X-Git-Url: https://err.no/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=fs%2Fpipe.c;h=20352573e025f0fe2574b6ffb300c44fc3605fa9;hb=8d79d088e88198d5456861ee9e6a8226dcd08799;hp=b941e1951eac3b24ca7af49cdec09b0c311db2cd;hpb=923f4f23940d2361e8d5c4245982163a8e9d1c91;p=linux-2.6

diff --git a/fs/pipe.c b/fs/pipe.c
index b941e1951e..20352573e0 100644
--- a/fs/pipe.c
+++ b/fs/pipe.c
@@ -44,7 +44,8 @@ void pipe_wait(struct pipe_inode_info *pipe)
 	 * Pipes are system-local resources, so sleeping on them
 	 * is considered a noninteractive wait:
 	 */
-	prepare_to_wait(&pipe->wait, &wait, TASK_INTERRUPTIBLE|TASK_NONINTERACTIVE);
+	prepare_to_wait(&pipe->wait, &wait,
+			TASK_INTERRUPTIBLE | TASK_NONINTERACTIVE);
 	if (pipe->inode)
 		mutex_unlock(&pipe->inode->i_mutex);
 	schedule();
@@ -54,7 +55,8 @@ void pipe_wait(struct pipe_inode_info *pipe)
 }
 
 static int
-pipe_iov_copy_from_user(void *to, struct iovec *iov, unsigned long len)
+pipe_iov_copy_from_user(void *to, struct iovec *iov, unsigned long len,
+			int atomic)
 {
 	unsigned long copy;
 
@@ -63,8 +65,13 @@ pipe_iov_copy_from_user(void *to, struct iovec *iov, unsigned long len)
 			iov++;
 		copy = min_t(unsigned long, len, iov->iov_len);
 
-		if (copy_from_user(to, iov->iov_base, copy))
-			return -EFAULT;
+		if (atomic) {
+			if (__copy_from_user_inatomic(to, iov->iov_base, copy))
+				return -EFAULT;
+		} else {
+			if (copy_from_user(to, iov->iov_base, copy))
+				return -EFAULT;
+		}
 		to += copy;
 		len -= copy;
 		iov->iov_base += copy;
@@ -74,7 +81,8 @@ pipe_iov_copy_from_user(void *to, struct iovec *iov, unsigned long len)
 }
 
 static int
-pipe_iov_copy_to_user(struct iovec *iov, const void *from, unsigned long len)
+pipe_iov_copy_to_user(struct iovec *iov, const void *from, unsigned long len,
+		      int atomic)
 {
 	unsigned long copy;
 
@@ -83,8 +91,13 @@ pipe_iov_copy_to_user(struct iovec *iov, const void *from, unsigned long len)
 			iov++;
 		copy = min_t(unsigned long, len, iov->iov_len);
 
-		if (copy_to_user(iov->iov_base, from, copy))
-			return -EFAULT;
+		if (atomic) {
+			if (__copy_to_user_inatomic(iov->iov_base, from, copy))
+				return -EFAULT;
+		} else {
+			if (copy_to_user(iov->iov_base, from, copy))
+				return -EFAULT;
+		}
 		from += copy;
 		len -= copy;
 		iov->iov_base += copy;
@@ -93,51 +106,115 @@ pipe_iov_copy_to_user(struct iovec *iov, const void *from, unsigned long len)
 	return 0;
 }
 
-static void anon_pipe_buf_release(struct pipe_inode_info *pipe, struct pipe_buffer *buf)
+/*
+ * Attempt to pre-fault in the user memory, so we can use atomic copies.
+ * Returns the number of bytes not faulted in.
+ */
+static int iov_fault_in_pages_write(struct iovec *iov, unsigned long len)
 {
-	struct page *page = buf->page;
+	while (!iov->iov_len)
+		iov++;
+
+	while (len > 0) {
+		unsigned long this_len;
+
+		this_len = min_t(unsigned long, len, iov->iov_len);
+		if (fault_in_pages_writeable(iov->iov_base, this_len))
+			break;
+
+		len -= this_len;
+		iov++;
+	}
+
+	return len;
+}
+
+/*
+ * Pre-fault in the user memory, so we can use atomic copies.
+ */
+static void iov_fault_in_pages_read(struct iovec *iov, unsigned long len)
+{
+	while (!iov->iov_len)
+		iov++;
 
