[PATCH v2 0/4] Enable holes in copy_file_range()

[PATCH v2 0/4] Enable holes in copy_file_range()

[Goldwyn Rodrigues]

If copy_file_range performs a copy using splice, it converts holes
to zeros. This effort primarily changes this behavior to create
holes when it is possible. 

Even if copy_file_range() or clone_file_range() does not work for different
mounted filesystems, We should be able to splice files if they do not
belong the same super_block.

Changes since v1:
- Fixed bug when hole/data offset is farther than len
- [Amir] Refactor flags parameter

Changes since v0:
- [Amir] Carved out do_copy_file_range() which can be used by overlayfs

-- 
Goldwyn

[PATCH 1/4] copy_file_range: refactor vfs_copy_file_range

[Goldwyn Rodrigues]

From: Goldwyn Rodrigues <rgoldwyn@suse.com>

Preparatory patch to carve out do_copy_file_range() from
vfs_copy_file_range

Signed-off-by: Goldwyn Rodrigues <rgoldwyn@suse.com>
---
 fs/read_write.c | 60 ++++++++++++++++++++++++++++++++-------------------------
 1 file changed, 34 insertions(+), 26 deletions(-)

diff --git a/fs/read_write.c b/fs/read_write.c
index c4eabbfc90df..525f2a67e15a 100644
--- a/fs/read_write.c
+++ b/fs/read_write.c
@@ -1541,6 +1541,38 @@ COMPAT_SYSCALL_DEFINE4(sendfile64, int, out_fd, int, in_fd,
 }
 #endif
 
+static ssize_t do_copy_file_range(struct file *file_in, loff_t pos_in,
+			    struct file *file_out, loff_t pos_out,
+			    size_t len, unsigned int flags)
+{
+	ssize_t ret = 0;
+
+	if (len == 0)
+		return 0;
+
+	/*
+	 * Try cloning first, this is supported by more file systems, and
+	 * more efficient if both clone and copy are supported (e.g. NFS).
+	 */
+	if (file_in->f_op->clone_file_range) {
+		ret = file_in->f_op->clone_file_range(file_in, pos_in,
+				file_out, pos_out, len);
+		if (ret == 0)
+			return len;
+	}
+
+	if (file_out->f_op->copy_file_range) {
+		ret = file_out->f_op->copy_file_range(file_in, pos_in, file_out,
+						      pos_out, len, flags);
+		if (ret != -EOPNOTSUPP)
+			return ret;
+	}
+
+	ret = do_splice_direct(file_in, &pos_in, file_out, &pos_out,
+			len > MAX_RW_COUNT ? MAX_RW_COUNT : len, 0);
+	return ret;
+}
+
 /*
  * copy_file_range() differs from regular file read and write in that it
  * specifically allows return partial success.  When it does so is up to
@@ -1579,35 +1611,11 @@ ssize_t vfs_copy_file_range(struct file *file_in, loff_t pos_in,
 	if (inode_in->i_sb != inode_out->i_sb)
 		return -EXDEV;
 
-	if (len == 0)
-		return 0;
-
 	file_start_write(file_out);
 
-	/*
-	 * Try cloning first, this is supported by more file systems, and
-	 * more efficient if both clone and copy are supported (e.g. NFS).
-	 */
-	if (file_in->f_op->clone_file_range) {
-		ret = file_in->f_op->clone_file_range(file_in, pos_in,
-				file_out, pos_out, len);
-		if (ret == 0) {
-			ret = len;
-			goto done;
-		}
-	}
-
-	if (file_out->f_op->copy_file_range) {
-		ret = file_out->f_op->copy_file_range(file_in, pos_in, file_out,
-						      pos_out, len, flags);
-		if (ret != -EOPNOTSUPP)
-			goto done;
-	}
-
-	ret = do_splice_direct(file_in, &pos_in, file_out, &pos_out,
-			len > MAX_RW_COUNT ? MAX_RW_COUNT : len, 0);
+	ret = do_copy_file_range(file_in, pos_in,
+			file_out, pos_out, len, flags);
 
-done:
 	if (ret > 0) {
 		fsnotify_access(file_in);
 		add_rchar(current, ret);
-- 
2.16.3

[PATCH 2/4] copy_file_range: Perform splice if in/out SB are not same

[Goldwyn Rodrigues]

From: Goldwyn Rodrigues <rgoldwyn@suse.com>

While performing copy_file_range(), if superblocks of file_in and
file_out don't match, instead of returning -EXDEV, perform
splice for a faster copy.

