From: Jeff Mahoney <jeffm@suse.com>
Subject: initramfs: add initramfs_{read,write}
References: bnc#568120
Patch-mainline: Probably never

 This patch adds initramfs_read and initramfs_write, which will read and
 write to the initramfs without traversing huge chunks of the VFS code.
 A previous incarnation of the ACPI dynamic table patches ended up
 causing "scheduling while atomic" warnings during boot, resulting in a
 whole lot of bug reports.

Signed-off-by: Jeff Mahoney <jeffm@suse.com>
---
 init/initramfs.c |  150 +++++++++++++++++++++++++++++++++++++++++++++++++++++--
 1 file changed, 147 insertions(+), 3 deletions(-)

--- a/init/initramfs.c
+++ b/init/initramfs.c
@@ -1,5 +1,6 @@
 #include <linux/init.h>
 #include <linux/fs.h>
+#include <linux/file.h>
 #include <linux/slab.h>
 #include <linux/types.h>
 #include <linux/fcntl.h>
@@ -8,6 +9,8 @@
 #include <linux/dirent.h>
 #include <linux/syscalls.h>
 #include <linux/utime.h>
+#include <linux/pagemap.h>
+#include <linux/uio.h>
 
 static __initdata char *message;
 static void __init error(char *x)
@@ -333,10 +335,152 @@ static int __init do_name(void)
 	return 0;
 }
 
+ssize_t initramfs_file_read(struct file *file, const char *buf,
+			    size_t count, loff_t *ppos)
+{
+	struct address_space *mapping = file->f_mapping;
+	struct iovec iov = { .iov_base = (void __user *) buf,
+			     .iov_len = count };
+	struct iov_iter i;
+	long status = 0;
+	loff_t pos = *ppos;
+	ssize_t read = 0;
+
+	iov_iter_init(&i, &iov, 1, count, 0);
+
+	do {
+		struct page *page;
+		pgoff_t index;
+		unsigned long offset;
+		unsigned long bytes;
+		char *data;
+
+		offset = (pos & (PAGE_CACHE_SIZE - 1));
+		index = pos >> PAGE_CACHE_SHIFT;
+		bytes = min_t(unsigned long, PAGE_CACHE_SIZE - offset,
+			      iov_iter_count(&i));
+
+		page = read_mapping_page(mapping, index, NULL);
+		if (IS_ERR(page)) {
+			status = PTR_ERR(page);
+			break;
+		}
+
+		data = page_address(page);
+		memcpy(i.iov->iov_base + i.iov_offset, data + offset, bytes);
+
+		iov_iter_advance(&i, bytes);
+		pos += bytes;
+		read += bytes;
+	} while (iov_iter_count(&i));
+
+	*ppos = pos;
+
+	return read ? read : status;
+}
+
+ssize_t initramfs_file_write(struct file *file, const char * __user buf,
+			     size_t count, loff_t *ppos)
+{
+	struct address_space *mapping = file->f_mapping;
+	struct iovec iov = { .iov_base = (void __user *) buf,
+			     .iov_len = count };
+	long status = 0;
+	ssize_t written = 0;
+	unsigned int flags = 0;
+	loff_t pos = *ppos;
+	struct iov_iter i;
+
+	iov_iter_init(&i, &iov, 1, count, 0);
+
+	/*
+	 * Copies from kernel address space cannot fail (NFSD is a big user).
+	 */
+	if (segment_eq(get_fs(), KERNEL_DS))
+		flags |= AOP_FLAG_UNINTERRUPTIBLE;
+
+	mutex_lock(&mapping->host->i_mutex);
+
+	do {
+		struct page *page;
+		pgoff_t index;		/* Pagecache index for current page */
+		unsigned long offset;	/* Offset into pagecache page */
+		unsigned long bytes;	/* Bytes to write to page */
+		size_t copied;		/* Bytes copied from user */
+		void *fsdata;
+		char *data;
+
+		offset = (pos & (PAGE_CACHE_SIZE - 1));
+		index = pos >> PAGE_CACHE_SHIFT;
+		bytes = min_t(unsigned long, PAGE_CACHE_SIZE - offset,
+						iov_iter_count(&i));
+
+		status = simple_write_begin(file, mapping, pos, bytes, flags,
+						&page, &fsdata);
+		if (unlikely(status))
+			break;
+		data = page_address(page);
+
+		memcpy(data + offset, i.iov->iov_base + i.iov_offset, bytes);
+		copied = bytes;
+
+		status = simple_write_end(file, mapping, pos, bytes, copied,
+						page, fsdata);
+		if (unlikely(status < 0))
+			break;
+		copied = status;
+
+		iov_iter_advance(&i, copied);
+		pos += copied;
+		written += copied;
+
+	} while (iov_iter_count(&i));
+
+	mutex_unlock(&mapping->host->i_mutex);
+
+	*ppos = pos;
+
+	return written ? written : status;
+}
+
+ssize_t
+initramfs_read(unsigned int fd, const char * buf, size_t count)
+{
+	struct file *file;
+	ssize_t ret = 0;
+
+	file = fget(fd);
+	if (file) {
+		loff_t pos = file->f_pos;
+		ret = initramfs_file_read(file, buf, count, &pos);
+		file->f_pos = pos;
+		fput(file);
+	}
+
+	return ret;
+}
+
+ssize_t
+initramfs_write(unsigned int fd, const char * buf, size_t count)
+{
+	struct file *file;
+	ssize_t ret = 0;
+
+	file = fget(fd);
+	if (file) {
+		loff_t pos = file->f_pos;
+		ret = initramfs_file_write(file, buf, count, &pos);
+		file->f_pos = pos;
+		fput(file);
+	}
+
+	return ret;
+}
+
 static int __init do_copy(void)
 {
 	if (count >= body_len) {
-		sys_write(wfd, victim, body_len);
+		initramfs_write(wfd, victim, body_len);
 		sys_close(wfd);
 		do_utime(vcollected, mtime);
 		kfree(vcollected);
@@ -344,7 +488,7 @@ static int __init do_copy(void)
 		state = SkipIt;
 		return 0;
 	} else {
-		sys_write(wfd, victim, count);
+		initramfs_write(wfd, victim, count);
 		body_len -= count;
 		eat(count);
 		return 1;
@@ -589,7 +733,7 @@ static int __init populate_rootfs(void)
 				"; looks like an initrd\n", err);
 		fd = sys_open("/initrd.image", O_WRONLY|O_CREAT, 0700);
 		if (fd >= 0) {
-			sys_write(fd, (char *)initrd_start,
+			initramfs_write(fd, (char *)initrd_start,
 					initrd_end - initrd_start);
 			sys_close(fd);
 			free_initrd();