ocfs2: fall back to buffer IO when append dio is disabled with file hole existing

The intention of this patch is to provide an option to ocfs2 users whether to allocate disk space while doing dio write.

Whether to make ocfs2 fall back to buffer io is up to ocfs2 users through toggling append-dio feature. It rather makes ocfs2 configurable and flexible.

So if something bad happens to dio write with space allocation, we can still make ocfs2 fall back to buffer io. It's an option not a mandatory action.:)

Besides, append-dio feature is key to whether to allocate space with dio writing. So writing to file hole and enlarging file(appending file) should have the same reflection to append-dio feature.

Signed-off-by Changwei Ge <[email protected]>

---
  fs/ocfs2/aops.c | 53 ++++++++++++++++++++++++++++++++++++++++++++++++++---
  1 file changed, 50 insertions(+), 3 deletions(-)

fs/ocfs2/aops.c +50 -3

index d151632..32e60c0 100644
@@ -2414,12 +2414,52 @@ static int ocfs2_dio_end_io(struct kiocb *iocb,
  	return ret;
  }
 
+/*
+ * Will look for holes and unwritten extents in the range starting at
+ * pos for count bytes (inclusive).
+ * Return value 1 indicates hole exists, 0 not exists, others indicate error.
+ */
+static int ocfs2_range_has_holes(struct inode *inode, loff_t pos,
+				 size_t count)
+{
+	int ret = 0;
+	unsigned int extent_flags;
+	u32 cpos, clusters, extent_len, phys_cpos;
+	struct super_block *sb = inode->i_sb;
+
+	cpos = pos >> OCFS2_SB(sb)->s_clustersize_bits;
+	clusters = ocfs2_clusters_for_bytes(sb, pos + count) - cpos;
+
+	while (clusters) {
+		ret = ocfs2_get_clusters(inode, cpos, &phys_cpos, &extent_len,
+					 &extent_flags);
+		if (ret < 0) {
+			mlog_errno(ret);
+			goto out;
+		}
+
+		if (phys_cpos == 0) {
+			ret = 1;
+			goto out;
+		}
+
+		if (extent_len > clusters)
+			extent_len = clusters;
+
+		clusters -= extent_len;
+		cpos += extent_len;
+	}
+out:
+	return ret;
+}
+
  static ssize_t ocfs2_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
  {
  	struct file *file = iocb->ki_filp;
  	struct inode *inode = file->f_mapping->host;
  	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
  	get_block_t *get_block;
+	int ret;
 
  	/*
  	 * Fallback to buffered I/O if we see an inode without

@@ -2429,9 +2469,16 @@ static ssize_t ocfs2_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
  		return 0;
 
  	/* Fallback to buffered I/O if we do not support append dio. */
-	if (iocb->ki_pos + iter->count > i_size_read(inode) &&
-	    !ocfs2_supports_append_dio(osb))
-		return 0;
+	if (!ocfs2_supports_append_dio(osb)) {
+		if (iocb->ki_pos + iter->count > i_size_read(inode))
+			return 0;
+
+		ret = ocfs2_range_has_holes(inode, iocb->ki_pos, iter->count);
+		if (ret == 1)
+			return 0;
+		else if (ret < 0)
+			return ret;
+	}
 
  	if (iov_iter_rw(iter) == READ)
  		get_block = ocfs2_lock_get_block;

Hi Changwei,

Quoted 7 lines · Changwei Ge

The intention of this patch is to provide an option to ocfs2 users whether to allocate disk space while doing dio write.

Whether to make ocfs2 fall back to buffer io is up to ocfs2 users through toggling append-dio feature. It rather makes ocfs2 configurable and flexible.

It is too strange to make ocfs2 fall back to buffer io by toggling append-dio feature.

Quoted 3 lines

So if something bad happens to dio write with space allocation, we can still make ocfs2 fall back to buffer io. It's an option not a mandatory action.:)

Now the ocfs2 supports fill holes during direct io whether or not supporting append-dio feature and we can directly fix the problem.
I think it is meaningless to provide an temporary option to turn off it.

Quoted 78 lines

Besides, append-dio feature is key to whether to allocate space with dio writing. So writing to file hole and enlarging file(appending file) should have the same reflection to append-dio feature.

