diff mbox series

[v2,3/4] ext4: Add iomap support for inline data

Message ID 20170914095047.23935-4-agruenba@redhat.com
State Superseded
Headers show
Series ext4: SEEK_HOLE / SEEK_DATA via iomap | expand

Commit Message

Andreas Gruenbacher Sept. 14, 2017, 9:50 a.m. UTC 
Report inline data as a IOMAP_F_DATA_INLINE mapping.  This allows to use
iomap_seek_hole and iomap_seek_data in ext4_llseek and makes switching
to iomap_fiemap in ext4_fiemap easier.

Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Reviewed-by: Jan Kara <jack@suse.cz>
---
 fs/ext4/ext4.h   |  4 ++++
 fs/ext4/inline.c | 33 +++++++++++++++++++++++++++++++++
 fs/ext4/inode.c  | 16 ++++++++++++++--
 3 files changed, 51 insertions(+), 2 deletions(-)

Comments

Theodore Ts'o Sept. 28, 2017, 5:57 p.m. UTC | #1 
On Thu, Sep 14, 2017 at 11:50:46AM +0200, Andreas Gruenbacher wrote:
> +int ext4_inline_data_iomap(struct inode *inode, struct iomap *iomap)
> +{
> +	__u64 addr;
> +	int error = -EAGAIN;
> +	struct ext4_iloc iloc;
> +
> +	down_read(&EXT4_I(inode)->xattr_sem);
> +	if (!ext4_has_inline_data(inode))
> +		goto out;
	....
> +out:
> +	up_read(&EXT4_I(inode)->xattr_sem);
> +	return error;
> +}


If we race with the inline data flag getting cleared,
ext4_iomap_begin() will return with -EAGAIN.  As near as I can tell,
this will get reflected all the way up to userspace, instead of having
the retry happen in the kernel.  Is this intentional?

It looks like a user visible change, no?

						- Ted
Andreas Gruenbacher Oct. 2, 2017, 2:39 p.m. UTC | #2 
On Thu, Sep 28, 2017 at 7:57 PM, Theodore Ts'o <tytso@mit.edu> wrote:
> On Thu, Sep 14, 2017 at 11:50:46AM +0200, Andreas Gruenbacher wrote:
>> +int ext4_inline_data_iomap(struct inode *inode, struct iomap *iomap)
>> +{
>> +     __u64 addr;
>> +     int error = -EAGAIN;
>> +     struct ext4_iloc iloc;
>> +
>> +     down_read(&EXT4_I(inode)->xattr_sem);
>> +     if (!ext4_has_inline_data(inode))
>> +             goto out;
>         ....
>> +out:
>> +     up_read(&EXT4_I(inode)->xattr_sem);
>> +     return error;
>> +}
>
>
> If we race with the inline data flag getting cleared,
> ext4_iomap_begin() will return with -EAGAIN.

Actually, in that case, ext4_iomap_begin will treat the file as non-inline.
It will not return -EAGAIN.

> As near as I can tell,
> this will get reflected all the way up to userspace, instead of having
> the retry happen in the kernel.  Is this intentional?
>
> It looks like a user visible change, no?

That would be a bug.

Thanks,
Andreas
Theodore Ts'o Oct. 2, 2017, 3:02 p.m. UTC | #3 
On Mon, Oct 02, 2017 at 04:39:59PM +0200, Andreas Gruenbacher wrote:
> > If we race with the inline data flag getting cleared,
> > ext4_iomap_begin() will return with -EAGAIN.
> 
> Actually, in that case, ext4_iomap_begin will treat the file as non-inline.
> It will not return -EAGAIN.

Ah, thanks, I missed that.

					- Ted
diff mbox series

Patch

diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h
index e2abe01c8c6b..ae3e4a25821a 100644
--- a/fs/ext4/ext4.h
+++ b/fs/ext4/ext4.h
@@ -3048,6 +3048,10 @@  extern struct buffer_head *ext4_get_first_inline_block(struct inode *inode,
 extern int ext4_inline_data_fiemap(struct inode *inode,
 				   struct fiemap_extent_info *fieinfo,
 				   int *has_inline, __u64 start, __u64 len);
+
+struct iomap;
+extern int ext4_inline_data_iomap(struct inode *inode, struct iomap *iomap);
+
 extern int ext4_try_to_evict_inline_data(handle_t *handle,
 					 struct inode *inode,
 					 int needed);
diff --git a/fs/ext4/inline.c b/fs/ext4/inline.c
index 28c5c3abddb3..f0bbc8cb6555 100644
--- a/fs/ext4/inline.c
+++ b/fs/ext4/inline.c
@@ -12,6 +12,7 @@ 
  * GNU General Public License for more details.
  */
 
+#include <linux/iomap.h>
 #include <linux/fiemap.h>
 
 #include "ext4_jbd2.h"
@@ -1827,6 +1828,38 @@  int ext4_destroy_inline_data(handle_t *handle, struct inode *inode)
 	return ret;
 }
 
+int ext4_inline_data_iomap(struct inode *inode, struct iomap *iomap)
+{
+	__u64 addr;
+	int error = -EAGAIN;
+	struct ext4_iloc iloc;
+
+	down_read(&EXT4_I(inode)->xattr_sem);
+	if (!ext4_has_inline_data(inode))
+		goto out;
+
+	error = ext4_get_inode_loc(inode, &iloc);
+	if (error)
+		goto out;
+
+	addr = (__u64)iloc.bh->b_blocknr << inode->i_sb->s_blocksize_bits;
+	addr += (char *)ext4_raw_inode(&iloc) - iloc.bh->b_data;
+	addr += offsetof(struct ext4_inode, i_block);
+
+	brelse(iloc.bh);
+
+	iomap->addr = addr;
+	iomap->offset = 0;
+	iomap->length = min_t(loff_t, ext4_get_inline_size(inode),
+			      i_size_read(inode));
+	iomap->type = 0;
+	iomap->flags = IOMAP_F_DATA_INLINE;
+
+out:
+	up_read(&EXT4_I(inode)->xattr_sem);
+	return error;
+}
+
 int ext4_inline_data_fiemap(struct inode *inode,
 			    struct fiemap_extent_info *fieinfo,
 			    int *has_inline, __u64 start, __u64 len)
diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
index d9e633c12aae..7755f41bdfc3 100644
--- a/fs/ext4/inode.c
+++ b/fs/ext4/inode.c
@@ -3404,8 +3404,20 @@  static int ext4_iomap_begin(struct inode *inode, loff_t offset, loff_t length,
 	struct ext4_map_blocks map;
 	int ret;
 
-	if (WARN_ON_ONCE(ext4_has_inline_data(inode)))
-		return -ERANGE;
+
+	if (flags & IOMAP_REPORT) {
+		if (ext4_has_inline_data(inode)) {
+			ret = ext4_inline_data_iomap(inode, iomap);
+			if (ret != -EAGAIN) {
+				if (ret == 0 && offset >= iomap->length)
+					ret = -ENOENT;
+				return ret;
+			}
+		}
+	} else {
+		if (WARN_ON_ONCE(ext4_has_inline_data(inode)))
+			return -ERANGE;
+	}
 
 	map.m_lblk = first_block;
 	map.m_len = last_block - first_block + 1;