fsck.gfs2: Support for checking gfs1 file systems

Hi,

I previously posted a series of 44 patches to fsck.gfs2 that fix numerous bugs and allow fsck.gfs2 to operate on GFS1 file systems. Since I posted those patches, I received feedback and changed the patches accordingly. I've also done extensive testing and discovered and fixed several more bugs. There are now 56 patches in all. As Andy Price suggested, I renamed their descriptions to make them shorter so they're not cut off.

My latest version has passed all my tests and I'm ready to start pushing them to the upstream gfs2-utils git repository. But before I do, I wanted to post the corrected patches and the new ones here. I'm not reposting the patches that got ACKed. Here is the complete list of patches:

1/56 fsck.gfs2: Make functions use sdp rather than sbp 2/56 fsck.gfs2: Change "if(" to "if ("
3/56 libgfs1: Add gfs1 variable to superblock structure 4/56 libgfs2: Make check_sb and read_sb operate on gfs1 5/56 libgfs2: move gfs1 structures to libgfs2
6/56 fsck.gfs2: Check for blocks wrongly inside resource groups 7/56 fsck.gfs2: Rename check_leaf to check_ealeaf_block 8/56 fsck.gfs2: fsck.gfs2: Delete vestigial buffer_head in check_leaf 9/56 fsck.gfs2: fsck.gfs2: Rename nlink functions to be intuitive
10/56 fsck.gfs2: fsck.gfs2: Sync di_nlink adding links for lost+found

11/56 fsck.gfs2: fsck.gfs2: Make dir entry count 32 bits 12/56 fsck.gfs2: get rid of triple negative logic
13/56 dirent_repair needs to mark the buffer as modified 14/56 fsck.gfs2: fsck.gfs2: Ask to reclaim unlinked meta per-rgrp only 15/56 fsck.gfs2: fsck.gfs2: Refactor add_dotdot function in lost+found 16/56 libgfs2: libgfs2: Use __FUNCTION__ rather than __FILE__ 17/56 fsck.gfs2: fsck.gfs2: Don't stop invalidating blocks on invalid 18/56 fsck.gfs2: fsck.gfs2: Find and clear duplicate leaf blocks refs 19/56 fsck.gfs2: fsck.gfs2: Move check_num_ptrs from metawalk to pass1 20/56 fsck.gfs2: fsck.gfs2: Duplicate ref processing for leaf blocks

21/56 fsck.gfs2: fsck.gfs2: split check_leaf_blks to be more readable 22/56 fsck.gfs2: Shorten output
23/56 fsck.gfs2: Make output messages more sensible
24/56 fsck.gfs pass2: Refactor function set_dotdor_dir 25/56 fsck.gfs2 pass2: Delete extended attributes with inode 26/56 fsck.gfs2 pass2: Don't delete invalid inode metadata 27/56 fsck.gfs2 pass3: Refactor mark_and_return_parent 28/56 fsck.gfs2: misc cosmetic changes
29/56 fsck.gfs2: Don't use old_leaf if it was a duplicate 30/56 fsck.gfs2: Add find_remove_dup, free_block_if_notdup

31/56 fsck.gfs2: don't free prev rgrp list repairing rgrps 32/56 libgfs2: eliminate gfs1_readi in favor of gfs2_readi 33/56 libgfs2: Mark buffer modified adding a new GFS1 block 34/56 libgfs2: Use dinode buffer to map gfs1 dinode blocks 35/56 libgfs2: move block_map functions to fsck.gfs2 36/56 libgfs2: eliminate gfs1_rindex_read
37/56 libgfs2: combine ri_update and gfs1_ri_update 38/56 libgfs2: combine gfs_inode_read and gfs_inode_get 39/56 libgfs2: move gfs1 functions from edit to libgfs2 40/56 gfs2_edit savemeta: save_inode_data backward for gfs1