Signed-off-by: Goldwyn Rodrigues <rgoldwyn@suse.com>
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
---
 fs/read_write.c | 11 ++++++-----
 1 file changed, 6 insertions(+), 5 deletions(-)

diff --git a/fs/read_write.c b/fs/read_write.c
index 525f2a67e15a..1b8fc9eada69 100644
--- a/fs/read_write.c
+++ b/fs/read_write.c
@@ -1545,11 +1545,16 @@ static ssize_t do_copy_file_range(struct file *file_in, loff_t pos_in,
 			    struct file *file_out, loff_t pos_out,
 			    size_t len, unsigned int flags)
 {
+	struct inode *inode_in = file_inode(file_in);
+	struct inode *inode_out = file_inode(file_out);
 	ssize_t ret = 0;
 
 	if (len == 0)
 		return 0;
 
+	if (inode_in->i_sb != inode_out->i_sb)
+		goto splice;
+
 	/*
 	 * Try cloning first, this is supported by more file systems, and
 	 * more efficient if both clone and copy are supported (e.g. NFS).
@@ -1567,7 +1572,7 @@ static ssize_t do_copy_file_range(struct file *file_in, loff_t pos_in,
 		if (ret != -EOPNOTSUPP)
 			return ret;
 	}
-
+splice:
 	ret = do_splice_direct(file_in, &pos_in, file_out, &pos_out,
 			len > MAX_RW_COUNT ? MAX_RW_COUNT : len, 0);
 	return ret;
@@ -1607,10 +1612,6 @@ ssize_t vfs_copy_file_range(struct file *file_in, loff_t pos_in,
 	    (file_out->f_flags & O_APPEND))
 		return -EBADF;
 
-	/* this could be relaxed once a method supports cross-fs copies */
-	if (inode_in->i_sb != inode_out->i_sb)
-		return -EXDEV;
-
 	file_start_write(file_out);
 
 	ret = do_copy_file_range(file_in, pos_in,
-- 
2.16.3

[PATCH 3/4] copy_file_range: splice with holes

[Goldwyn Rodrigues]

From: Goldwyn Rodrigues <rgoldwyn@suse.com>

copy_file_range calls do_splice_direct() if fs->clone_file_range
or fs->copy_file_range() is not available. However, do_splice_direct()
converts holes to zeros. Detect holes in the file_in range, and
create them in the corresponding file_out range.

If there is already data present at the offset in file_out, attempt
to punch a hole there. If the operation is not supported, fall
back to performing splice on the whole range.

Signed-off-by: Goldwyn Rodrigues <rgoldwyn@suse.com>
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
---
 fs/read_write.c | 62 +++++++++++++++++++++++++++++++++++++++++++++++++++++----
 1 file changed, 58 insertions(+), 4 deletions(-)

diff --git a/fs/read_write.c b/fs/read_write.c
index 1b8fc9eada69..e765fec656af 100644
--- a/fs/read_write.c
+++ b/fs/read_write.c
@@ -20,6 +20,7 @@
 #include <linux/compat.h>
 #include <linux/mount.h>
 #include <linux/fs.h>
+#include <linux/falloc.h>
 #include "internal.h"
 
 #include <linux/uaccess.h>
@@ -1547,7 +1548,8 @@ static ssize_t do_copy_file_range(struct file *file_in, loff_t pos_in,
 {
 	struct inode *inode_in = file_inode(file_in);
 	struct inode *inode_out = file_inode(file_out);
-	ssize_t ret = 0;
+	ssize_t ret = 0, total = 0;
+	loff_t size, end;
 
 	if (len == 0)
 		return 0;
@@ -1572,10 +1574,62 @@ static ssize_t do_copy_file_range(struct file *file_in, loff_t pos_in,
 		if (ret != -EOPNOTSUPP)
 			return ret;
 	}
+
 splice:
-	ret = do_splice_direct(file_in, &pos_in, file_out, &pos_out,
-			len > MAX_RW_COUNT ? MAX_RW_COUNT : len, 0);
-	return ret;
+	while (total < len) {
+		end = vfs_llseek(file_in, pos_in, SEEK_HOLE);
+
+		/* Starting position is already in a hole */
+		if (end == pos_in)
+			goto hole;
+		size = end - pos_in;
+do_splice:
+		if (size > len - total)
+			size = len - total;
+		ret = do_splice_direct(file_in, &pos_in, file_out, &pos_out,
+				size, 0);
+		if (ret < 0)
+			goto out;
+		total += ret;
+		if (total == len)
+			break;
+hole:
+		end = vfs_llseek(file_in, pos_in, SEEK_DATA);
+		if (end < 0) {
+			ret = end;
+			goto out;
+		}
+		size = end - pos_in;
+		if (size > len - total)
+			size = len - total;
+		/* Data on offset, punch holes */
+		if (i_size_read(file_out->f_inode) > pos_out) {
+			ret = vfs_fallocate(file_out,
+					FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE,
+					pos_out, size);
+			if (ret < 0) {
+				/*
+				 * The filesystem does not support punching
+				 * holes. Perform splice on the remaining range.
+				 */
+				if (ret == -EOPNOTSUPP) {
+					size = len - total;
+					goto do_splice;
+				}
+				goto out;
+			}
+		}
+		if (ret < 0) {
+			ret = end;
+			goto out;
+		}
+		pos_out += size;
+		pos_in = end;
+		total += size;
+	}
+
+out:
+	return total ? total : ret;
 }
 