-	buf->flags &= ~PIPE_BUF_FLAG_STOLEN;
+	while (len > 0) {
+		unsigned long this_len;
+
+		this_len = min_t(unsigned long, len, iov->iov_len);
+		fault_in_pages_readable(iov->iov_base, this_len);
+		len -= this_len;
+		iov++;
+	}
+}
+
+static void anon_pipe_buf_release(struct pipe_inode_info *pipe,
+				  struct pipe_buffer *buf)
+{
+	struct page *page = buf->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
+	 * allocation cache. (Otherwise just release our reference to it)
 	 */
-	if (page_count(page) == 1 && !pipe->tmp_page) {
+	if (page_count(page) == 1 && !pipe->tmp_page)
 		pipe->tmp_page = page;
-		return;
+	else
+		page_cache_release(page);
+}
+
+void *generic_pipe_buf_map(struct pipe_inode_info *pipe,
+			   struct pipe_buffer *buf, int atomic)
+{
+	if (atomic) {
+		buf->flags |= PIPE_BUF_FLAG_ATOMIC;
+		return kmap_atomic(buf->page, KM_USER0);
 	}
 
-	/*
-	 * Otherwise just release our reference to it
-	 */
-	page_cache_release(page);
+	return kmap(buf->page);
 }
 
-static void *anon_pipe_buf_map(struct file *file, struct pipe_inode_info *pipe, struct pipe_buffer *buf)
+void generic_pipe_buf_unmap(struct pipe_inode_info *pipe,
+			    struct pipe_buffer *buf, void *map_data)
 {
-	return kmap(buf->page);
+	if (buf->flags & PIPE_BUF_FLAG_ATOMIC) {
+		buf->flags &= ~PIPE_BUF_FLAG_ATOMIC;
+		kunmap_atomic(map_data, KM_USER0);
+	} else
+		kunmap(buf->page);
+}
+
+int generic_pipe_buf_steal(struct pipe_inode_info *pipe,
+			   struct pipe_buffer *buf)
+{
+	struct page *page = buf->page;
+
+	if (page_count(page) == 1) {
+		lock_page(page);
+		return 0;
+	}
+
+	return 1;
 }
 
-static void anon_pipe_buf_unmap(struct pipe_inode_info *pipe, struct pipe_buffer *buf)
+void generic_pipe_buf_get(struct pipe_inode_info *info, struct pipe_buffer *buf)
 {
-	kunmap(buf->page);
+	page_cache_get(buf->page);
 }
 
-static int anon_pipe_buf_steal(struct pipe_inode_info *pipe,
-			       struct pipe_buffer *buf)
+int generic_pipe_buf_pin(struct pipe_inode_info *info, struct pipe_buffer *buf)
 {
-	buf->flags |= PIPE_BUF_FLAG_STOLEN;
 	return 0;
 }
 
 static struct pipe_buf_operations anon_pipe_buf_ops = {
 	.can_merge = 1,
-	.map = anon_pipe_buf_map,
-	.unmap = anon_pipe_buf_unmap,
+	.map = generic_pipe_buf_map,
+	.unmap = generic_pipe_buf_unmap,
+	.pin = generic_pipe_buf_pin,
 	.release = anon_pipe_buf_release,
-	.steal = anon_pipe_buf_steal,
+	.steal = generic_pipe_buf_steal,
+	.get = generic_pipe_buf_get,
 };
 
 static ssize_t
@@ -168,21 +245,33 @@ pipe_readv(struct file *filp, const struct iovec *_iov,
 			struct pipe_buf_operations *ops = buf->ops;
 			void *addr;
 			size_t chars = buf->len;
-			int error;
+			int error, atomic;
 
 			if (chars > total_len)
 				chars = total_len;
 
-			addr = ops->map(filp, pipe, buf);
-			if (IS_ERR(addr)) {
+			error = ops->pin(pipe, buf);
+			if (error) {
 				if (!ret)
-					ret = PTR_ERR(addr);
+					error = ret;
 				break;
 			}
-			error = pipe_iov_copy_to_user(iov, addr + buf->offset, chars);
-			ops->unmap(pipe, buf);
+
+			atomic = !iov_fault_in_pages_write(iov, chars);
+redo:
+			addr = ops->map(pipe, buf, atomic);
+			error = pipe_iov_copy_to_user(iov, addr + buf->offset, chars, atomic);
+			ops->unmap(pipe, buf, addr);
 			if (unlikely(error)) {
-				if (!ret) ret = -EFAULT;
+				/*
+				 * Just retry with the slow path if we failed.
+				 */
+				if (atomic) {
+					atomic = 0;
+					goto redo;
+				}
+				if (!ret)
+					ret = error;
 				break;
 			}
 			ret += chars;
@@ -218,7 +307,8 @@ pipe_readv(struct file *filp, const struct iovec *_iov,
 			}
 		}
 		if (signal_pending(current)) {
-			if (!ret) ret = -ERESTARTSYS;
+			if (!ret)
+				ret = -ERESTARTSYS;
 			break;
 		}
 		if (do_wakeup) {
@@ -228,7 +318,8 @@ pipe_readv(struct file *filp, const struct iovec *_iov,
 		pipe_wait(pipe);
 	}
 	mutex_unlock(&inode->i_mutex);
-	/* Signal writers asynchronously that there is more room.  */
+
+	/* Signal writers asynchronously that there is more room. */
 	if (do_wakeup) {
 		wake_up_interruptible(&pipe->wait);
 		kill_fasync(&pipe->fasync_writers, SIGIO, POLL_OUT);
@@ -242,6 +333,7 @@ static ssize_t
 pipe_read(struct file *filp, char __user *buf, size_t count, loff_t *ppos)
 {
 	struct iovec iov = { .iov_base = buf, .iov_len = count };
+
 	return pipe_readv(filp, &iov, 1, ppos);
 }
 
