1 files changed, 88 insertions, 49 deletions

diff --git a/fs/overlayfs/file.c b/fs/overlayfs/file.c
index 21245b00722a..131621daeb13 100644
--- a/fs/overlayfs/file.c
+++ b/fs/overlayfs/file.c
@@ -15,11 +15,15 @@
 #include <linux/fs.h>
 #include "overlayfs.h"
 
+#include "../internal.h"	/* for sb_init_dio_done_wq */
+
 struct ovl_aio_req {
 	struct kiocb iocb;
 	refcount_t ref;
 	struct kiocb *orig_iocb;
-	struct fd fd;
+	/* used for aio completion */
+	struct work_struct work;
+	long res;
 };
 
 static struct kmem_cache *ovl_aio_request_cachep;
@@ -115,8 +119,8 @@ static int ovl_real_fdget_meta(const struct file *file, struct fd *real,
 	if (allow_meta) {
 		ovl_path_real(dentry, &realpath);
 	} else {
-		/* lazy lookup of lowerdata */
-		err = ovl_maybe_lookup_lowerdata(dentry);
+		/* lazy lookup and verify of lowerdata */
+		err = ovl_verify_lowerdata(dentry);
 		if (err)
 			return err;
 
@@ -159,8 +163,8 @@ static int ovl_open(struct inode *inode, struct file *file)
 	struct path realpath;
 	int err;
 
-	/* lazy lookup of lowerdata */
-	err = ovl_maybe_lookup_lowerdata(dentry);
+	/* lazy lookup and verify lowerdata */
+	err = ovl_verify_lowerdata(dentry);
 	if (err)
 		return err;
 
@@ -236,9 +240,17 @@ static loff_t ovl_llseek(struct file *file, loff_t offset, int whence)
 	return ret;
 }
 
+static void ovl_file_modified(struct file *file)
+{
+	/* Update size/mtime */
+	ovl_copyattr(file_inode(file));
+}
+
 static void ovl_file_accessed(struct file *file)
 {
 	struct inode *inode, *upperinode;
+	struct timespec64 ctime, uctime;
+	struct timespec64 mtime, umtime;
 
 	if (file->f_flags & O_NOATIME)
 		return;
@@ -249,35 +261,31 @@ static void ovl_file_accessed(struct file *file)
 	if (!upperinode)
 		return;
 
-	if ((!timespec64_equal(&inode->i_mtime, &upperinode->i_mtime) ||
-	     !timespec64_equal(&inode->i_ctime, &upperinode->i_ctime))) {
-		inode->i_mtime = upperinode->i_mtime;
-		inode->i_ctime = upperinode->i_ctime;
+	ctime = inode_get_ctime(inode);
+	uctime = inode_get_ctime(upperinode);
+	mtime = inode_get_mtime(inode);
+	umtime = inode_get_mtime(upperinode);
+	if ((!timespec64_equal(&mtime, &umtime)) ||
+	     !timespec64_equal(&ctime, &uctime)) {
+		inode_set_mtime_to_ts(inode, inode_get_mtime(upperinode));
+		inode_set_ctime_to_ts(inode, uctime);
 	}
 
 	touch_atime(&file->f_path);
 }
 
-static rwf_t ovl_iocb_to_rwf(int ifl)
+#define OVL_IOCB_MASK \
+	(IOCB_NOWAIT | IOCB_HIPRI | IOCB_DSYNC | IOCB_SYNC | IOCB_APPEND)
+
+static rwf_t iocb_to_rw_flags(int flags)
 {
-	rwf_t flags = 0;
-
-	if (ifl & IOCB_NOWAIT)
-		flags |= RWF_NOWAIT;
-	if (ifl & IOCB_HIPRI)
-		flags |= RWF_HIPRI;
-	if (ifl & IOCB_DSYNC)
-		flags |= RWF_DSYNC;
-	if (ifl & IOCB_SYNC)
-		flags |= RWF_SYNC;
-
-	return flags;
+	return (__force rwf_t)(flags & OVL_IOCB_MASK);
 }
 
 static inline void ovl_aio_put(struct ovl_aio_req *aio_req)
 {
 	if (refcount_dec_and_test(&aio_req->ref)) {
-		fdput(aio_req->fd);
+		fput(aio_req->iocb.ki_filp);
 		kmem_cache_free(ovl_aio_request_cachep, aio_req);
 	}
 }
@@ -288,13 +296,8 @@ static void ovl_aio_cleanup_handler(struct ovl_aio_req *aio_req)
 	struct kiocb *orig_iocb = aio_req->orig_iocb;
 
 	if (iocb->ki_flags & IOCB_WRITE) {
-		struct inode *inode = file_inode(orig_iocb->ki_filp);
-
-		/* Actually acquired in ovl_write_iter() */
-		__sb_writers_acquired(file_inode(iocb->ki_filp)->i_sb,
-				      SB_FREEZE_WRITE);
-		file_end_write(iocb->ki_filp);
-		ovl_copyattr(inode);
+		kiocb_end_write(iocb);
+		ovl_file_modified(orig_iocb->ki_filp);
 	}
 
 	orig_iocb->ki_pos = iocb->ki_pos;
@@ -311,6 +314,37 @@ static void ovl_aio_rw_complete(struct kiocb *iocb, long res)
 	orig_iocb->ki_complete(orig_iocb, res);
 }
 
