2019-07-01 17:17:21

by Luis Henriques

[permalink] [raw]
Subject: [PATCH] ceph: fix end offset in truncate_inode_pages_range call

Commit e450f4d1a5d6 ("ceph: pass inclusive lend parameter to
filemap_write_and_wait_range()") fixed the end offset parameter used to
call filemap_write_and_wait_range and invalidate_inode_pages2_range.
Unfortunately it missed truncate_inode_pages_range, introducing a
regression that is easily detected by xfstest generic/130.

The problem is that when doing direct IO it is possible that an extra page
is truncated from the page cache when the end offset is page aligned.
This can cause data loss if that page hasn't been sync'ed to the OSDs.

While there, change code to use PAGE_ALIGN macro instead.

Fixes: e450f4d1a5d6 ("ceph: pass inclusive lend parameter to filemap_write_and_wait_range()")
Signed-off-by: Luis Henriques <[email protected]>
---
fs/ceph/file.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/fs/ceph/file.c b/fs/ceph/file.c
index 183c37c0a8fc..7a57db8e2fa9 100644
--- a/fs/ceph/file.c
+++ b/fs/ceph/file.c
@@ -1007,7 +1007,7 @@ ceph_direct_read_write(struct kiocb *iocb, struct iov_iter *iter,
* may block.
*/
truncate_inode_pages_range(inode->i_mapping, pos,
- (pos+len) | (PAGE_SIZE - 1));
+ PAGE_ALIGN(pos + len) - 1);

req->r_mtime = mtime;
}


2019-07-01 18:19:54

by Jeff Layton

[permalink] [raw]
Subject: Re: [PATCH] ceph: fix end offset in truncate_inode_pages_range call

On Mon, 2019-07-01 at 18:16 +0100, Luis Henriques wrote:
> Commit e450f4d1a5d6 ("ceph: pass inclusive lend parameter to
> filemap_write_and_wait_range()") fixed the end offset parameter used to
> call filemap_write_and_wait_range and invalidate_inode_pages2_range.
> Unfortunately it missed truncate_inode_pages_range, introducing a
> regression that is easily detected by xfstest generic/130.
>
> The problem is that when doing direct IO it is possible that an extra page
> is truncated from the page cache when the end offset is page aligned.
> This can cause data loss if that page hasn't been sync'ed to the OSDs.
>
> While there, change code to use PAGE_ALIGN macro instead.
>
> Fixes: e450f4d1a5d6 ("ceph: pass inclusive lend parameter to filemap_write_and_wait_range()")
> Signed-off-by: Luis Henriques <[email protected]>
> ---
> fs/ceph/file.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/fs/ceph/file.c b/fs/ceph/file.c
> index 183c37c0a8fc..7a57db8e2fa9 100644
> --- a/fs/ceph/file.c
> +++ b/fs/ceph/file.c
> @@ -1007,7 +1007,7 @@ (struct kiocb *iocb, struct iov_iter *iter,
> * may block.
> */
> truncate_inode_pages_range(inode->i_mapping, pos,
> - (pos+len) | (PAGE_SIZE - 1));
> + PAGE_ALIGN(pos + len) - 1);
>
> req->r_mtime = mtime;
> }

Reviewed-by: Jeff Layton <[email protected]>

2019-07-02 13:48:59

by Jeff Layton

[permalink] [raw]
Subject: Re: [PATCH] ceph: fix end offset in truncate_inode_pages_range call

On Mon, 2019-07-01 at 18:16 +0100, Luis Henriques wrote:
> Commit e450f4d1a5d6 ("ceph: pass inclusive lend parameter to
> filemap_write_and_wait_range()") fixed the end offset parameter used to
> call filemap_write_and_wait_range and invalidate_inode_pages2_range.
> Unfortunately it missed truncate_inode_pages_range, introducing a
> regression that is easily detected by xfstest generic/130.
>
> The problem is that when doing direct IO it is possible that an extra page
> is truncated from the page cache when the end offset is page aligned.
> This can cause data loss if that page hasn't been sync'ed to the OSDs.
>
> While there, change code to use PAGE_ALIGN macro instead.
>
> Fixes: e450f4d1a5d6 ("ceph: pass inclusive lend parameter to filemap_write_and_wait_range()")
> Signed-off-by: Luis Henriques <[email protected]>
> ---
> fs/ceph/file.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/fs/ceph/file.c b/fs/ceph/file.c
> index 183c37c0a8fc..7a57db8e2fa9 100644
> --- a/fs/ceph/file.c
> +++ b/fs/ceph/file.c
> @@ -1007,7 +1007,7 @@ ceph_direct_read_write(struct kiocb *iocb, struct iov_iter *iter,
> * may block.
> */
> truncate_inode_pages_range(inode->i_mapping, pos,
> - (pos+len) | (PAGE_SIZE - 1));
> + PAGE_ALIGN(pos + len) - 1);
>
> req->r_mtime = mtime;
> }

Luis, should this be sent to stable? It seems like a data corruption
problem...

Thanks,
--
Jeff Layton <[email protected]>

2019-07-02 14:00:16

by Luis Henriques

[permalink] [raw]
Subject: Re: [PATCH] ceph: fix end offset in truncate_inode_pages_range call

"Jeff Layton" <[email protected]> writes:

> On Mon, 2019-07-01 at 18:16 +0100, Luis Henriques wrote:
>> Commit e450f4d1a5d6 ("ceph: pass inclusive lend parameter to
>> filemap_write_and_wait_range()") fixed the end offset parameter used to
>> call filemap_write_and_wait_range and invalidate_inode_pages2_range.
>> Unfortunately it missed truncate_inode_pages_range, introducing a
>> regression that is easily detected by xfstest generic/130.
>>
>> The problem is that when doing direct IO it is possible that an extra page
>> is truncated from the page cache when the end offset is page aligned.
>> This can cause data loss if that page hasn't been sync'ed to the OSDs.
>>
>> While there, change code to use PAGE_ALIGN macro instead.
>>
>> Fixes: e450f4d1a5d6 ("ceph: pass inclusive lend parameter to filemap_write_and_wait_range()")
>> Signed-off-by: Luis Henriques <[email protected]>
>> ---
>> fs/ceph/file.c | 2 +-
>> 1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/fs/ceph/file.c b/fs/ceph/file.c
>> index 183c37c0a8fc..7a57db8e2fa9 100644
>> --- a/fs/ceph/file.c
>> +++ b/fs/ceph/file.c
>> @@ -1007,7 +1007,7 @@ ceph_direct_read_write(struct kiocb *iocb, struct iov_iter *iter,
>> * may block.
>> */
>> truncate_inode_pages_range(inode->i_mapping, pos,
>> - (pos+len) | (PAGE_SIZE - 1));
>> + PAGE_ALIGN(pos + len) - 1);
>>
>> req->r_mtime = mtime;
>> }
>
> Luis, should this be sent to stable? It seems like a data corruption
> problem...

Yes, I believe so. But I believe all the active stable kernels that
include commit e450f4d1a5d6 (or a backport of it) will pick it anyway
due to the 'Fixes:' tag. AFAIK only 5.1 and 5.2 are affected.

Cheers,
--
Luis