 /*
-- 
2.16.3

[PATCH 4/4] ovl: Use do_copy_file_range() in copy_up_data()

[Goldwyn Rodrigues]

From: Goldwyn Rodrigues <rgoldwyn@suse.com>

This will preserve the holes by copying the chunks of data.
If available it will use clone().

Signed-off-by: Goldwyn Rodrigues <rgoldwyn@suse.com>
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
---
 fs/overlayfs/copy_up.c | 28 ++++++++--------------------
 fs/read_write.c        | 10 ++++++----
 include/linux/fs.h     |  3 +++
 3 files changed, 17 insertions(+), 24 deletions(-)

diff --git a/fs/overlayfs/copy_up.c b/fs/overlayfs/copy_up.c
index 8bede0742619..1f89380873ce 100644
--- a/fs/overlayfs/copy_up.c
+++ b/fs/overlayfs/copy_up.c
@@ -138,8 +138,7 @@ static int ovl_copy_up_data(struct path *old, struct path *new, loff_t len)
 {
 	struct file *old_file;
 	struct file *new_file;
-	loff_t old_pos = 0;
-	loff_t new_pos = 0;
+	loff_t pos = 0;
 	int error = 0;
 
 	if (len == 0)
@@ -155,38 +154,27 @@ static int ovl_copy_up_data(struct path *old, struct path *new, loff_t len)
 		goto out_fput;
 	}
 
-	/* Try to use clone_file_range to clone up within the same fs */
-	error = vfs_clone_file_range(old_file, 0, new_file, 0, len);
-	if (!error)
-		goto out;
-	/* Couldn't clone, so now we try to copy the data */
-	error = 0;
-
-	/* FIXME: copy up sparse files efficiently */
-	while (len) {
+	while (pos < len) {
 		size_t this_len = OVL_COPY_UP_CHUNK_SIZE;
 		long bytes;
 
-		if (len < this_len)
-			this_len = len;
+		if (len - pos < this_len)
+			this_len = len - pos;
 
 		if (signal_pending_state(TASK_KILLABLE, current)) {
 			error = -EINTR;
 			break;
 		}
 
-		bytes = do_splice_direct(old_file, &old_pos,
-					 new_file, &new_pos,
-					 this_len, SPLICE_F_MOVE);
+		bytes = do_copy_file_range(old_file, pos,
+					 new_file, pos,
+					 this_len, 0, SPLICE_F_MOVE);
 		if (bytes <= 0) {
 			error = bytes;
 			break;
 		}
-		WARN_ON(old_pos != new_pos);
-
-		len -= bytes;
+		pos += bytes;
 	}
-out:
 	if (!error)
 		error = vfs_fsync(new_file, 0);
 	fput(new_file);
diff --git a/fs/read_write.c b/fs/read_write.c
index e765fec656af..50d7ef77410f 100644
--- a/fs/read_write.c
+++ b/fs/read_write.c
@@ -1542,9 +1542,10 @@ COMPAT_SYSCALL_DEFINE4(sendfile64, int, out_fd, int, in_fd,
 }
 #endif
 
-static ssize_t do_copy_file_range(struct file *file_in, loff_t pos_in,
+ssize_t do_copy_file_range(struct file *file_in, loff_t pos_in,
 			    struct file *file_out, loff_t pos_out,
-			    size_t len, unsigned int flags)
+			    size_t len, unsigned int flags,
+			    unsigned int splice_flags)
 {
 	struct inode *inode_in = file_inode(file_in);
 	struct inode *inode_out = file_inode(file_out);
@@ -1587,7 +1588,7 @@ static ssize_t do_copy_file_range(struct file *file_in, loff_t pos_in,
 		if (size > len - total)
 			size = len - total;
 		ret = do_splice_direct(file_in, &pos_in, file_out, &pos_out,
-				size, 0);
+				size, splice_flags);
 		if (ret < 0)
 			goto out;
 		total += ret;
@@ -1631,6 +1632,7 @@ static ssize_t do_copy_file_range(struct file *file_in, loff_t pos_in,
 out:
 	return total ? total : ret;
 }
+EXPORT_SYMBOL(do_copy_file_range);
 
