Ocfs2 devel - Nov 2018 - [PATCH V2 2/2] ocfs2: clear journal dirty flag after shutdown journal

Dirty flag of the journal should be cleared at the last stage of umount,
if do it before jbd2_journal_destroy(), then some metadata in uncommitted
transaction could be lost due to io error, but as dirty flag of journal
was already cleared, we can't find that until run a full fsck. This may
cause system panic or other corruption.

Signed-off-by: Junxiao Bi <junxiao.bi at oracle.com>
Cc: Yiwen Jiang <jiangyiwen at huawei.com>
Cc: Jun Piao <piaojun at huawei.com>
---
 fs/ocfs2/journal.c | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)

 V1 -> V2:
 pointed by Yiwen, need check return value of jbd2_journal_destroy

diff --git a/fs/ocfs2/journal.c b/fs/ocfs2/journal.c
index 13f8e097babf..b51bb873441f 100644
--- a/fs/ocfs2/journal.c
+++ b/fs/ocfs2/journal.c
@@ -986,7 +986,8 @@ void ocfs2_journal_shutdown(struct ocfs2_super *osb)
 			mlog_errno(status);
 	}
 
-	if (status == 0) {
+	/* Shutdown the kernel journal system */
+	if (!jbd2_journal_destroy(journal->j_journal) && !status) {
 		/*
 		 * Do not toggle if flush was unsuccessful otherwise
 		 * will leave dirty metadata in a "clean" journal
@@ -995,9 +996,6 @@ void ocfs2_journal_shutdown(struct ocfs2_super *osb)
 		if (status < 0)
 			mlog_errno(status);
 	}
-
-	/* Shutdown the kernel journal system */
-	jbd2_journal_destroy(journal->j_journal);
 	journal->j_journal = NULL;
 
 	OCFS2_I(inode)->ip_open_count--;
-- 
2.17.1

On 2018/11/19 16:07, Junxiao Bi wrote:
> Dirty flag of the journal should be cleared at the last stage of umount,
> if do it before jbd2_journal_destroy(), then some metadata in uncommitted
> transaction could be lost due to io error, but as dirty flag of journal
> was already cleared, we can't find that until run a full fsck. This may
> cause system panic or other corruption.
> 
> Signed-off-by: Junxiao Bi <junxiao.bi at oracle.com>
Reviewed-by: Yiwen Jiang <jiangyiwen at huawei.com>
> Cc: Yiwen Jiang <jiangyiwen at huawei.com>
> Cc: Jun Piao <piaojun at huawei.com>
> ---
>  fs/ocfs2/journal.c | 6 ++----
>  1 file changed, 2 insertions(+), 4 deletions(-)
> 
>  V1 -> V2:
>  pointed by Yiwen, need check return value of jbd2_journal_destroy
> 
> diff --git a/fs/ocfs2/journal.c b/fs/ocfs2/journal.c
> index 13f8e097babf..b51bb873441f 100644
> --- a/fs/ocfs2/journal.c
> +++ b/fs/ocfs2/journal.c
> @@ -986,7 +986,8 @@ void ocfs2_journal_shutdown(struct ocfs2_super *osb)
>  			mlog_errno(status);
>  	}
>  
> -	if (status == 0) {
> +	/* Shutdown the kernel journal system */
> +	if (!jbd2_journal_destroy(journal->j_journal) && !status) {
>  		/*
>  		 * Do not toggle if flush was unsuccessful otherwise
>  		 * will leave dirty metadata in a "clean" journal
> @@ -995,9 +996,6 @@ void ocfs2_journal_shutdown(struct ocfs2_super *osb)
>  		if (status < 0)
>  			mlog_errno(status);
>  	}
> -
> -	/* Shutdown the kernel journal system */
> -	jbd2_journal_destroy(journal->j_journal);
>  	journal->j_journal = NULL;
>  
>  	OCFS2_I(inode)->ip_open_count--;
>

Hi Junxiao,

On 18/11/19 16:07, Junxiao Bi wrote:
> Dirty flag of the journal should be cleared at the last stage of umount,
> if do it before jbd2_journal_destroy(), then some metadata in uncommitted
> transaction could be lost due to io error, but as dirty flag of journal
> was already cleared, we can't find that until run a full fsck. This may
> cause system panic or other corruption.
> 
> Signed-off-by: Junxiao Bi <junxiao.bi at oracle.com>
> Cc: Yiwen Jiang <jiangyiwen at huawei.com>
> Cc: Jun Piao <piaojun at huawei.com>
> ---
>  fs/ocfs2/journal.c | 6 ++----
>  1 file changed, 2 insertions(+), 4 deletions(-)
> 
>  V1 -> V2:
>  pointed by Yiwen, need check return value of jbd2_journal_destroy
> 
> diff --git a/fs/ocfs2/journal.c b/fs/ocfs2/journal.c
> index 13f8e097babf..b51bb873441f 100644
> --- a/fs/ocfs2/journal.c
> +++ b/fs/ocfs2/journal.c
> @@ -986,7 +986,8 @@ void ocfs2_journal_shutdown(struct ocfs2_super *osb)
>  			mlog_errno(status);
>  	}
>  
> -	if (status == 0) {
> +	/* Shutdown the kernel journal system */
> +	if (!jbd2_journal_destroy(journal->j_journal) && !status) {
>  		/*
>  		 * Do not toggle if flush was unsuccessful otherwise
>  		 * will leave dirty metadata in a "clean" journal
> @@ -995,9 +996,6 @@ void ocfs2_journal_shutdown(struct ocfs2_super *osb)
>  		if (status < 0)
>  			mlog_errno(status);
>  	}
> -
> -	/* Shutdown the kernel journal system */
> -	jbd2_journal_destroy(journal->j_journal);
Now we will write journal inode after journal has been destroyed.
I wonder if it the right way as expected.

Thanks,
Joseph
>  	journal->j_journal = NULL;
>  
>  	OCFS2_I(inode)->ip_open_count--;
>