41/56 libgfs2: expand capabilities to operate on gfs1 42/56 fsck.gfs2: Combine block and char device inode types 43/56 fsck.gfs2: four-step duplicate elimination process 44/56 fsck.gfs2: Add ability to check gfs1 file systems 45/56 fsck.gfs2: Remove bad inodes from duplicate tree 46/56 fsck.gfs2: Handle duplicate reference to dinode blocks 47/56 fsck.gfs2: Bad extended attributes not deleted properly 48/56 libgfs2: Make rebuild functions not re-read ip 49/56 fsck.gfs2: Shorten debug output
50/56 fsck.gfs2: Increment link count reporting wrong dinode

51/56 fsck.gfs2: system dinodes take priority over user dinodes 52/56 fsck.gfs2: Recognize partially gfs2-converted dinodes 53/56 fsck.gfs2: Print step 2 duplicate debug msg first 54/56 fsck.gfs2: pass1c counts percentage backward 55/56 fsck.gfs2: Speed up rangecheck functions
56/56 libgfs2: Make in-core rgrps use rbtree

In addition, I just put my experimental new fsck.gfs2, which contains all these patches, on my people page at this location:

http://people.redhat.com/rpeterso/Experimental/RHEL6.x/fsck.gfs2

It is experimental, so make backups and try it AT YOUR OWN RISK. However, it is also better tested than all of its predecessors.

Regards,

Bob Peterson
Red Hat File Systems

Hi Bob,

Could you send the latest version of patch 32? Using the original one I'm getting:

Applying: libgfs2: eliminate gfs1_readi in favor of gfs2_readi
error: patch failed: gfs2/edit/extended.c:516
error: gfs2/edit/extended.c: patch does not apply
error: patch failed: gfs2/libgfs2/super.c:169
error: gfs2/libgfs2/super.c: patch does not apply

Cheers,

Andy

----- Original Message -----

| Hi Bob,
| 
| On 25/08/11 17:49, Bob Peterson wrote:
| > 32/56 libgfs2: eliminate gfs1_readi in favor of gfs2_readi
| 
| Could you send the latest version of patch 32? Using the original one
| I'm getting:
| 
| Applying: libgfs2: eliminate gfs1_readi in favor of gfs2_readi
| error: patch failed: gfs2/edit/extended.c:516
| error: gfs2/edit/extended.c: patch does not apply
| error: patch failed: gfs2/libgfs2/super.c:169
| error: gfs2/libgfs2/super.c: patch does not apply
| 
| Cheers,
| 
| Andy

Hi Andy,

Here it is:

Bob
--
From 03f0ed42c95b10f02bfe52ae08cb7c18402ce662 Mon Sep 17 00:00:00 2001
From: Bob Peterson <[email protected]>
Date: Wed, 10 Aug 2011 10:21:00 -0500
Subject: [PATCH 32/56] libgfs2: eliminate gfs1_readi in favor of gfs2_readi

This patch eliminates function gfs1_readi because it's nearly identical to gfs2_readi. The gfs1-specific bits have been made generic in gfs2_readi based on the sdp->gfs1 setting.

rhbz#675723

---
 gfs2/edit/extended.c   |    8 +----
 gfs2/libgfs2/fs_ops.c  |   21 ++++++++++----
 gfs2/libgfs2/gfs1.c    |   68 +-----------------------------------------------
 gfs2/libgfs2/libgfs2.h |    2 -
 gfs2/libgfs2/super.c   |    5 ---
 5 files changed, 18 insertions(+), 86 deletions(-)

index 792b724..071f589 100644
@@ -516,12 +516,8 @@ static int parse_rindex(struct gfs2_inode *dip, int print_rindex)
 
 		roff = print_entry_ndx * sizeof(struct gfs2_rindex);
 
