diff mbox

ocfs2: fix issue that ocfs2_setattr() does not deal with new_i_size==i_size

Message ID 51CBEFEA.3010308@huawei.com
State New, archived
Headers show

Commit Message

Younger Liu June 27, 2013, 7:55 a.m. UTC
The issue scenario is as following:
1. fallocate a large disk space(eg. 30G) with FALLOC_FL_KEEP_SIZE
for a file whose i_size and disk size are 512(or other size). 
After executing fallocate, i_size file is still 512, and the disk 
size became to 30G+512?
2. ftruncate the file to new_i_size which equal to inode->i_size.
After executing ftruncate, disk space does not changes. 
In other words, i_size file is still 512, and disk size is
30G+512. But we want disk size to be 512.
This does not meet our expectations.

In order to Solving the issue above, we modified ocfs2_setattr(), 
if attr->ia_size != i_size_read(inode), It calls 
ocfs2_truncate_file(), and truncate disk space to attr->ia_size.

Signed-off-by: Younger Liu <younger.liu@huawei.com>
---
 fs/ocfs2/alloc.c |    2 +-
 fs/ocfs2/file.c  |    7 ++-----
 2 files changed, 3 insertions(+), 6 deletions(-)

Comments

Shen Canquan June 28, 2013, 1:42 a.m. UTC | #1
It looks good to me.

reviewed-by: Jensen <shencanquan@huawei.com>

On 2013/6/27 15:55, Younger Liu wrote:

> The issue scenario is as following:
> 1. fallocate a large disk space(eg. 30G) with FALLOC_FL_KEEP_SIZE
> for a file whose i_size and disk size are 512(or other size). 
> After executing fallocate, i_size file is still 512, and the disk 
> size became to 30G+512?
> 2. ftruncate the file to new_i_size which equal to inode->i_size.
> After executing ftruncate, disk space does not changes. 
> In other words, i_size file is still 512, and disk size is
> 30G+512. But we want disk size to be 512.
> This does not meet our expectations.
> 
> In order to Solving the issue above, we modified ocfs2_setattr(), 
> if attr->ia_size != i_size_read(inode), It calls 
> ocfs2_truncate_file(), and truncate disk space to attr->ia_size.
> 
> Signed-off-by: Younger Liu <younger.liu@huawei.com>
> ---
>  fs/ocfs2/alloc.c |    2 +-
>  fs/ocfs2/file.c  |    7 ++-----
>  2 files changed, 3 insertions(+), 6 deletions(-)
> 
> diff --git a/fs/ocfs2/alloc.c b/fs/ocfs2/alloc.c
> index b8a9d87..19837d4 100644
> --- a/fs/ocfs2/alloc.c
> +++ b/fs/ocfs2/alloc.c
> @@ -7126,7 +7126,7 @@ int ocfs2_truncate_inline(struct inode *inode, struct buffer_head *di_bh,
>  	if (end > i_size_read(inode))
>  		end = i_size_read(inode);
>  
> -	BUG_ON(start >= end);
> +	BUG_ON(start > end);
>  
>  	if (!(OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) ||
>  	    !(le16_to_cpu(di->i_dyn_features) & OCFS2_INLINE_DATA_FL) ||
> diff --git a/fs/ocfs2/file.c b/fs/ocfs2/file.c
> index 793c010..2e405e8 100644
> --- a/fs/ocfs2/file.c
> +++ b/fs/ocfs2/file.c
> @@ -476,9 +476,6 @@ static int ocfs2_truncate_file(struct inode *inode,
>  
>  	/* lets handle the simple truncate cases before doing any more
>  	 * cluster locking. */
> -	if (new_i_size == le64_to_cpu(fe->i_size))
> -		goto bail;
> -
>  	down_write(&OCFS2_I(inode)->ip_alloc_sem);
>  
>  	ocfs2_resv_discard(&osb->osb_la_resmap,
> @@ -1150,14 +1147,14 @@ int ocfs2_setattr(struct dentry *dentry, struct iattr *attr)
>  		goto bail_unlock_rw;
>  	}
>  
> -	if (size_change && attr->ia_size != i_size_read(inode)) {
> +	if (size_change) {
>  		status = inode_newsize_ok(inode, attr->ia_size);
>  		if (status)
>  			goto bail_unlock;
>  
>  		inode_dio_wait(inode);
>  
> -		if (i_size_read(inode) > attr->ia_size) {
> +		if (i_size_read(inode) >= attr->ia_size) {
>  			if (ocfs2_should_order_data(inode)) {
>  				status = ocfs2_begin_ordered_truncate(inode,
>  								      attr->ia_size);
jeff.liu June 28, 2013, 4:06 a.m. UTC | #2
Thanks for your patch, it indeed fixed a bug.  However, your description
is not correct, please see my inline comments.