 /*
  * copy_file_range() differs from regular file read and write in that it
@@ -1669,7 +1671,7 @@ ssize_t vfs_copy_file_range(struct file *file_in, loff_t pos_in,
 	file_start_write(file_out);
 
 	ret = do_copy_file_range(file_in, pos_in,
-			file_out, pos_out, len, flags);
+			file_out, pos_out, len, flags, 0);
 
 	if (ret > 0) {
 		fsnotify_access(file_in);
diff --git a/include/linux/fs.h b/include/linux/fs.h
index 760d8da1b6c7..d5349b17fa10 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -1799,6 +1799,9 @@ extern ssize_t vfs_read(struct file *, char __user *, size_t, loff_t *);
 extern ssize_t vfs_write(struct file *, const char __user *, size_t, loff_t *);
 extern ssize_t vfs_readv(struct file *, const struct iovec __user *,
 		unsigned long, loff_t *, rwf_t);
+extern ssize_t do_copy_file_range(struct file *, loff_t , struct file *,
+				   loff_t , size_t, unsigned int,
+				   unsigned int);
 extern ssize_t vfs_copy_file_range(struct file *, loff_t , struct file *,
 				   loff_t, size_t, unsigned int);
 extern int vfs_clone_file_prep_inodes(struct inode *inode_in, loff_t pos_in,
-- 
2.16.3

Re: [PATCH 3/4] copy_file_range: splice with holes

[Amir Goldstein]

On Thu, May 10, 2018 at 4:58 AM, Goldwyn Rodrigues <rgoldwyn@suse.de> wrote:
> From: Goldwyn Rodrigues <rgoldwyn@suse.com>
>
> copy_file_range calls do_splice_direct() if fs->clone_file_range
> or fs->copy_file_range() is not available. However, do_splice_direct()
> converts holes to zeros. Detect holes in the file_in range, and
> create them in the corresponding file_out range.
>
> If there is already data present at the offset in file_out, attempt
> to punch a hole there. If the operation is not supported, fall
> back to performing splice on the whole range.
>
> Signed-off-by: Goldwyn Rodrigues <rgoldwyn@suse.com>
> Reviewed-by: Amir Goldstein <amir73il@gmail.com>
> ---
>  fs/read_write.c | 62 +++++++++++++++++++++++++++++++++++++++++++++++++++++----
>  1 file changed, 58 insertions(+), 4 deletions(-)
>
> diff --git a/fs/read_write.c b/fs/read_write.c
> index 1b8fc9eada69..e765fec656af 100644
> --- a/fs/read_write.c
> +++ b/fs/read_write.c
> @@ -20,6 +20,7 @@
>  #include <linux/compat.h>
>  #include <linux/mount.h>
>  #include <linux/fs.h>
> +#include <linux/falloc.h>
>  #include "internal.h"
>
>  #include <linux/uaccess.h>
> @@ -1547,7 +1548,8 @@ static ssize_t do_copy_file_range(struct file *file_in, loff_t pos_in,
>  {
>         struct inode *inode_in = file_inode(file_in);
>         struct inode *inode_out = file_inode(file_out);
> -       ssize_t ret = 0;
> +       ssize_t ret = 0, total = 0;
> +       loff_t size, end;
>
>         if (len == 0)
>                 return 0;
> @@ -1572,10 +1574,62 @@ static ssize_t do_copy_file_range(struct file *file_in, loff_t pos_in,
>                 if (ret != -EOPNOTSUPP)
>                         return ret;
>         }
> +
>  splice:
> -       ret = do_splice_direct(file_in, &pos_in, file_out, &pos_out,
> -                       len > MAX_RW_COUNT ? MAX_RW_COUNT : len, 0);
> -       return ret;
> +       while (total < len) {
> +               end = vfs_llseek(file_in, pos_in, SEEK_HOLE);
> +
> +               /* Starting position is already in a hole */
> +               if (end == pos_in)
> +                       goto hole;
> +               size = end - pos_in;
> +do_splice:
> +               if (size > len - total)
> +                       size = len - total;
> +               ret = do_splice_direct(file_in, &pos_in, file_out, &pos_out,
> +                               size, 0);

I wonder, can do_splice_direct() return short copy (< size)?
If so, code below will try to punch a zero length hole.
Best put some protection here, don't you think?

> +               if (ret < 0)
> +                       goto out;
> +               total += ret;
> +               if (total == len)
> +                       break;
> +hole:
> +               end = vfs_llseek(file_in, pos_in, SEEK_DATA);
> +               if (end < 0) {
> +                       ret = end;
> +                       goto out;
> +               }
> +               size = end - pos_in;
> +               if (size > len - total)
> +                       size = len - total;
> +               /* Data on offset, punch holes */
> +               if (i_size_read(file_out->f_inode) > pos_out) {
> +                       ret = vfs_fallocate(file_out,
> +                                       FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE,
> +                                       pos_out, size);

I'm afraid you have more re-factoring to do vfs_fallocate() does
file_start_write() -
you probably need do_fallocate().