@@ -276,26 +368,35 @@ pipe_writev(struct file *filp, const struct iovec *_iov,
 	/* We try to merge small writes */
 	chars = total_len & (PAGE_SIZE-1); /* size of the last buffer */
 	if (pipe->nrbufs && chars != 0) {
-		int lastbuf = (pipe->curbuf + pipe->nrbufs - 1) & (PIPE_BUFFERS-1);
+		int lastbuf = (pipe->curbuf + pipe->nrbufs - 1) &
+							(PIPE_BUFFERS-1);
 		struct pipe_buffer *buf = pipe->bufs + lastbuf;
 		struct pipe_buf_operations *ops = buf->ops;
 		int offset = buf->offset + buf->len;
+
 		if (ops->can_merge && offset + chars <= PAGE_SIZE) {
+			int error, atomic = 1;
 			void *addr;
-			int error;
 
-			addr = ops->map(filp, pipe, buf);
-			if (IS_ERR(addr)) {
-				error = PTR_ERR(addr);
+			error = ops->pin(pipe, buf);
+			if (error)
 				goto out;
-			}
+
+			iov_fault_in_pages_read(iov, chars);
+redo1:
+			addr = ops->map(pipe, buf, atomic);
 			error = pipe_iov_copy_from_user(offset + addr, iov,
-							chars);
-			ops->unmap(pipe, buf);
+							chars, atomic);
+			ops->unmap(pipe, buf, addr);
 			ret = error;
 			do_wakeup = 1;
-			if (error)
+			if (error) {
+				if (atomic) {
+					atomic = 0;
+					goto redo1;
+				}
 				goto out;
+			}
 			buf->len += chars;
 			total_len -= chars;
 			ret = chars;
@@ -306,9 +407,11 @@ pipe_writev(struct file *filp, const struct iovec *_iov,
 
 	for (;;) {
 		int bufs;
+
 		if (!pipe->readers) {
 			send_sig(SIGPIPE, current, 0);
-			if (!ret) ret = -EPIPE;
+			if (!ret)
+				ret = -EPIPE;
 			break;
 		}
 		bufs = pipe->nrbufs;
@@ -316,7 +419,8 @@ pipe_writev(struct file *filp, const struct iovec *_iov,
 			int newbuf = (pipe->curbuf + bufs) & (PIPE_BUFFERS-1);
 			struct pipe_buffer *buf = pipe->bufs + newbuf;
 			struct page *page = pipe->tmp_page;
-			int error;
+			char *src;
+			int error, atomic = 1;
 
 			if (!page) {
 				page = alloc_page(GFP_HIGHUSER);
@@ -326,7 +430,7 @@ pipe_writev(struct file *filp, const struct iovec *_iov,
 				}
 				pipe->tmp_page = page;
 			}
-			/* Always wakeup, even if the copy fails. Otherwise
+			/* Always wake up, even if the copy fails. Otherwise
 			 * we lock up (O_NONBLOCK-)readers that sleep due to
 			 * syscall merging.
 			 * FIXME! Is this really true?
@@ -336,10 +440,27 @@ pipe_writev(struct file *filp, const struct iovec *_iov,
 			if (chars > total_len)
 				chars = total_len;
 
-			error = pipe_iov_copy_from_user(kmap(page), iov, chars);
-			kunmap(page);
+			iov_fault_in_pages_read(iov, chars);
+redo2:
+			if (atomic)
+				src = kmap_atomic(page, KM_USER0);
+			else
+				src = kmap(page);
+
+			error = pipe_iov_copy_from_user(src, iov, chars,
+							atomic);
+			if (atomic)
+				kunmap_atomic(src, KM_USER0);
+			else
+				kunmap(page);
+
 			if (unlikely(error)) {
-				if (!ret) ret = -EFAULT;
+				if (atomic) {
+					atomic = 0;
+					goto redo2;
+				}
+				if (!ret)
+					ret = error;
 				break;
 			}
 			ret += chars;
@@ -359,11 +480,13 @@ pipe_writev(struct file *filp, const struct iovec *_iov,
 		if (bufs < PIPE_BUFFERS)
 			continue;
 		if (filp->f_flags & O_NONBLOCK) {
-			if (!ret) ret = -EAGAIN;
+			if (!ret)
+				ret = -EAGAIN;
 			break;
 		}
 		if (signal_pending(current)) {
-			if (!ret) ret = -ERESTARTSYS;
+			if (!ret)
+				ret = -ERESTARTSYS;
 			break;
 		}
 		if (do_wakeup) {
@@ -391,6 +514,7 @@ pipe_write(struct file *filp, const char __user *buf,
 	   size_t count, loff_t *ppos)
 {
 	struct iovec iov = { .iov_base = (void __user *)buf, .iov_len = count };
+
 	return pipe_writev(filp, &iov, 1, ppos);
 }
 