On 06/27/2013 03:55 PM, Younger Liu wrote:

> The issue scenario is as following:
> 1. fallocate a large disk space(eg. 30G) with FALLOC_FL_KEEP_SIZE
> for a file whose i_size and disk size are 512(or other size). 
> After executing fallocate, i_size file is still 512, and the disk 
> size became to 30G+512?

The file size won't be changed if FALLOC_FL_KEEP_SIZE is specified,
Pls refer to the man page of fallocate(2) for detail.

Disk size should be changed to reflect this operation, why not?

> 2. ftruncate the file to new_i_size which equal to inode->i_size.
> After executing ftruncate, disk space does not changes. 

This is a real bug that be fixed in this patch.

> In other words, i_size file is still 512, and disk size is
> 30G+512. But we want disk size to be 512.

Ditto.

> This does not meet our expectations.

...

> 
> In order to Solving the issue above, we modified ocfs2_setattr(), 
> if attr->ia_size != i_size_read(inode), It calls 
> ocfs2_truncate_file(), and truncate disk space to attr->ia_size.

Right.

> 
> Signed-off-by: Younger Liu <younger.liu@huawei.com>
> ---
>  fs/ocfs2/alloc.c |    2 +-
>  fs/ocfs2/file.c  |    7 ++-----
>  2 files changed, 3 insertions(+), 6 deletions(-)
> 
> diff --git a/fs/ocfs2/alloc.c b/fs/ocfs2/alloc.c
> index b8a9d87..19837d4 100644
> --- a/fs/ocfs2/alloc.c
> +++ b/fs/ocfs2/alloc.c
> @@ -7126,7 +7126,7 @@ int ocfs2_truncate_inline(struct inode *inode, struct buffer_head *di_bh,
>  	if (end > i_size_read(inode))
>  		end = i_size_read(inode);
>  
> -	BUG_ON(start >= end);
> +	BUG_ON(start > end);
>  
>  	if (!(OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) ||
>  	    !(le16_to_cpu(di->i_dyn_features) & OCFS2_INLINE_DATA_FL) ||
> diff --git a/fs/ocfs2/file.c b/fs/ocfs2/file.c
> index 793c010..2e405e8 100644
> --- a/fs/ocfs2/file.c
> +++ b/fs/ocfs2/file.c
> @@ -476,9 +476,6 @@ static int ocfs2_truncate_file(struct inode *inode,
>  
>  	/* lets handle the simple truncate cases before doing any more
>  	 * cluster locking. */

Above pointless comments should be removed with this change as well.

> -	if (new_i_size == le64_to_cpu(fe->i_size))
> -		goto bail;
> -
>  	down_write(&OCFS2_I(inode)->ip_alloc_sem);
>  
>  	ocfs2_resv_discard(&osb->osb_la_resmap,
> @@ -1150,14 +1147,14 @@ int ocfs2_setattr(struct dentry *dentry, struct iattr *attr)
>  		goto bail_unlock_rw;
>  	}
>  
> -	if (size_change && attr->ia_size != i_size_read(inode)) {
> +	if (size_change) {
>  		status = inode_newsize_ok(inode, attr->ia_size);
>  		if (status)
>  			goto bail_unlock;
>  
>  		inode_dio_wait(inode);
>  
> -		if (i_size_read(inode) > attr->ia_size) {
> +		if (i_size_read(inode) >= attr->ia_size) {
>  			if (ocfs2_should_order_data(inode)) {
>  				status = ocfs2_begin_ordered_truncate(inode,
>  								      attr->ia_size);


So I also run a quick test to verify this patch, it looks fine.

