xfs: ensure EFD trans aborts on log recovery extent free failure
authorBrian Foster <bfoster@redhat.com>
Tue, 18 Aug 2015 23:51:43 +0000 (09:51 +1000)
committerDave Chinner <david@fromorbit.com>
Tue, 18 Aug 2015 23:51:43 +0000 (09:51 +1000)
Log recovery attempts to free extents with leftover EFIs in the AIL
after initial processing. If the extent free fails (e.g., due to
unrelated fs corruption), the transaction is cancelled, though it
might not be dirtied at the time. If this is the case, the EFD does
not abort and thus does not release the EFI. This can lead to hangs
as the EFI pins the AIL.

Update xlog_recover_process_efi() to log the EFD in the transaction
before xfs_free_extent() errors are handled to ensure the
transaction is dirty, aborts the EFD and releases the EFI on error.
Since this is a requirement for EFD processing (and consistent with
xfs_bmap_finish()), update the EFD logging helper to do the extent
free and unconditionally log the EFD. This encodes the required EFD
logging behavior into the helper and reduces the likelihood of
errors down the road.

[dchinner: re-add xfs_alloc.h to xfs_log_recover.c to fix build
 failure.]

Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
fs/xfs/xfs_bmap_util.c
fs/xfs/xfs_log_recover.c
fs/xfs/xfs_trans.h
fs/xfs/xfs_trans_extfree.c

index fa65f67..73aab0d 100644 (file)
@@ -112,24 +112,17 @@ xfs_bmap_finish(
                return error;
        }
 
+       /*
+        * Get an EFD and free each extent in the list, logging to the EFD in
+        * the process. The remaining bmap free list is cleaned up by the caller
+        * on error.
+        */
        efd = xfs_trans_get_efd(*tp, efi, flist->xbf_count);
        for (free = flist->xbf_first; free != NULL; free = next) {
                next = free->xbfi_next;
 
-               /*
-                * Free the extent and log the EFD to dirty the transaction
-                * before handling errors. This ensures that the transaction is
-                * aborted, which:
-                *
-                * 1.) releases the EFI and frees the EFD
-                * 2.) shuts down the filesystem
-                *
-                * The bmap free list is cleaned up at a higher level.
-                */
-               error = xfs_free_extent(*tp, free->xbfi_startblock,
-                                       free->xbfi_blockcount);
-               xfs_trans_log_efd_extent(*tp, efd, free->xbfi_startblock,
-                                        free->xbfi_blockcount);
+               error = xfs_trans_free_extent(*tp, efd, free->xbfi_startblock,
+                                             free->xbfi_blockcount);
                if (error)
                        return error;
 
index 578bc2d..434ba2c 100644 (file)
@@ -3752,11 +3752,11 @@ xlog_recover_process_efi(
 
        for (i = 0; i < efip->efi_format.efi_nextents; i++) {
                extp = &(efip->efi_format.efi_extents[i]);
-               error = xfs_free_extent(tp, extp->ext_start, extp->ext_len);
+               error = xfs_trans_free_extent(tp, efdp, extp->ext_start,
+                                             extp->ext_len);
                if (error)
                        goto abort_error;
-               xfs_trans_log_efd_extent(tp, efdp, extp->ext_start,
-                                        extp->ext_len);
+
        }
 
        set_bit(XFS_EFI_RECOVERED, &efip->efi_flags);
index ba1660b..4643070 100644 (file)
@@ -220,10 +220,9 @@ void               xfs_trans_log_efi_extent(xfs_trans_t *,
 struct xfs_efd_log_item        *xfs_trans_get_efd(xfs_trans_t *,
                                  struct xfs_efi_log_item *,
                                  uint);
-void           xfs_trans_log_efd_extent(xfs_trans_t *,
-                                        struct xfs_efd_log_item *,
-                                        xfs_fsblock_t,
-                                        xfs_extlen_t);
+int            xfs_trans_free_extent(struct xfs_trans *,
+                                     struct xfs_efd_log_item *, xfs_fsblock_t,
+                                     xfs_extlen_t);
 int            xfs_trans_commit(struct xfs_trans *);
 int            __xfs_trans_roll(struct xfs_trans **, struct xfs_inode *, int *);
 int            xfs_trans_roll(struct xfs_trans **, struct xfs_inode *);
index 284397d..a96ae54 100644 (file)
@@ -25,6 +25,7 @@
 #include "xfs_trans.h"
 #include "xfs_trans_priv.h"
 #include "xfs_extfree_item.h"
+#include "xfs_alloc.h"
 
 /*
  * This routine is called to allocate an "extent free intention"
@@ -108,19 +109,30 @@ xfs_trans_get_efd(xfs_trans_t             *tp,
 }
 
 /*
- * This routine is called to indicate that the described
- * extent is to be logged as having been freed.  It should
- * be called once for each extent freed.
+ * Free an extent and log it to the EFD. Note that the transaction is marked
+ * dirty regardless of whether the extent free succeeds or fails to support the
+ * EFI/EFD lifecycle rules.
  */
-void
-xfs_trans_log_efd_extent(xfs_trans_t           *tp,
-                        xfs_efd_log_item_t     *efdp,
-                        xfs_fsblock_t          start_block,
-                        xfs_extlen_t           ext_len)
+int
+xfs_trans_free_extent(
+       struct xfs_trans        *tp,
+       struct xfs_efd_log_item *efdp,
+       xfs_fsblock_t           start_block,
+       xfs_extlen_t            ext_len)
 {
        uint                    next_extent;
-       xfs_extent_t            *extp;
+       struct xfs_extent       *extp;
+       int                     error;
 
+       error = xfs_free_extent(tp, start_block, ext_len);
+
+       /*
+        * Mark the transaction dirty, even on error. This ensures the
+        * transaction is aborted, which:
+        *
+        * 1.) releases the EFI and frees the EFD
+        * 2.) shuts down the filesystem
+        */
        tp->t_flags |= XFS_TRANS_DIRTY;
        efdp->efd_item.li_desc->lid_flags |= XFS_LID_DIRTY;
 
@@ -130,4 +142,6 @@ xfs_trans_log_efd_extent(xfs_trans_t                *tp,
        extp->ext_start = start_block;
        extp->ext_len = ext_len;
        efdp->efd_next_extent++;
+
+       return error;
 }