@@ -401,7 +525,8 @@ bad_pipe_r(struct file *filp, char __user *buf, size_t count, loff_t *ppos)
 }
 
 static ssize_t
-bad_pipe_w(struct file *filp, const char __user *buf, size_t count, loff_t *ppos)
+bad_pipe_w(struct file *filp, const char __user *buf, size_t count,
+	   loff_t *ppos)
 {
 	return -EBADF;
 }
@@ -475,6 +600,7 @@ pipe_release(struct inode *inode, int decr, int decw)
 	pipe = inode->i_pipe;
 	pipe->readers -= decr;
 	pipe->writers -= decw;
+
 	if (!pipe->readers && !pipe->writers) {
 		free_pipe_info(inode);
 	} else {
@@ -525,14 +651,15 @@ static int
 pipe_rdwr_fasync(int fd, struct file *filp, int on)
 {
 	struct inode *inode = filp->f_dentry->d_inode;
+	struct pipe_inode_info *pipe = inode->i_pipe;
 	int retval;
 
 	mutex_lock(&inode->i_mutex);
 
-	retval = fasync_helper(fd, filp, on, &inode->i_pipe->fasync_readers);
+	retval = fasync_helper(fd, filp, on, &pipe->fasync_readers);
 
 	if (retval >= 0)
-		retval = fasync_helper(fd, filp, on, &inode->i_pipe->fasync_writers);
+		retval = fasync_helper(fd, filp, on, &pipe->fasync_writers);
 
 	mutex_unlock(&inode->i_mutex);
 
@@ -720,6 +847,7 @@ static int pipefs_delete_dentry(struct dentry *dentry)
 {
 	return 1;
 }
+
 static struct dentry_operations pipefs_dentry_operations = {
 	.d_delete	= pipefs_delete_dentry,
 };
@@ -757,6 +885,7 @@ static struct inode * get_pipe_inode(void)
 
 fail_iput:
 	iput(inode);
+
 fail_inode:
 	return NULL;
 }
@@ -769,7 +898,7 @@ int do_pipe(int *fd)
 	struct inode * inode;
 	struct file *f1, *f2;
 	int error;
-	int i,j;
+	int i, j;
 
 	error = -ENFILE;
 	f1 = get_empty_filp();
@@ -802,6 +931,7 @@ int do_pipe(int *fd)
 	dentry = d_alloc(pipe_mnt->mnt_sb->s_root, &this);
 	if (!dentry)
 		goto close_f12_inode_i_j;
+
 	dentry->d_op = &pipefs_dentry_operations;
 	d_add(dentry, inode);
 	f1->f_vfsmnt = f2->f_vfsmnt = mntget(mntget(pipe_mnt));
@@ -825,6 +955,7 @@ int do_pipe(int *fd)
 	fd_install(j, f2);
 	fd[0] = i;
 	fd[1] = j;
+
 	return 0;
 
 close_f12_inode_i_j:
@@ -848,11 +979,11 @@ no_files:
  * any operations on the root directory. However, we need a non-trivial
  * d_name - pipe: will go nicely and kill the special-casing in procfs.
  */
-
-static struct super_block *pipefs_get_sb(struct file_system_type *fs_type,
-	int flags, const char *dev_name, void *data)
+static int pipefs_get_sb(struct file_system_type *fs_type,
+			 int flags, const char *dev_name, void *data,
+			 struct vfsmount *mnt)
 {
-	return get_sb_pseudo(fs_type, "pipe:", NULL, PIPEFS_MAGIC);
+	return get_sb_pseudo(fs_type, "pipe:", NULL, PIPEFS_MAGIC, mnt);
 }
 
 static struct file_system_type pipe_fs_type = {
@@ -864,6 +995,7 @@ static struct file_system_type pipe_fs_type = {
 static int __init init_pipe_fs(void)
 {
 	int err = register_filesystem(&pipe_fs_type);
+
 	if (!err) {
 		pipe_mnt = kern_mount(&pipe_fs_type);
 		if (IS_ERR(pipe_mnt)) {