# dd if=/dev/zero of=/ocfs2/test bs=512 count=1
1+0 records in
1+0 records out
512 bytes (512 B) copied, 0.000150857 s, 3.4 MB/s

# xfs_io -c 'falloc -k 512 3G' /ocfs2/test
# ls -l /ocfs2/test
-rw-r--r-- 1 root root 512 Jun 28 11:47 /ocfs2/test
# df -h
Filesystem      Size  Used Avail Use% Mounted on
....
/dev/sdb1       8.0G  3.4G  4.7G  42% /ocfs2

# xfs_io -c 'truncate 512' /ocfs2/test

# df -h
Filesystem      Size  Used Avail Use% Mounted on
.....
/dev/sdb1       8.0G  363M  7.7G   5% /ocfs2

# ls -l /ocfs2/test
-rw-r--r-- 1 root root 512 Jun 28 11:47 /ocfs2/test


With above bug description problems were fixed, you can
consider to add:
Reviewed-by: Jie Liu <jeff.liu@oracle.com>
Tested-by: Jie Liu <jeff.liu@oracle.com>

-Jeff
Younger Liu June 29, 2013, 5:20 a.m. UTC | #3
On 2013/6/28 12:06, Jeff Liu wrote:
> Thanks for your patch, it indeed fixed a bug.  However, your description
> is not correct, please see my inline comments.
> 
> On 06/27/2013 03:55 PM, Younger Liu wrote:
> 
>> The issue scenario is as following:
>> 1. fallocate a large disk space(eg. 30G) with FALLOC_FL_KEEP_SIZE
>> for a file whose i_size and disk size are 512(or other size). 
>> After executing fallocate, i_size file is still 512, and the disk 
>> size became to 30G+512?
> 
> The file size won't be changed if FALLOC_FL_KEEP_SIZE is specified,
> Pls refer to the man page of fallocate(2) for detail.
> 
> Disk size should be changed to reflect this operation, why not?
> 
>> 2. ftruncate the file to new_i_size which equal to inode->i_size.
>> After executing ftruncate, disk space does not changes. 
> 
> This is a real bug that be fixed in this patch.
> 
>> In other words, i_size file is still 512, and disk size is
>> 30G+512. But we want disk size to be 512.
> 
> Ditto.
> 
>> This does not meet our expectations.
> 
> ...
> 
>>
>> In order to Solving the issue above, we modified ocfs2_setattr(), 
>> if attr->ia_size != i_size_read(inode), It calls 
>> ocfs2_truncate_file(), and truncate disk space to attr->ia_size.
> 
> Right.
> 
>>
>> Signed-off-by: Younger Liu <younger.liu@huawei.com>
>> ---
>>  fs/ocfs2/alloc.c |    2 +-
>>  fs/ocfs2/file.c  |    7 ++-----
>>  2 files changed, 3 insertions(+), 6 deletions(-)
>>
>> diff --git a/fs/ocfs2/alloc.c b/fs/ocfs2/alloc.c
>> index b8a9d87..19837d4 100644
>> --- a/fs/ocfs2/alloc.c
>> +++ b/fs/ocfs2/alloc.c
>> @@ -7126,7 +7126,7 @@ int ocfs2_truncate_inline(struct inode *inode, struct buffer_head *di_bh,
>>  	if (end > i_size_read(inode))
>>  		end = i_size_read(inode);
>>  
>> -	BUG_ON(start >= end);
>> +	BUG_ON(start > end);
>>  
>>  	if (!(OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) ||
>>  	    !(le16_to_cpu(di->i_dyn_features) & OCFS2_INLINE_DATA_FL) ||
>> diff --git a/fs/ocfs2/file.c b/fs/ocfs2/file.c
>> index 793c010..2e405e8 100644
>> --- a/fs/ocfs2/file.c
>> +++ b/fs/ocfs2/file.c
>> @@ -476,9 +476,6 @@ static int ocfs2_truncate_file(struct inode *inode,
>>  
>>  	/* lets handle the simple truncate cases before doing any more
>>  	 * cluster locking. */
> 
> Above pointless comments should be removed with this change as well.
> 
>> -	if (new_i_size == le64_to_cpu(fe->i_size))
>> -		goto bail;
>> -
>>  	down_write(&OCFS2_I(inode)->ip_alloc_sem);
>>  
>>  	ocfs2_resv_discard(&osb->osb_la_resmap,
>> @@ -1150,14 +1147,14 @@ int ocfs2_setattr(struct dentry *dentry, struct iattr *attr)
>>  		goto bail_unlock_rw;
>>  	}
>>  
>> -	if (size_change && attr->ia_size != i_size_read(inode)) {
>> +	if (size_change) {
>>  		status = inode_newsize_ok(inode, attr->ia_size);
>>  		if (status)
>>  			goto bail_unlock;
>>  
>>  		inode_dio_wait(inode);
>>  
>> -		if (i_size_read(inode) > attr->ia_size) {
>> +		if (i_size_read(inode) >= attr->ia_size) {
>>  			if (ocfs2_should_order_data(inode)) {
>>  				status = ocfs2_begin_ordered_truncate(inode,
>>  								      attr->ia_size);
> 
> 
> So I also run a quick test to verify this patch, it looks fine.
> 
> # dd if=/dev/zero of=/ocfs2/test bs=512 count=1
> 1+0 records in
> 1+0 records out
> 512 bytes (512 B) copied, 0.000150857 s, 3.4 MB/s
> 
> # xfs_io -c 'falloc -k 512 3G' /ocfs2/test
> # ls -l /ocfs2/test
> -rw-r--r-- 1 root root 512 Jun 28 11:47 /ocfs2/test
> # df -h
> Filesystem      Size  Used Avail Use% Mounted on
> ....
> /dev/sdb1       8.0G  3.4G  4.7G  42% /ocfs2
> 
> # xfs_io -c 'truncate 512' /ocfs2/test
> 
> # df -h
> Filesystem      Size  Used Avail Use% Mounted on
> .....
> /dev/sdb1       8.0G  363M  7.7G   5% /ocfs2
> 
> # ls -l /ocfs2/test
> -rw-r--r-- 1 root root 512 Jun 28 11:47 /ocfs2/test
> 
> 
> With above bug description problems were fixed, you can
> consider to add:
> Reviewed-by: Jie Liu <jeff.liu@oracle.com>
> Tested-by: Jie Liu <jeff.liu@oracle.com>
> 
> -Jeff
> 
> .
> 
Thanks, I will resent the patch in a moment.
					Younger
diff mbox