+static void ovl_aio_complete_work(struct work_struct *work)
+{
+	struct ovl_aio_req *aio_req = container_of(work,
+						   struct ovl_aio_req, work);
+
+	ovl_aio_rw_complete(&aio_req->iocb, aio_req->res);
+}
+
+static void ovl_aio_queue_completion(struct kiocb *iocb, long res)
+{
+	struct ovl_aio_req *aio_req = container_of(iocb,
+						   struct ovl_aio_req, iocb);
+	struct kiocb *orig_iocb = aio_req->orig_iocb;
+
+	/*
+	 * Punt to a work queue to serialize updates of mtime/size.
+	 */
+	aio_req->res = res;
+	INIT_WORK(&aio_req->work, ovl_aio_complete_work);
+	queue_work(file_inode(orig_iocb->ki_filp)->i_sb->s_dio_done_wq,
+		   &aio_req->work);
+}
+
+static int ovl_init_aio_done_wq(struct super_block *sb)
+{
+	if (sb->s_dio_done_wq)
+		return 0;
+
+	return sb_init_dio_done_wq(sb);
+}
+
 static ssize_t ovl_read_iter(struct kiocb *iocb, struct iov_iter *iter)
 {
 	struct file *file = iocb->ki_filp;
@@ -332,8 +366,9 @@ static ssize_t ovl_read_iter(struct kiocb *iocb, struct iov_iter *iter)
 
 	old_cred = ovl_override_creds(file_inode(file)->i_sb);
 	if (is_sync_kiocb(iocb)) {
-		ret = vfs_iter_read(real.file, iter, &iocb->ki_pos,
-				    ovl_iocb_to_rwf(iocb->ki_flags));
+		rwf_t rwf = iocb_to_rw_flags(iocb->ki_flags);
+
+		ret = vfs_iter_read(real.file, iter, &iocb->ki_pos, rwf);
 	} else {
 		struct ovl_aio_req *aio_req;
 
@@ -342,10 +377,8 @@ static ssize_t ovl_read_iter(struct kiocb *iocb, struct iov_iter *iter)
 		if (!aio_req)
 			goto out;
 
-		aio_req->fd = real;
-		real.flags = 0;
 		aio_req->orig_iocb = iocb;
-		kiocb_clone(&aio_req->iocb, iocb, real.file);
+		kiocb_clone(&aio_req->iocb, iocb, get_file(real.file));
 		aio_req->iocb.ki_complete = ovl_aio_rw_complete;
 		refcount_set(&aio_req->ref, 2);
 		ret = vfs_iocb_iter_read(real.file, &aio_req->iocb, iter);
@@ -393,33 +426,39 @@ static ssize_t ovl_write_iter(struct kiocb *iocb, struct iov_iter *iter)
 	if (!ovl_should_sync(OVL_FS(inode->i_sb)))
 		ifl &= ~(IOCB_DSYNC | IOCB_SYNC);
 
+	/*
+	 * Overlayfs doesn't support deferred completions, don't copy
+	 * this property in case it is set by the issuer.
+	 */
+	ifl &= ~IOCB_DIO_CALLER_COMP;
+
 	old_cred = ovl_override_creds(file_inode(file)->i_sb);
 	if (is_sync_kiocb(iocb)) {
+		rwf_t rwf = iocb_to_rw_flags(ifl);
+
 		file_start_write(real.file);
-		ret = vfs_iter_write(real.file, iter, &iocb->ki_pos,
-				     ovl_iocb_to_rwf(ifl));
+		ret = vfs_iter_write(real.file, iter, &iocb->ki_pos, rwf);
 		file_end_write(real.file);
 		/* Update size */
-		ovl_copyattr(inode);
+		ovl_file_modified(file);
 	} else {
 		struct ovl_aio_req *aio_req;
 
+		ret = ovl_init_aio_done_wq(inode->i_sb);
+		if (ret)
+			goto out;
+
 		ret = -ENOMEM;
 		aio_req = kmem_cache_zalloc(ovl_aio_request_cachep, GFP_KERNEL);
 		if (!aio_req)
 			goto out;
 
-		file_start_write(real.file);
-		/* Pacify lockdep, same trick as done in aio_write() */
-		__sb_writers_release(file_inode(real.file)->i_sb,
-				     SB_FREEZE_WRITE);
-		aio_req->fd = real;
-		real.flags = 0;
 		aio_req->orig_iocb = iocb;
-		kiocb_clone(&aio_req->iocb, iocb, real.file);
+		kiocb_clone(&aio_req->iocb, iocb, get_file(real.file));
 		aio_req->iocb.ki_flags = ifl;
-		aio_req->iocb.ki_complete = ovl_aio_rw_complete;
+		aio_req->iocb.ki_complete = ovl_aio_queue_completion;
 		refcount_set(&aio_req->ref, 2);
+		kiocb_start_write(&aio_req->iocb);
 		ret = vfs_iocb_iter_write(real.file, &aio_req->iocb, iter);
 		ovl_aio_put(aio_req);
 		if (ret != -EIOCBQUEUED)
@@ -491,7 +530,7 @@ static ssize_t ovl_splice_write(struct pipe_inode_info *pipe, struct file *out,
 
 	file_end_write(real.file);
 	/* Update size */
-	ovl_copyattr(inode);
+	ovl_file_modified(out);
 	revert_creds(old_cred);
 	fdput(real);
 
@@ -572,7 +611,7 @@ static long ovl_fallocate(struct file *file, int mode, loff_t offset, loff_t len
 	revert_creds(old_cred);
 
 	/* Update size */
-	ovl_copyattr(inode);
+	ovl_file_modified(file);
 
 	fdput(real);
 
@@ -656,7 +695,7 @@ static loff_t ovl_copyfile(struct file *file_in, loff_t pos_in,
 	revert_creds(old_cred);
 
 	/* Update size */
-	ovl_copyattr(inode_out);
+	ovl_file_modified(file_out);
 
 	fdput(real_in);
 	fdput(real_out);