[PATCH] Btrfs: fix btrfs_ordered_update_i_size to update disk_i_size properly

[PATCH] Btrfs: fix btrfs_ordered_update_i_size to update disk_i_size properly

[Liu Bo]

btrfs_ordered_update_i_size can be called by truncate and endio, but only endio
takes ordered_extent which contains the completed IO.

while truncating down a file, if there are some in-flight IOs,
btrfs_ordered_update_i_size in endio will set disk_i_size to @orig_offset that
is zero.  If truncating-down fails somehow, we try to recover in memory isize
with this zero'd disk_i_size.

Fix it by only updating disk_i_size with @orig_offset when
btrfs_ordered_update_i_size is not called from endio while truncating down and
waiting for in-flight IOs completing their work before recover in-memory size.

Besides fixing the above issue, add an assertion for last_size to double check
we truncate down to the desired size.

Signed-off-by: Liu Bo <bo.li.liu@oracle.com>
---
 fs/btrfs/inode.c        | 14 ++++++++++++++
 fs/btrfs/ordered-data.c |  9 +++++++--
 2 files changed, 21 insertions(+), 2 deletions(-)

diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 09157dd..ef3594d 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -4682,6 +4682,13 @@ int btrfs_truncate_inode_items(struct btrfs_trans_handle *trans,
 
 	btrfs_free_path(path);
 
+	if (err == 0) {
+		/* only inline file may have last_size != new_size */
+		if (new_size >= root->sectorsize ||
+		    new_size > root->fs_info->max_inline)
+			ASSERT(last_size == new_size);
+	}
+
 	if (be_nice && bytes_deleted > SZ_32M) {
 		unsigned long updates = trans->delayed_ref_updates;
 		if (updates) {
@@ -5064,6 +5071,13 @@ static int btrfs_setsize(struct inode *inode, struct iattr *attr)
 		if (ret && inode->i_nlink) {
 			int err;
 
+			/* To get a stable disk_i_size */
+			err = btrfs_wait_ordered_range(inode, 0, (u64)-1);
+			if (err) {
+				btrfs_orphan_del(NULL, inode);
+				return err;
+			}
+
 			/*
 			 * failed to truncate, disk_i_size is only adjusted down
 			 * as we remove extents, so it should represent the true
diff --git a/fs/btrfs/ordered-data.c b/fs/btrfs/ordered-data.c
index b2d1e95..5eaa25a 100644
--- a/fs/btrfs/ordered-data.c
+++ b/fs/btrfs/ordered-data.c
@@ -982,8 +982,13 @@ int btrfs_ordered_update_i_size(struct inode *inode, u64 offset,
 	}
 	disk_i_size = BTRFS_I(inode)->disk_i_size;
 
-	/* truncate file */
-	if (disk_i_size > i_size) {
+	/*
+	 * truncate file.
+	 * If ordered is not NULL, then this is called from endio and
+	 * disk_i_size will be updated by either truncate itself or any
+	 * in-flight IOs which are inside the disk_i_size.
+	 */
+	if (!ordered && disk_i_size > i_size) {
 		BTRFS_I(inode)->disk_i_size = orig_offset;
 		ret = 0;
 		goto out;
-- 
2.5.5

Re: [PATCH] Btrfs: fix btrfs_ordered_update_i_size to update disk_i_size properly

[Liu Bo]

On Thu, Dec 01, 2016 at 01:46:10PM -0800, Liu Bo wrote:
> btrfs_ordered_update_i_size can be called by truncate and endio, but only endio
> takes ordered_extent which contains the completed IO.
> 
> while truncating down a file, if there are some in-flight IOs,
> btrfs_ordered_update_i_size in endio will set disk_i_size to @orig_offset that
> is zero.  If truncating-down fails somehow, we try to recover in memory isize
> with this zero'd disk_i_size.
> 
> Fix it by only updating disk_i_size with @orig_offset when
> btrfs_ordered_update_i_size is not called from endio while truncating down and
> waiting for in-flight IOs completing their work before recover in-memory size.
> 
> Besides fixing the above issue, add an assertion for last_size to double check
> we truncate down to the desired size.
> 
> Signed-off-by: Liu Bo <bo.li.liu@oracle.com>
> ---
>  fs/btrfs/inode.c        | 14 ++++++++++++++
>  fs/btrfs/ordered-data.c |  9 +++++++--
>  2 files changed, 21 insertions(+), 2 deletions(-)
> 
> diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
> index 09157dd..ef3594d 100644
> --- a/fs/btrfs/inode.c
> +++ b/fs/btrfs/inode.c
> @@ -4682,6 +4682,13 @@ int btrfs_truncate_inode_items(struct btrfs_trans_handle *trans,
>  
>  	btrfs_free_path(path);
>  
> +	if (err == 0) {
> +		/* only inline file may have last_size != new_size */
> +		if (new_size >= root->sectorsize ||
> +		    new_size > root->fs_info->max_inline)
> +			ASSERT(last_size == new_size);
> +	}
> +

This ASSERT has been hit by fstests/generic/299, and it didn't show up
the first time I tested, I'm trying to figure out whether we have
problems in code or in this ASSERT.

Thanks,

-liubo

>  	if (be_nice && bytes_deleted > SZ_32M) {
>  		unsigned long updates = trans->delayed_ref_updates;
>  		if (updates) {
> @@ -5064,6 +5071,13 @@ static int btrfs_setsize(struct inode *inode, struct iattr *attr)
>  		if (ret && inode->i_nlink) {
>  			int err;
>  
> +			/* To get a stable disk_i_size */
> +			err = btrfs_wait_ordered_range(inode, 0, (u64)-1);
> +			if (err) {
> +				btrfs_orphan_del(NULL, inode);
> +				return err;
> +			}
> +
>  			/*
>  			 * failed to truncate, disk_i_size is only adjusted down
>  			 * as we remove extents, so it should represent the true
> diff --git a/fs/btrfs/ordered-data.c b/fs/btrfs/ordered-data.c
> index b2d1e95..5eaa25a 100644
> --- a/fs/btrfs/ordered-data.c
> +++ b/fs/btrfs/ordered-data.c
> @@ -982,8 +982,13 @@ int btrfs_ordered_update_i_size(struct inode *inode, u64 offset,
>  	}
>  	disk_i_size = BTRFS_I(inode)->disk_i_size;
>  
> -	/* truncate file */
> -	if (disk_i_size > i_size) {
> +	/*
> +	 * truncate file.
> +	 * If ordered is not NULL, then this is called from endio and
> +	 * disk_i_size will be updated by either truncate itself or any
> +	 * in-flight IOs which are inside the disk_i_size.
> +	 */
> +	if (!ordered && disk_i_size > i_size) {
>  		BTRFS_I(inode)->disk_i_size = orig_offset;
>  		ret = 0;
>  		goto out;
> -- 
> 2.5.5
> 
> --
> To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html