[Ocfs2-devel] [DRAFT] ocfs2: commit the transtion before goto out to avoid blocked.

[Darrick J. Wong]

On Sat, Apr 01, 2017 at 11:42:36AM +0000, Guozhonghua wrote:
> Hi, 
> 
> For the commit: 2b0ad0085aa47ace4756aa501274a7de0325c09c 
> 
> Mar 31 14:42:59 wy-ost209 kernel: [795291.811547] OCFS2: ERROR (device dm-11): ocfs2_block_group_clear_bits: Group descriptor # 33030144 has bit count 32256 but claims 32294 are freed. num_bits 48
> Mar 31 14:42:59 wy-ost209 kernel: [795291.811574] File system is now read-only due to the potential of on-disk corruption. Please run fsck.ocfs2 once the file system is unmounted.
> Mar 31 14:42:59 wy-ost209 kernel: [795291.811591] (libvirtd,1223183,4):_ocfs2_free_suballoc_bits:2502 ERROR: status = -30
> Mar 31 14:42:59 wy-ost209 kernel: [795291.811605] (libvirtd,1223183,4):_ocfs2_free_suballoc_bits:2525 ERROR: status = -30
> Mar 31 14:42:59 wy-ost209 kernel: [795291.811619] (libvirtd,1223183,4):_ocfs2_free_clusters:2588 ERROR: status = -30
> Mar 31 14:42:59 wy-ost209 kernel: [795291.811632] (libvirtd,1223183,4):_ocfs2_free_clusters:2597 ERROR: status = -30
> Mar 31 14:42:59 wy-ost209 kernel: [795291.811646] (libvirtd,1223183,4):ocfs2_replay_truncate_records:5996 ERROR: status = -30
> Mar 31 14:42:59 wy-ost209 kernel: [795291.811660] (libvirtd,1223183,4):__ocfs2_flush_truncate_log:6062 ERROR: status = -30
> Mar 31 14:42:59 wy-ost209 kernel: [795291.811675] (libvirtd,1223183,4):ocfs2_remove_btree_range:5769 ERROR: status = -30
> Mar 31 14:42:59 wy-ost209 kernel: [795291.811689] (libvirtd,1223183,4):ocfs2_commit_truncate:7194 ERROR: status = -30
> Mar 31 14:42:59 wy-ost209 kernel: [795291.811722] (libvirtd,1223183,4):ocfs2_truncate_for_delete:607 ERROR: status = -30
> Mar 31 14:42:59 wy-ost209 kernel: [795291.811736] (libvirtd,1223183,4):ocfs2_wipe_inode:776 ERROR: status = -30
> Mar 31 14:42:59 wy-ost209 kernel: [795291.811746] (libvirtd,1223183,4):ocfs2_delete_inode:1066 ERROR: status = -30
> 
> 
> ocfs2_replay_truncate_records
>     ocfs2_start_trans
> 	   down_read(&osb->journal->j_trans_barrier)
> 	
> 	Something is wrong, so it goes out. 
> 	The ocfs2_commit_trans is not called, so up_read is not called, as to cause the node blocked.
> 	
> 	ocfs2_commit_trans(osb, handle)
> 		up_read(&journal->j_trans_barrier)
> 		
> 
> So we have a patch to avoid it.
> 
> 
> >From 9ae1f5550c4743356f68685276ffcddecf698e9d Mon Sep 17 00:00:00 2001
> From: guozhonghua <guozhonghua at h3c.com>
> Date: Sat, 1 Apr 2017 19:30:54 +0800
> Subject: [PATCH] If there is error in the function
>  ocfs2_replay_truncate_records, the function
>  ocfs2_commit_trans should be called before return to avoid
>  node blocked
> 
> 
> Signed-off-by: guozhonghua <guozhonghua at h3c.com>
> ---
>  fs/ocfs2/alloc.c |    2 ++
>  1 file changed, 2 insertions(+)
> 
> diff --git a/fs/ocfs2/alloc.c b/fs/ocfs2/alloc.c
> index fb15a96..dc52f7a 100644
> --- a/fs/ocfs2/alloc.c
> +++ b/fs/ocfs2/alloc.c
> @@ -5953,6 +5953,7 @@ static int ocfs2_replay_truncate_records(struct ocfs2_super *osb,
>  						 OCFS2_JOURNAL_ACCESS_WRITE);
>  		if (status < 0) {
>  			mlog_errno(status);
> +			ocfs2_commit_trans(osb, handle);

Um... if there's an error, shouldn't we /abort/ the transaction?

--D

>  			goto bail;
>  		}
>  
> @@ -5977,6 +5978,7 @@ static int ocfs2_replay_truncate_records(struct ocfs2_super *osb,
>  						     num_clusters);
>  			if (status < 0) {
>  				mlog_errno(status);
> +				ocfs2_commit_trans(osb, handle);
>  				goto bail;
>  			}
>  		}
> -- 
> 1.7.9.5
> 
> _______________________________________________
> Ocfs2-devel mailing list
> Ocfs2-devel at oss.oracle.com
> https://oss.oracle.com/mailman/listinfo/ocfs2-devel