> +                       if (ret < 0) {
> +                               /*
> +                                * The filesystem does not support punching
> +                                * holes. Perform splice on the remaining range.
> +                                */
> +                               if (ret == -EOPNOTSUPP) {
> +                                       size = len - total;
> +                                       goto do_splice;
> +                               }
> +                               goto out;
> +                       }
> +               }
> +               if (ret < 0) {
> +                       ret = end;
> +                       goto out;
> +               }
> +               pos_out += size;
> +               pos_in = end;
> +               total += size;
> +       }
> +
> +out:
> +       return total ? total : ret;
>  }
>
>  /*
> --
> 2.16.3
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-unionfs" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH v2 0/4] Enable holes in copy_file_range()

[Amir Goldstein]

On Thu, May 10, 2018 at 4:58 AM, Goldwyn Rodrigues <rgoldwyn@suse.de> wrote:
> If copy_file_range performs a copy using splice, it converts holes
> to zeros. This effort primarily changes this behavior to create
> holes when it is possible.
>
> Even if copy_file_range() or clone_file_range() does not work for different
> mounted filesystems, We should be able to splice files if they do not
> belong the same super_block.
>
> Changes since v1:
> - Fixed bug when hole/data offset is farther than len

Please specify how you tested this work.
Did you run the xfstests generic/copy tests that Anna wrote?
If the tests caught your bugs - good, if they didn't, please
write more tests to cover the issues you found.
One thing I do not see in the existing tests is coverage for the
trivial case of "only holes" (i.e. touch x; truncate -s 10 x).

Thanks,
Amir.

Re: [PATCH 3/4] copy_file_range: splice with holes

[Amir Goldstein]

On Thu, May 10, 2018 at 7:42 AM, Amir Goldstein <amir73il@gmail.com> wrote:
> On Thu, May 10, 2018 at 4:58 AM, Goldwyn Rodrigues <rgoldwyn@suse.de> wrote:
>> From: Goldwyn Rodrigues <rgoldwyn@suse.com>
>>
>> copy_file_range calls do_splice_direct() if fs->clone_file_range
>> or fs->copy_file_range() is not available. However, do_splice_direct()
>> converts holes to zeros. Detect holes in the file_in range, and
>> create them in the corresponding file_out range.
>>
>> If there is already data present at the offset in file_out, attempt
>> to punch a hole there. If the operation is not supported, fall
>> back to performing splice on the whole range.
>>
>> Signed-off-by: Goldwyn Rodrigues <rgoldwyn@suse.com>
>> Reviewed-by: Amir Goldstein <amir73il@gmail.com>
>> ---
>>  fs/read_write.c | 62 +++++++++++++++++++++++++++++++++++++++++++++++++++++----
>>  1 file changed, 58 insertions(+), 4 deletions(-)
>>
>> diff --git a/fs/read_write.c b/fs/read_write.c
>> index 1b8fc9eada69..e765fec656af 100644
>> --- a/fs/read_write.c
>> +++ b/fs/read_write.c
>> @@ -20,6 +20,7 @@
>>  #include <linux/compat.h>
>>  #include <linux/mount.h>
>>  #include <linux/fs.h>
>> +#include <linux/falloc.h>
>>  #include "internal.h"
>>
>>  #include <linux/uaccess.h>
>> @@ -1547,7 +1548,8 @@ static ssize_t do_copy_file_range(struct file *file_in, loff_t pos_in,
>>  {
>>         struct inode *inode_in = file_inode(file_in);
>>         struct inode *inode_out = file_inode(file_out);
>> -       ssize_t ret = 0;
>> +       ssize_t ret = 0, total = 0;
>> +       loff_t size, end;
>>
>>         if (len == 0)
>>                 return 0;
>> @@ -1572,10 +1574,62 @@ static ssize_t do_copy_file_range(struct file *file_in, loff_t pos_in,
>>                 if (ret != -EOPNOTSUPP)
>>                         return ret;
>>         }
>> +
>>  splice:
>> -       ret = do_splice_direct(file_in, &pos_in, file_out, &pos_out,
>> -                       len > MAX_RW_COUNT ? MAX_RW_COUNT : len, 0);
>> -       return ret;
>> +       while (total < len) {
>> +               end = vfs_llseek(file_in, pos_in, SEEK_HOLE);
>> +
>> +               /* Starting position is already in a hole */
>> +               if (end == pos_in)
>> +                       goto hole;
>> +               size = end - pos_in;
>> +do_splice:
>> +               if (size > len - total)
>> +                       size = len - total;
>> +               ret = do_splice_direct(file_in, &pos_in, file_out, &pos_out,
>> +                               size, 0);
>
> I wonder, can do_splice_direct() return short copy (< size)?
> If so, code below will try to punch a zero length hole.
> Best put some protection here, don't you think?
>
>> +               if (ret < 0)
>> +                       goto out;
>> +               total += ret;
>> +               if (total == len)
>> +                       break;
>> +hole:
>> +               end = vfs_llseek(file_in, pos_in, SEEK_DATA);
>> +               if (end < 0) {
>> +                       ret = end;
>> +                       goto out;
>> +               }
>> +               size = end - pos_in;
>> +               if (size > len - total)
>> +                       size = len - total;
>> +               /* Data on offset, punch holes */
>> +               if (i_size_read(file_out->f_inode) > pos_out) {
>> +                       ret = vfs_fallocate(file_out,
>> +                                       FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE,
>> +                                       pos_out, size);
>
> I'm afraid you have more re-factoring to do vfs_fallocate() does
> file_start_write() -
> you probably need do_fallocate().
>
>