Signed-off-by Changwei Ge <[email protected]>

---
  fs/ocfs2/aops.c | 53 ++++++++++++++++++++++++++++++++++++++++++++++++++---
  1 file changed, 50 insertions(+), 3 deletions(-)

fs/ocfs2/aops.c +43 -3

index d151632..32e60c0 100644
@@ -2414,12 +2414,52 @@ static int ocfs2_dio_end_io(struct kiocb *iocb,
  	return ret;
  }
 
+/*
+ * Will look for holes and unwritten extents in the range starting at
+ * pos for count bytes (inclusive).
+ * Return value 1 indicates hole exists, 0 not exists, others indicate error.
+ */
+static int ocfs2_range_has_holes(struct inode *inode, loff_t pos,
+				 size_t count)
+{
+	int ret = 0;
+	unsigned int extent_flags;
+	u32 cpos, clusters, extent_len, phys_cpos;
+	struct super_block *sb = inode->i_sb;
+
+	cpos = pos >> OCFS2_SB(sb)->s_clustersize_bits;
+	clusters = ocfs2_clusters_for_bytes(sb, pos + count) - cpos;
+
+	while (clusters) {
+		ret = ocfs2_get_clusters(inode, cpos, &phys_cpos, &extent_len,
+					 &extent_flags);
+		if (ret < 0) {
+			mlog_errno(ret);
+			goto out;
+		}
+
+		if (phys_cpos == 0) {
+			ret = 1;
+			goto out;
+		}
+
+		if (extent_len > clusters)
+			extent_len = clusters;
+
+		clusters -= extent_len;
+		cpos += extent_len;
+	}
+out:
+	return ret;
+}
+
  static ssize_t ocfs2_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
  {
  	struct file *file = iocb->ki_filp;
  	struct inode *inode = file->f_mapping->host;
  	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
  	get_block_t *get_block;
+	int ret;
 
  	/*
  	 * Fallback to buffered I/O if we see an inode without

@@ -2429,9 +2469,16 @@ static ssize_t ocfs2_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
  		return 0;
 
  	/* Fallback to buffered I/O if we do not support append dio. */
-	if (iocb->ki_pos + iter->count > i_size_read(inode) &&
-	    !ocfs2_supports_append_dio(osb))
-		return 0;
+	if (!ocfs2_supports_append_dio(osb)) {
+		if (iocb->ki_pos + iter->count > i_size_read(inode))
+			return 0;

fs/ocfs2/aops.c static ssize_t ocfs2_direct_IO(struct kiocb *iocb, struct iov_iter *iter) · line 2484

Quoted 10 lines

fs/ocfs2/aops.c +7 -0

+
+		ret = ocfs2_range_has_holes(inode, iocb->ki_pos, iter->count);
+		if (ret == 1)
+			return 0;
+		else if (ret < 0)
+			return ret;
+	}
 
  	if (iov_iter_rw(iter) == READ)
  		get_block = ocfs2_lock_get_block;

_______________________________________________
Ocfs2-devel mailing list
Ocfs2-devel at oss.oracle.com
https://oss.oracle.com/mailman/listinfo/ocfs2-devel

Hi Changwei,

Quoted 7 lines · Changwei Ge

The intention of this patch is to provide an option to ocfs2 users whether to allocate disk space while doing dio write.

Whether to make ocfs2 fall back to buffer io is up to ocfs2 users through toggling append-dio feature. It rather makes ocfs2 configurable and flexible.

It is too strange to make ocfs2 fall back to buffer io by toggling append-dio feature.

Quoted 4 lines

So if something bad happens to dio write with space allocation, we can still make ocfs2 fall back to buffer io. It's an option not a mandatory action.:)

Now the ocfs2 supports fill holes during direct io whether or not supporting append-dio feature and we can directly fix the problem.
I think it is meaningless to provide an temporary option to turn off it.

Quoted 89 lines

Besides, append-dio feature is key to whether to allocate space with dio writing. So writing to file hole and enlarging file(appending file) should have the same reflection to append-dio feature.

