diff mbox series

[f2fs-dev,v2,3/4] f2fs: atomic: fix to truncate pagecache before on-disk metadata truncation

Message ID 20240726012204.1306174-1-chao@kernel.org (mailing list archive)
State New
Headers show
Series None | expand

Commit Message

Chao Yu July 26, 2024, 1:22 a.m. UTC
We should always truncate pagecache while truncating on-disk data.

Fixes: a46bebd502fe ("f2fs: synchronize atomic write aborts")
Signed-off-by: Chao Yu <chao@kernel.org>
---
v2:
- fix to use cow_inode instead of inode
 fs/f2fs/file.c | 1 +
 1 file changed, 1 insertion(+)

Comments

Jaegeuk Kim July 29, 2024, 4:26 p.m. UTC | #1
On 07/26, Chao Yu wrote:
> We should always truncate pagecache while truncating on-disk data.
> 
> Fixes: a46bebd502fe ("f2fs: synchronize atomic write aborts")
> Signed-off-by: Chao Yu <chao@kernel.org>
> ---
> v2:
> - fix to use cow_inode instead of inode
>  fs/f2fs/file.c | 1 +
>  1 file changed, 1 insertion(+)
> 
> diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c
> index 6c62f76474d1..54886ddcb8ab 100644
> --- a/fs/f2fs/file.c
> +++ b/fs/f2fs/file.c
> @@ -2202,6 +2202,7 @@ static int f2fs_ioc_start_atomic_write(struct file *filp, bool truncate)
>  		F2FS_I(fi->cow_inode)->atomic_inode = inode;
>  	} else {
>  		/* Reuse the already created COW inode */
> +		truncate_setsize(fi->cow_inode, 0);

What if the below truncation failed?

>  		ret = f2fs_do_truncate_blocks(fi->cow_inode, 0, true);
>  		if (ret) {
>  			f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
> -- 
> 2.40.1
Chao Yu July 30, 2024, 1:29 a.m. UTC | #2
On 2024/7/30 0:26, Jaegeuk Kim wrote:
> On 07/26, Chao Yu wrote:
>> We should always truncate pagecache while truncating on-disk data.
>>
>> Fixes: a46bebd502fe ("f2fs: synchronize atomic write aborts")
>> Signed-off-by: Chao Yu <chao@kernel.org>
>> ---
>> v2:
>> - fix to use cow_inode instead of inode
>>   fs/f2fs/file.c | 1 +
>>   1 file changed, 1 insertion(+)
>>
>> diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c
>> index 6c62f76474d1..54886ddcb8ab 100644
>> --- a/fs/f2fs/file.c
>> +++ b/fs/f2fs/file.c
>> @@ -2202,6 +2202,7 @@ static int f2fs_ioc_start_atomic_write(struct file *filp, bool truncate)
>>   		F2FS_I(fi->cow_inode)->atomic_inode = inode;
>>   	} else {
>>   		/* Reuse the already created COW inode */
>> +		truncate_setsize(fi->cow_inode, 0);
> 
> What if the below truncation failed?

What about just dropping page cache and do not set isize to 0?

Thanks,

> 
>>   		ret = f2fs_do_truncate_blocks(fi->cow_inode, 0, true);
>>   		if (ret) {
>>   			f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
>> -- 
>> 2.40.1
Jaegeuk Kim July 30, 2024, 4:26 p.m. UTC | #3
On 07/30, Chao Yu wrote:
> On 2024/7/30 0:26, Jaegeuk Kim wrote:
> > On 07/26, Chao Yu wrote:
> > > We should always truncate pagecache while truncating on-disk data.
> > > 
> > > Fixes: a46bebd502fe ("f2fs: synchronize atomic write aborts")
> > > Signed-off-by: Chao Yu <chao@kernel.org>
> > > ---
> > > v2:
> > > - fix to use cow_inode instead of inode
> > >   fs/f2fs/file.c | 1 +
> > >   1 file changed, 1 insertion(+)
> > > 
> > > diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c
> > > index 6c62f76474d1..54886ddcb8ab 100644
> > > --- a/fs/f2fs/file.c
> > > +++ b/fs/f2fs/file.c
> > > @@ -2202,6 +2202,7 @@ static int f2fs_ioc_start_atomic_write(struct file *filp, bool truncate)
> > >   		F2FS_I(fi->cow_inode)->atomic_inode = inode;
> > >   	} else {
> > >   		/* Reuse the already created COW inode */
> > > +		truncate_setsize(fi->cow_inode, 0);
> > 
> > What if the below truncation failed?
> 
> What about just dropping page cache and do not set isize to 0?

Can we also check if there's any dirty page before truncating it?