I was trying to look for a pattern of what goes in vfs_ helpers and their
corresponding do_ helpers and I can't say I found a single pattern.

What stood out for me is the do_clone_file_range() is a wrapper
around vfs_clone_file_range() while do_truncate() is a helper
of vfs_truncate(). I did not survey all of those helpers, but I have
a feeling that the latter is the more common pattern and I know
who to blame for the former...

Anyway, this anomaly, explains why overlayfs calls
vfs_clone_file_range() and it cannot call vfs_fallocate()
from the copy up loop context.

I advise you to turn on LOCKDEP while testing to be warned
about this sort of things.

Thanks,
Amir.

Re: [PATCH v2 0/4] Enable holes in copy_file_range()

[Anna Schumaker]

On 05/10/2018 12:47 AM, Amir Goldstein wrote:
> On Thu, May 10, 2018 at 4:58 AM, Goldwyn Rodrigues <rgoldwyn@suse.de> wrote:
>> If copy_file_range performs a copy using splice, it converts holes
>> to zeros. This effort primarily changes this behavior to create
>> holes when it is possible.
>>
>> Even if copy_file_range() or clone_file_range() does not work for different
>> mounted filesystems, We should be able to splice files if they do not
>> belong the same super_block.
>>
>> Changes since v1:
>> - Fixed bug when hole/data offset is farther than len
> 
> Please specify how you tested this work.
> Did you run the xfstests generic/copy tests that Anna wrote?
> If the tests caught your bugs - good, if they didn't, please
> write more tests to cover the issues you found.
> One thing I do not see in the existing tests is coverage for the
> trivial case of "only holes" (i.e. touch x; truncate -s 10 x).

Yeah,  that's probably an oversight on my part.  I hadn't even thought about sparse files as I was writing those tests.  Please add tests for holes!

Anna

> 
> Thanks,
> Amir.
> 

[PATCH 2/4] copy_file_range: Perform splice if in/out SB are not same

[Goldwyn Rodrigues]

From: Goldwyn Rodrigues <rgoldwyn@suse.com>

While performing copy_file_range(), if superblocks of file_in and
file_out don't match, instead of returning -EXDEV, perform
splice for a faster copy.

Signed-off-by: Goldwyn Rodrigues <rgoldwyn@suse.com>
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
---
 fs/read_write.c | 11 ++++++-----
 1 file changed, 6 insertions(+), 5 deletions(-)

diff --git a/fs/read_write.c b/fs/read_write.c
index 525f2a67e15a..1b8fc9eada69 100644
--- a/fs/read_write.c
+++ b/fs/read_write.c
@@ -1545,11 +1545,16 @@ static ssize_t do_copy_file_range(struct file *file_in, loff_t pos_in,
 			    struct file *file_out, loff_t pos_out,
 			    size_t len, unsigned int flags)
 {
+	struct inode *inode_in = file_inode(file_in);
+	struct inode *inode_out = file_inode(file_out);
 	ssize_t ret = 0;
 
 	if (len == 0)
 		return 0;
 
+	if (inode_in->i_sb != inode_out->i_sb)
+		goto splice;
+
 	/*
 	 * Try cloning first, this is supported by more file systems, and
 	 * more efficient if both clone and copy are supported (e.g. NFS).
@@ -1567,7 +1572,7 @@ static ssize_t do_copy_file_range(struct file *file_in, loff_t pos_in,
 		if (ret != -EOPNOTSUPP)
 			return ret;
 	}
-
+splice:
 	ret = do_splice_direct(file_in, &pos_in, file_out, &pos_out,
 			len > MAX_RW_COUNT ? MAX_RW_COUNT : len, 0);
 	return ret;
@@ -1607,10 +1612,6 @@ ssize_t vfs_copy_file_range(struct file *file_in, loff_t pos_in,
 	    (file_out->f_flags & O_APPEND))
 		return -EBADF;
 