Patch

diff --git a/fs/ocfs2/alloc.c b/fs/ocfs2/alloc.c
index b8a9d87..19837d4 100644
--- a/fs/ocfs2/alloc.c
+++ b/fs/ocfs2/alloc.c
@@ -7126,7 +7126,7 @@  int ocfs2_truncate_inline(struct inode *inode, struct buffer_head *di_bh,
 	if (end > i_size_read(inode))
 		end = i_size_read(inode);
 
-	BUG_ON(start >= end);
+	BUG_ON(start > end);
 
 	if (!(OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) ||
 	    !(le16_to_cpu(di->i_dyn_features) & OCFS2_INLINE_DATA_FL) ||
diff --git a/fs/ocfs2/file.c b/fs/ocfs2/file.c
index 793c010..2e405e8 100644
--- a/fs/ocfs2/file.c
+++ b/fs/ocfs2/file.c
@@ -476,9 +476,6 @@  static int ocfs2_truncate_file(struct inode *inode,
 
 	/* lets handle the simple truncate cases before doing any more
 	 * cluster locking. */
-	if (new_i_size == le64_to_cpu(fe->i_size))
-		goto bail;
-
 	down_write(&OCFS2_I(inode)->ip_alloc_sem);
 
 	ocfs2_resv_discard(&osb->osb_la_resmap,
@@ -1150,14 +1147,14 @@  int ocfs2_setattr(struct dentry *dentry, struct iattr *attr)
 		goto bail_unlock_rw;
 	}
 
-	if (size_change && attr->ia_size != i_size_read(inode)) {
+	if (size_change) {
 		status = inode_newsize_ok(inode, attr->ia_size);
 		if (status)
 			goto bail_unlock;
 
 		inode_dio_wait(inode);
 
-		if (i_size_read(inode) > attr->ia_size) {
+		if (i_size_read(inode) >= attr->ia_size) {
 			if (ocfs2_should_order_data(inode)) {
 				status = ocfs2_begin_ordered_truncate(inode,
 								      attr->ia_size);