Signed-off-by Changwei Ge <[email protected]>

---
   fs/ocfs2/aops.c | 53 ++++++++++++++++++++++++++++++++++++++++++++++++++---
   1 file changed, 50 insertions(+), 3 deletions(-)

fs/ocfs2/aops.c +50 -3

index d151632..32e60c0 100644
@@ -2414,12 +2414,52 @@ static int ocfs2_dio_end_io(struct kiocb *iocb,
   	return ret;
   }
 
+/*
+ * Will look for holes and unwritten extents in the range starting at
+ * pos for count bytes (inclusive).
+ * Return value 1 indicates hole exists, 0 not exists, others indicate error.
+ */
+static int ocfs2_range_has_holes(struct inode *inode, loff_t pos,
+				 size_t count)
+{
+	int ret = 0;
+	unsigned int extent_flags;
+	u32 cpos, clusters, extent_len, phys_cpos;
+	struct super_block *sb = inode->i_sb;
+
+	cpos = pos >> OCFS2_SB(sb)->s_clustersize_bits;
+	clusters = ocfs2_clusters_for_bytes(sb, pos + count) - cpos;
+
+	while (clusters) {
+		ret = ocfs2_get_clusters(inode, cpos, &phys_cpos, &extent_len,
+					 &extent_flags);
+		if (ret < 0) {
+			mlog_errno(ret);
+			goto out;
+		}
+
+		if (phys_cpos == 0) {
+			ret = 1;
+			goto out;
+		}
+
+		if (extent_len > clusters)
+			extent_len = clusters;
+
+		clusters -= extent_len;
+		cpos += extent_len;
+	}
+out:
+	return ret;
+}
+
   static ssize_t ocfs2_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
   {
   	struct file *file = iocb->ki_filp;
   	struct inode *inode = file->f_mapping->host;
   	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
   	get_block_t *get_block;
+	int ret;
 
   	/*
   	 * Fallback to buffered I/O if we see an inode without

@@ -2429,9 +2469,16 @@ static ssize_t ocfs2_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
   		return 0;
 
   	/* Fallback to buffered I/O if we do not support append dio. */
-	if (iocb->ki_pos + iter->count > i_size_read(inode) &&
-	    !ocfs2_supports_append_dio(osb))
-		return 0;
+	if (!ocfs2_supports_append_dio(osb)) {
+		if (iocb->ki_pos + iter->count > i_size_read(inode))
+			return 0;
+
+		ret = ocfs2_range_has_holes(inode, iocb->ki_pos, iter->count);
+		if (ret == 1)
+			return 0;
+		else if (ret < 0)
+			return ret;
+	}
 
   	if (iov_iter_rw(iter) == READ)
   		get_block = ocfs2_lock_get_block;

Hi Alex

Quoted 10 lines · alex chen

Hi Changwei,

Quoted 7 lines · Changwei Ge

The intention of this patch is to provide an option to ocfs2 users whether to allocate disk space while doing dio write.

Whether to make ocfs2 fall back to buffer io is up to ocfs2 users through toggling append-dio feature. It rather makes ocfs2 configurable and flexible.

Quoted 1 line

It is too strange to make ocfs2 fall back to buffer io by toggling append-dio feature.

It might be.
But as my changelog said, I think, append-dio is key to whether to allocate space with dio writing. So filling hole and appending file should have the same reflection.

Besides, in the early days, ocfs2 truly falls back to buffer io when append-dio is disabled and file hole is encountered.

Thanks,
Changwei

Quoted 7 lines

Quoted 4 lines

So if something bad happens to dio write with space allocation, we can still make ocfs2 fall back to buffer io. It's an option not a mandatory action.:)

Now the ocfs2 supports fill holes during direct io whether or not supporting append-dio feature and we can directly fix the problem.
I think it is meaningless to provide an temporary option to turn off it.

IMO, this patch is NOT a temporary solution.
Instead, it provides an extra option to ocfs2 users. And append-dio is enabled while making fs by *default*.

Thanks,
Changwei

Quoted 91 lines

Quoted 89 lines

Besides, append-dio feature is key to whether to allocate space with dio writing. So writing to file hole and enlarging file(appending file) should have the same reflection to append-dio feature.