-	/* this could be relaxed once a method supports cross-fs copies */
-	if (inode_in->i_sb != inode_out->i_sb)
-		return -EXDEV;
-
 	file_start_write(file_out);
 
 	ret = do_copy_file_range(file_in, pos_in,
-- 
2.16.3

[PATCH 2/4] copy_file_range: Perform splice if in/out SB are not same

[Goldwyn Rodrigues]

From: Goldwyn Rodrigues <rgoldwyn@suse.com>

While performing copy_file_range(), if superblocks of file_in and
file_out don't match, instead of returning -EXDEV, perform
splice for a faster copy.

Signed-off-by: Goldwyn Rodrigues <rgoldwyn@suse.com>
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
---
 fs/read_write.c | 11 ++++++-----
 1 file changed, 6 insertions(+), 5 deletions(-)

diff --git a/fs/read_write.c b/fs/read_write.c
index 525f2a67e15a..1b8fc9eada69 100644
--- a/fs/read_write.c
+++ b/fs/read_write.c
@@ -1545,11 +1545,16 @@ static ssize_t do_copy_file_range(struct file *file_in, loff_t pos_in,
 			    struct file *file_out, loff_t pos_out,
 			    size_t len, unsigned int flags)
 {
+	struct inode *inode_in = file_inode(file_in);
+	struct inode *inode_out = file_inode(file_out);
 	ssize_t ret = 0;
 
 	if (len == 0)
 		return 0;
 
+	if (inode_in->i_sb != inode_out->i_sb)
+		goto splice;
+
 	/*
 	 * Try cloning first, this is supported by more file systems, and
 	 * more efficient if both clone and copy are supported (e.g. NFS).
@@ -1567,7 +1572,7 @@ static ssize_t do_copy_file_range(struct file *file_in, loff_t pos_in,
 		if (ret != -EOPNOTSUPP)
 			return ret;
 	}
-
+splice:
 	ret = do_splice_direct(file_in, &pos_in, file_out, &pos_out,
 			len > MAX_RW_COUNT ? MAX_RW_COUNT : len, 0);
 	return ret;
@@ -1607,10 +1612,6 @@ ssize_t vfs_copy_file_range(struct file *file_in, loff_t pos_in,
 	    (file_out->f_flags & O_APPEND))
 		return -EBADF;
 
-	/* this could be relaxed once a method supports cross-fs copies */
-	if (inode_in->i_sb != inode_out->i_sb)
-		return -EXDEV;
-
 	file_start_write(file_out);
 
 	ret = do_copy_file_range(file_in, pos_in,
-- 
2.16.3

Re: [PATCH 2/4] copy_file_range: Perform splice if in/out SB are not same

[Goldwyn Rodrigues]

On 05/08/2018 04:57 PM, Florian Weimer wrote:
> On 05/08/2018 11:24 PM, Goldwyn Rodrigues wrote:
>> While performing copy_file_range(), if superblocks of file_in and
>> file_out don't match, instead of returning -EXDEV, perform
>> splice for a faster copy.
> 
> We have a userspace emulation in glibc which used to be quite faithful,
> including the EXDEV error (which is not strictly necessary to produce).
> 
> Should we change glibc to perform a userspace copy if the system call
> returns EXDEV due to an older kernel?
> 

I don't seen any purpose. The user would anyways have to perform a copy
if it receives -EXDEV.

-- 
Goldwyn

Re: [PATCH 2/4] copy_file_range: Perform splice if in/out SB are not same

[Amir Goldstein]

On Wed, May 9, 2018 at 12:24 AM, Goldwyn Rodrigues <rgoldwyn@suse.de> wrote:
> From: Goldwyn Rodrigues <rgoldwyn@suse.com>
>
> While performing copy_file_range(), if superblocks of file_in and
> file_out don't match, instead of returning -EXDEV, perform
> splice for a faster copy.
>
> Signed-off-by: Goldwyn Rodrigues <rgoldwyn@suse.com>

