[Cluster-devel] [PATCH] dlm: send_bast_queue() skip list loop not only sending basts to convertqueue

[cmaiolino@redhat.com <cmaiolino@redhat.com>]

From: Carlos Maiolino <cmaiolino@redhat.com>

with the current check condition: if (gr == lkb), the list will skip not only basts sent to convertqueue
since gr == lkb could be true on another situations, where such can cause a gfs2 corruption.

Corruption checked on gfs2 resource groups, adding a file on a node and removing the file on another node.
The resource groups got corrupted without this patch:

This is an output of the gfs2_edit without this patch:

Environment:
	-a two node cluster
	-gfs2 filesystem with 1G (counting journal space)
	- /dev/sda5             1.0G  259M  766M  26% /mnt (new fs after mkfs.gfs2)

- A new filesystem shared by two nodes:

RG #1 located at: 65551 (0x1000f)
  mh_magic              0x01161970(hex)
  mh_type               2                   0x2
  mh_format             200                 0xc8
  rg_flags              0                   0x0
  rg_free               64858               0xfd5a
  rg_dinodes            11                  0xb

- After fill the filesystem with just one file (a 764M file using dd) from node 1

RG #1 located at: 65551 (0x1000f)
  mh_magic              0x01161970(hex)
  mh_type               2                   0x2
  mh_format             200                 0xc8
  rg_flags              0                   0x0
  rg_free               18                  0x12
  rg_dinodes            12                  0xc

- After remove the file from the filesystem from node 2

RG #1 located at: 65551 (0x1000f)
  mh_magic              0x01161970(hex)
  mh_type               2                   0x2
  mh_format             200                 0xc8
  rg_flags              0                   0x0
  rg_free               18                  0x12
  rg_dinodes            12                  0xc

- The space of the filesystem is not freed:

[root at node2-vm stats]# df -h /mnt
Filesystem            Size  Used Avail Use% Mounted on
/dev/sda              1.0G  1.0G  216K 100% /mnt

- The problem also persists after umount the filesystem

- Applying the patch, the problem is not reproducible anymore and the resource group blocks are properly freed

Signed-off-by: Carlos Maiolino <cmaiolino@redhat.com>
---
 fs/dlm/lock.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/fs/dlm/lock.c b/fs/dlm/lock.c
index 64e5f3e..565c519 100644
--- a/fs/dlm/lock.c
+++ b/fs/dlm/lock.c
@@ -1847,7 +1847,7 @@ static void send_bast_queue(struct dlm_rsb *r, struct list_head *head,
 
 	list_for_each_entry(gr, head, lkb_statequeue) {
 		/* skip self when sending basts to convertqueue */
-		if (gr == lkb)
+		if (head == &r->res_grantqueue && gr == lkb)
 			continue;
 		if (gr->lkb_bastfn && modes_require_bast(gr, lkb)) {
 			queue_bast(r, gr, lkb->lkb_rqmode);
-- 
1.7.1