[Ocfs2-tools-devel] [PATCH 2/3] fsck.ocfs2: Fix the cluster count if we changed it in pass 0.

[Sunil Mushran]

Signed-off-by: Sunil Mushran <sunil.mushran at oracle.com>

Joel Becker wrote:
> At the start of pass 0, we fix the superblock's cluster count to match
> the global bitmap.  After we do that, we actually verify the global
> bitmap.  If we end up repairing the global bitmap in any way, we never
> check to see if its cluster count changed.
>
> This patch checks to see if the cluster count changed during repair.  If
> so, the user is prompted to trust the new values.  This triggers a
> reinit of the fsck state and a rescan of the global bitmap.  While the
> fixed bitmap should be fine, the rescan is necessary to pick up the
> clusters used.
>
> Signed-off-by: Joel Becker <joel.becker at oracle.com>
> ---
>  fsck.ocfs2/fsck.ocfs2.checks.8.in |    7 ++++++
>  fsck.ocfs2/pass0.c                |   42 ++++++++++++++++++++++++++++++++++++-
>  2 files changed, 48 insertions(+), 1 deletions(-)
>
> diff --git a/fsck.ocfs2/fsck.ocfs2.checks.8.in b/fsck.ocfs2/fsck.ocfs2.checks.8.in
> index b780561..5aa8023 100644
> --- a/fsck.ocfs2/fsck.ocfs2.checks.8.in
> +++ b/fsck.ocfs2/fsck.ocfs2.checks.8.in
> @@ -118,6 +118,13 @@ bitmap. This is only possible due to a failed volume resize operation.
>  Answering yes updates the total clusters in the super block to the value
>  specified in the global bitmap.
>  
> +.SS "FIXED_CHAIN_CLUSTERS"
> +The global bitmap inode was repaired, resulting in a change to the total
> +cluster count of the filesystem.
> +
> +Answering yes updates the total clusters in the super block to the value
> +specified in the global bitmap.
> +
>  \" pass0.c
>  
>  .SS "GROUP_UNEXPECTED_DESC"
> diff --git a/fsck.ocfs2/pass0.c b/fsck.ocfs2/pass0.c
> index 66e18c7..a32bd18 100644
> --- a/fsck.ocfs2/pass0.c
> +++ b/fsck.ocfs2/pass0.c
> @@ -1060,13 +1060,14 @@ errcode_t o2fsck_pass0(o2fsck_state *ost)
>  {
>  	errcode_t ret;
>  	uint64_t blkno;
> +	uint32_t pre_repair_clusters;
>  	char *blocks = NULL;
>  	char *pre_cache_buf = NULL;
>  	struct ocfs2_dinode *di = NULL;
>  	ocfs2_filesys *fs = ost->ost_fs;
>  	ocfs2_cached_inode **ci;
>  	int max_slots = OCFS2_RAW_SB(fs->fs_super)->s_max_slots;
> -	int i, type;
> +	int i, type, bitmap_retried = 0;
>  
>  	printf("Pass 0a: Checking cluster allocation chains\n");
>  
> @@ -1159,12 +1160,51 @@ errcode_t o2fsck_pass0(o2fsck_state *ost)
>  		}
>  	}
>  
> +retry_bitmap:
> +	pre_repair_clusters = di->i_clusters;
>  	ret = verify_bitmap_descs(ost, di, blocks + ost->ost_fs->fs_blocksize,
>  				  blocks + (ost->ost_fs->fs_blocksize * 2));
>  
>  	if (ret)
>  		goto out;
>  
> +	if (pre_repair_clusters != di->i_clusters) {
> +		if (prompt(ost, PY, PR_FIXED_CHAIN_CLUSTERS,
> +			   "Repair of global_bitmap changed the filesystem "
> +			   "from %u clusters to %u clusters.  Trust "
> +			   "global_bitmap?",
> +			   pre_repair_clusters, di->i_clusters)) {
> +			ost->ost_num_clusters = di->i_clusters;
> +			fs->fs_clusters = di->i_clusters;
> +			fs->fs_blocks = ocfs2_clusters_to_blocks(fs,
> +							 fs->fs_clusters);
> +			ret = o2fsck_state_reinit(fs, ost);
> +			if (ret) {
> +				com_err(whoami, ret, "while reinit "
> +					"o2fsck_state.");
> +				goto out;
> +			}
> +
> +			/*
> +			 * The reinit clears the bits found during the
> +			 * scan of the global bitmap.  We need to go over
> +			 * them again.  They really should come out clean
> +			 * this time.  If they don't, we probably have
> +			 * a serious problem.
> +			 *
> +			 * In an interactive run, the user can keep
> +			 * retrying and abort when they give up.  In a
> +			 * non-interactive mode, we can't loop forever.
> +			 */
> +			if (ost->ost_ask || !bitmap_retried) {
> +				bitmap_retried = 1;
> +				verbosef("Restarting global_bitmap %s\n",
> +					 "scan");
> +				goto retry_bitmap;
> +			}
> +		}
> +	}
> +
>  	printf("Pass 0b: Checking inode allocation chains\n");
>  
>  	/* first the global inode alloc and then each of the node's
>