-		if (sbd.gfs1)
-			error = gfs1_readi(dip, (void *)&rbuf, roff,
-					   sizeof(struct gfs2_rindex));
-		else
-			error = gfs2_readi(dip, (void *)&rbuf, roff,
-					   sizeof(struct gfs2_rindex));
+		error = gfs2_readi(dip, (void *)&rbuf, roff,
+				   sizeof(struct gfs2_rindex));
 		if (!error) /* end of file */
 			break;
 		gfs2_rindex_in(&ri, rbuf);

index 60ee467..4c8d31e 100644
@@ -510,6 +510,7 @@ int gfs2_readi(struct gfs2_inode *ip, void *buf,
 	unsigned int amount;
 	int not_new = 0;
 	int isdir = !!(S_ISDIR(ip->i_di.di_mode));
+	int journaled = ip->i_di.di_flags & GFS2_DIF_JDATA;
 	int copied = 0;
 
 	if (offset >= ip->i_di.di_size)

@@ -521,7 +522,7 @@ int gfs2_readi(struct gfs2_inode *ip, void *buf,
 	if (!size)
 		return 0;
 
-	if (isdir) {
+	if ((sdp->gfs1 && journaled) || (!sdp->gfs1 && isdir)) {
 		lblock = offset;
 		o = lblock % sdp->sd_jbsize;
 		lblock /= sdp->sd_jbsize;

@@ -532,7 +533,7 @@ int gfs2_readi(struct gfs2_inode *ip, void *buf,
 
 	if (inode_is_stuffed(ip))
 		o += sizeof(struct gfs2_dinode);
-	else if (isdir)
+	else if ((sdp->gfs1 && journaled) || (!sdp->gfs1 && isdir))
 		o += sizeof(struct gfs2_meta_header);
 
 	while (copied < size) {

@@ -540,9 +541,14 @@ int gfs2_readi(struct gfs2_inode *ip, void *buf,
 		if (amount > sdp->bsize - o)
 			amount = sdp->bsize - o;
 
-		if (!extlen)
-			block_map(ip, lblock, &not_new, &dblock, &extlen,
-				  FALSE);
+		if (!extlen) {
+			if (sdp->gfs1)
+				gfs1_block_map(ip, lblock, &not_new, &dblock,
+					       &extlen, FALSE);
+			else
+				block_map(ip, lblock, &not_new, &dblock,
+					  &extlen, FALSE);
+		}
 
 		if (dblock) {
 			if (dblock == ip->i_di.di_num.no_addr)

@@ -561,7 +567,10 @@ int gfs2_readi(struct gfs2_inode *ip, void *buf,
 		copied += amount;
 		lblock++;
 
-		o = (isdir) ? sizeof(struct gfs2_meta_header) : 0;
+		if (sdp->gfs1)
+			o = (journaled) ? sizeof(struct gfs2_meta_header) : 0;
+		else
+			o = (isdir) ? sizeof(struct gfs2_meta_header) : 0;
 	}
 
 	return copied;

index f8f55b1..0ac5f3d 100644
@@ -669,8 +669,6 @@ extern void gfs1_lookup_block(struct gfs2_inode *ip,
 			      int create, int *new, uint64_t *block);
 extern void gfs1_block_map(struct gfs2_inode *ip, uint64_t lblock, int *new,
 			   uint64_t *dblock, uint32_t *extlen, int prealloc);
-extern int gfs1_readi(struct gfs2_inode *ip, void *buf, uint64_t offset,
-		      unsigned int size);
 extern int gfs1_rindex_read(struct gfs2_sbd *sdp, int fd, int *count1);
 extern int gfs1_ri_update(struct gfs2_sbd *sdp, int fd, int *rgcount, int quiet);
 extern struct gfs2_inode *gfs_inode_get(struct gfs2_sbd *sdp,

index 9e8d079..693e547 100644
@@ -160,11 +160,6 @@ int rindex_read(struct gfs2_sbd *sdp, int fd, int *count1, int *sane)
 	for (rg = 0; ; rg++) {
 		if (fd > 0)
 			error = read(fd, &buf, sizeof(struct gfs2_rindex));
-		else if (sdp->gfs1)
-			error = gfs1_readi(sdp->md.riinode,
-					   (char *)&buf.bufgfs1,
-					   rg * sizeof(struct gfs2_rindex),
-					   sizeof(struct gfs2_rindex));
 		else
 			error = gfs2_readi(sdp->md.riinode,
 					   (char *)&buf.bufgfs2,

1.7.4.4