> 
> Thanks,
> 
> > 
> > >   		ret = f2fs_do_truncate_blocks(fi->cow_inode, 0, true);
> > >   		if (ret) {
> > >   			f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
> > > -- 
> > > 2.40.1
Chao Yu July 31, 2024, 8:18 a.m. UTC | #4
On 2024/7/31 0:26, Jaegeuk Kim wrote:
> On 07/30, Chao Yu wrote:
>> On 2024/7/30 0:26, Jaegeuk Kim wrote:
>>> On 07/26, Chao Yu wrote:
>>>> We should always truncate pagecache while truncating on-disk data.
>>>>
>>>> Fixes: a46bebd502fe ("f2fs: synchronize atomic write aborts")
>>>> Signed-off-by: Chao Yu <chao@kernel.org>
>>>> ---
>>>> v2:
>>>> - fix to use cow_inode instead of inode
>>>>    fs/f2fs/file.c | 1 +
>>>>    1 file changed, 1 insertion(+)
>>>>
>>>> diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c
>>>> index 6c62f76474d1..54886ddcb8ab 100644
>>>> --- a/fs/f2fs/file.c
>>>> +++ b/fs/f2fs/file.c
>>>> @@ -2202,6 +2202,7 @@ static int f2fs_ioc_start_atomic_write(struct file *filp, bool truncate)
>>>>    		F2FS_I(fi->cow_inode)->atomic_inode = inode;
>>>>    	} else {
>>>>    		/* Reuse the already created COW inode */
>>>> +		truncate_setsize(fi->cow_inode, 0);
>>>
>>> What if the below truncation failed?
>>
>> What about just dropping page cache and do not set isize to 0?
> 
> Can we also check if there's any dirty page before truncating it?

Sure, so how about this?

		/* Reuse the already created COW inode */
		if (get_dirty_pages(fi->cow_inode)) {
			ret = filemap_write_and_wait(fi->cow_inode->i_mapping);
			if (ret) {
				f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
				goto out;
			}
		}

		truncate_inode_pages(fi->cow_inode->i_mapping, 0);

		ret = f2fs_do_truncate_blocks(fi->cow_inode, 0, true);
		if (ret) {
			f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
			goto out;
		}

		i_size_write(fi->cow_inode, 0);

Thanks,

> 
>>
>> Thanks,
>>
>>>
>>>>    		ret = f2fs_do_truncate_blocks(fi->cow_inode, 0, true);
>>>>    		if (ret) {
>>>>    			f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
>>>> -- 
>>>> 2.40.1
Jaegeuk Kim July 31, 2024, 8:27 p.m. UTC | #5
On 07/31, Chao Yu wrote:
> On 2024/7/31 0:26, Jaegeuk Kim wrote:
> > On 07/30, Chao Yu wrote:
> > > On 2024/7/30 0:26, Jaegeuk Kim wrote:
> > > > On 07/26, Chao Yu wrote:
> > > > > We should always truncate pagecache while truncating on-disk data.
> > > > > 
> > > > > Fixes: a46bebd502fe ("f2fs: synchronize atomic write aborts")
> > > > > Signed-off-by: Chao Yu <chao@kernel.org>
> > > > > ---
> > > > > v2:
> > > > > - fix to use cow_inode instead of inode
> > > > >    fs/f2fs/file.c | 1 +
> > > > >    1 file changed, 1 insertion(+)
> > > > > 
> > > > > diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c
> > > > > index 6c62f76474d1..54886ddcb8ab 100644
> > > > > --- a/fs/f2fs/file.c
> > > > > +++ b/fs/f2fs/file.c
> > > > > @@ -2202,6 +2202,7 @@ static int f2fs_ioc_start_atomic_write(struct file *filp, bool truncate)
> > > > >    		F2FS_I(fi->cow_inode)->atomic_inode = inode;
> > > > >    	} else {
> > > > >    		/* Reuse the already created COW inode */
> > > > > +		truncate_setsize(fi->cow_inode, 0);
> > > > 
> > > > What if the below truncation failed?
> > > 
> > > What about just dropping page cache and do not set isize to 0?
> > 
> > Can we also check if there's any dirty page before truncating it?
> 
> Sure, so how about this?
> 
> 		/* Reuse the already created COW inode */
> 		if (get_dirty_pages(fi->cow_inode)) {

Was supposed to be a bug here?

> 			ret = filemap_write_and_wait(fi->cow_inode->i_mapping);
> 			if (ret) {
> 				f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
> 				goto out;
> 			}
> 		}
> 
> 		truncate_inode_pages(fi->cow_inode->i_mapping, 0);

invalidate_mapping_pages() would be much safer?

> 
> 		ret = f2fs_do_truncate_blocks(fi->cow_inode, 0, true);
> 		if (ret) {
> 			f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
> 			goto out;
> 		}
> 
> 		i_size_write(fi->cow_inode, 0);
> 
> Thanks,
> 
> > 
> > > 
> > > Thanks,
> > > 
> > > > 
> > > > >    		ret = f2fs_do_truncate_blocks(fi->cow_inode, 0, true);
> > > > >    		if (ret) {
> > > > >    			f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
> > > > > -- 
> > > > > 2.40.1
diff mbox series

Patch

diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c
index 6c62f76474d1..54886ddcb8ab 100644
--- a/fs/f2fs/file.c
+++ b/fs/f2fs/file.c
@@ -2202,6 +2202,7 @@  static int f2fs_ioc_start_atomic_write(struct file *filp, bool truncate)
 		F2FS_I(fi->cow_inode)->atomic_inode = inode;
 	} else {
 		/* Reuse the already created COW inode */
+		truncate_setsize(fi->cow_inode, 0);
 		ret = f2fs_do_truncate_blocks(fi->cow_inode, 0, true);
 		if (ret) {
 			f2fs_up_write(&fi->i_gc_rwsem[WRITE]);