Signed-off-by Changwei Ge <[email protected]>

---
   fs/ocfs2/aops.c | 53 ++++++++++++++++++++++++++++++++++++++++++++++++++---
   1 file changed, 50 insertions(+), 3 deletions(-)

fs/ocfs2/aops.c +50 -3

index d151632..32e60c0 100644
@@ -2414,12 +2414,52 @@ static int ocfs2_dio_end_io(struct kiocb *iocb,
   	return ret;
   }
 
+/*
+ * Will look for holes and unwritten extents in the range starting at
+ * pos for count bytes (inclusive).
+ * Return value 1 indicates hole exists, 0 not exists, others indicate error.
+ */
+static int ocfs2_range_has_holes(struct inode *inode, loff_t pos,
+				 size_t count)
+{
+	int ret = 0;
+	unsigned int extent_flags;
+	u32 cpos, clusters, extent_len, phys_cpos;
+	struct super_block *sb = inode->i_sb;
+
+	cpos = pos >> OCFS2_SB(sb)->s_clustersize_bits;
+	clusters = ocfs2_clusters_for_bytes(sb, pos + count) - cpos;
+
+	while (clusters) {
+		ret = ocfs2_get_clusters(inode, cpos, &phys_cpos, &extent_len,
+					 &extent_flags);
+		if (ret < 0) {
+			mlog_errno(ret);
+			goto out;
+		}
+
+		if (phys_cpos == 0) {
+			ret = 1;
+			goto out;
+		}
+
+		if (extent_len > clusters)
+			extent_len = clusters;
+
+		clusters -= extent_len;
+		cpos += extent_len;
+	}
+out:
+	return ret;
+}
+
   static ssize_t ocfs2_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
   {
   	struct file *file = iocb->ki_filp;
   	struct inode *inode = file->f_mapping->host;
   	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
   	get_block_t *get_block;
+	int ret;
 
   	/*
   	 * Fallback to buffered I/O if we see an inode without

@@ -2429,9 +2469,16 @@ static ssize_t ocfs2_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
   		return 0;
 
   	/* Fallback to buffered I/O if we do not support append dio. */
-	if (iocb->ki_pos + iter->count > i_size_read(inode) &&
-	    !ocfs2_supports_append_dio(osb))
-		return 0;
+	if (!ocfs2_supports_append_dio(osb)) {
+		if (iocb->ki_pos + iter->count > i_size_read(inode))
+			return 0;
+
+		ret = ocfs2_range_has_holes(inode, iocb->ki_pos, iter->count);
+		if (ret == 1)
+			return 0;
+		else if (ret < 0)
+			return ret;
+	}
 
   	if (iov_iter_rw(iter) == READ)
   		get_block = ocfs2_lock_get_block;

Hi Changwei,

Quoted 15 lines · Changwei Ge

Hi Alex

Quoted 11 lines · alex chen

Hi Changwei,

Quoted 7 lines · Changwei Ge

The intention of this patch is to provide an option to ocfs2 users whether to allocate disk space while doing dio write.

Whether to make ocfs2 fall back to buffer io is up to ocfs2 users through toggling append-dio feature. It rather makes ocfs2 configurable and flexible.

It is too strange to make ocfs2 fall back to buffer io by toggling append-dio feature.

Quoted 5 lines

It might be.
But as my changelog said, I think, append-dio is key to whether to allocate space with dio writing. So filling hole and appending file should have the same reflection.

Besides, in the early days, ocfs2 truly falls back to buffer io when append-dio is disabled and file hole is encountered.

I think we have discussed a lot in your v1 version. This wasn't fit for mainline, you can keep one off-mainline patch.

Thanks,
Junxiao.

Hi Junxiao,

Quoted 21 lines · Junxiao Bi

Hi Changwei,

Quoted 18 lines · Changwei Ge

Hi Alex

Quoted 11 lines · alex chen

Hi Changwei,

Quoted 7 lines · Changwei Ge

The intention of this patch is to provide an option to ocfs2 users whether to allocate disk space while doing dio write.

Whether to make ocfs2 fall back to buffer io is up to ocfs2 users through toggling append-dio feature. It rather makes ocfs2 configurable and flexible.

It is too strange to make ocfs2 fall back to buffer io by toggling append-dio feature.

It might be.
But as my changelog said, I think, append-dio is key to whether to allocate space with dio writing. So filling hole and appending file should have the same reflection.

Quoted 3 lines

Quoted 1 line

Besides, in the early days, ocfs2 truly falls back to buffer io when append-dio is disabled and file hole is encountered.

I think we have discussed a lot in your v1 version. This wasn't fit for mainline, you can keep one off-mainline patch.

Fine. I agree now.
I give up trying to push this patch into mainline.
BTW, could you please help review my another patch (ocfs2: try to reuse extent block in dealloc without meta_alloc) which is used to fix a dio crash issue? So without this patch, dio still won't hit such an issue.
Your comments are very important to me.

Thanks,
Changwei

Quoted 3 lines

Thanks,
Junxiao.