)
+ pipe = in->f_dentry->d_inode->i_pipe;
+ if (pipe)
return do_splice_from(pipe, out, len, flags);
- pipe = out->f_dentry->d_inode;
- if (pipe->i_pipe)
+ pipe = out->f_dentry->d_inode->i_pipe;
+ if (pipe)
return do_splice_to(in, pipe, len, flags);
return -EINVAL;