Reviewed-by: Amir Goldstein <amir73il@gmail.com>

> ---
>  fs/read_write.c | 11 ++++++-----
>  1 file changed, 6 insertions(+), 5 deletions(-)
>
> diff --git a/fs/read_write.c b/fs/read_write.c
> index 7d9dfb62ba7d..2c9e7a5ea806 100644
> --- a/fs/read_write.c
> +++ b/fs/read_write.c
> @@ -1546,6 +1546,8 @@ ssize_t do_copy_file_range(struct file *file_in, loff_t pos_in,
>                             size_t len, unsigned int flags,
>                             unsigned int splice_flags)
>  {
> +       struct inode *inode_in = file_inode(file_in);
> +       struct inode *inode_out = file_inode(file_out);
>         ssize_t ret = 0;
>
>         if (flags != 0)
> @@ -1554,6 +1556,9 @@ ssize_t do_copy_file_range(struct file *file_in, loff_t pos_in,
>         if (len == 0)
>                 return 0;
>
> +       if (inode_in->i_sb != inode_out->i_sb)
> +               goto splice;
> +
>         /*
>          * Try cloning first, this is supported by more file systems, and
>          * more efficient if both clone and copy are supported (e.g. NFS).
> @@ -1571,7 +1576,7 @@ ssize_t do_copy_file_range(struct file *file_in, loff_t pos_in,
>                 if (ret != -EOPNOTSUPP)
>                         return ret;
>         }
> -
> +splice:
>         ret = do_splice_direct(file_in, &pos_in, file_out, &pos_out,
>                         len > MAX_RW_COUNT ? MAX_RW_COUNT : len, splice_flags);
>         return ret;
> @@ -1608,10 +1613,6 @@ ssize_t vfs_copy_file_range(struct file *file_in, loff_t pos_in,
>             (file_out->f_flags & O_APPEND))
>                 return -EBADF;
>
> -       /* this could be relaxed once a method supports cross-fs copies */
> -       if (inode_in->i_sb != inode_out->i_sb)
> -               return -EXDEV;
> -
>         file_start_write(file_out);
>
>         ret = do_copy_file_range(file_in, pos_in,
> --
> 2.16.3
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-unionfs" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH 2/4] copy_file_range: Perform splice if in/out SB are not same

[Florian Weimer]

On 05/08/2018 11:24 PM, Goldwyn Rodrigues wrote:
> While performing copy_file_range(), if superblocks of file_in and
> file_out don't match, instead of returning -EXDEV, perform
> splice for a faster copy.

We have a userspace emulation in glibc which used to be quite faithful, 
including the EXDEV error (which is not strictly necessary to produce).

Should we change glibc to perform a userspace copy if the system call 
returns EXDEV due to an older kernel?

Thanks,
Florian

[PATCH 2/4] copy_file_range: Perform splice if in/out SB are not same

[Goldwyn Rodrigues]

From: Goldwyn Rodrigues <rgoldwyn@suse.com>

While performing copy_file_range(), if superblocks of file_in and
file_out don't match, instead of returning -EXDEV, perform
splice for a faster copy.

Signed-off-by: Goldwyn Rodrigues <rgoldwyn@suse.com>
---
 fs/read_write.c | 11 ++++++-----
 1 file changed, 6 insertions(+), 5 deletions(-)

diff --git a/fs/read_write.c b/fs/read_write.c
index 7d9dfb62ba7d..2c9e7a5ea806 100644
--- a/fs/read_write.c
+++ b/fs/read_write.c
@@ -1546,6 +1546,8 @@ ssize_t do_copy_file_range(struct file *file_in, loff_t pos_in,
 			    size_t len, unsigned int flags,
 			    unsigned int splice_flags)
 {
+	struct inode *inode_in = file_inode(file_in);
+	struct inode *inode_out = file_inode(file_out);
 	ssize_t ret = 0;
 
 	if (flags != 0)
@@ -1554,6 +1556,9 @@ ssize_t do_copy_file_range(struct file *file_in, loff_t pos_in,
 	if (len == 0)
 		return 0;
 
+	if (inode_in->i_sb != inode_out->i_sb)
+		goto splice;
+
 	/*
 	 * Try cloning first, this is supported by more file systems, and
 	 * more efficient if both clone and copy are supported (e.g. NFS).
@@ -1571,7 +1576,7 @@ ssize_t do_copy_file_range(struct file *file_in, loff_t pos_in,
 		if (ret != -EOPNOTSUPP)
 			return ret;
 	}
-
+splice:
 	ret = do_splice_direct(file_in, &pos_in, file_out, &pos_out,
 			len > MAX_RW_COUNT ? MAX_RW_COUNT : len, splice_flags);
 	return ret;
@@ -1608,10 +1613,6 @@ ssize_t vfs_copy_file_range(struct file *file_in, loff_t pos_in,
 	    (file_out->f_flags & O_APPEND))
 		return -EBADF;
 
-	/* this could be relaxed once a method supports cross-fs copies */
-	if (inode_in->i_sb != inode_out->i_sb)
-		return -EXDEV;
-
 	file_start_write(file_out);
 
 	ret = do_copy_file_range(file_in, pos_in,
-- 
2.16.3