Merge tag 'xfs-5.13-merge-3' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux

Pull xfs updates from Darrick Wong:
 "The notable user-visible addition this cycle is ability to remove
  space from the last AG in a filesystem. This is the first of many
  changes needed for full-fledged support for shrinking a filesystem.
  Still needed are (a) the ability to reorganize files and metadata away
  from the end of the fs; (b) the ability to remove entire allocation
  groups; (c) shrink support for realtime volumes; and (d) thorough
  testing of (a-c).

  There are a number of performance improvements in this code drop: Dave
  streamlined various parts of the buffer logging code and reduced the
  cost of various debugging checks, and added the ability to pre-create
  the xattr structures while creating files. Brian eliminated
  transaction reservations that were being held across writeback (thus
  reducing livelock potential.

  Other random pieces: Pavel fixed the repetitve warnings about
  deprecated mount options, I fixed online fsck to behave itself when a
  readonly remount comes in during scrub, and refactored various other
  parts of that code, Christoph contributed a lot of refactoring this
  cycle. The xfs_icdinode structure has been absorbed into the (incore)
  xfs_inode structure, and the format and flags handling around
  xfs_inode_fork structures has been simplified. Chandan provided a
  number of fixes for extent count overflow related problems that have
  been shaken out by debugging knobs added during 5.12.

  Summary:

   - Various minor fixes in online scrub.

   - Prevent metadata files from being automatically inactivated.

   - Validate btree heights by the computed per-btree limits.

   - Don't warn about remounting with deprecated mount options.

   - Initialize attr forks at create time if we suspect we're going to
     need to store them.

   - Reduce memory reallocation workouts in the logging code.

   - Fix some theoretical math calculation errors in logged buffers that
     span multiple discontig memory ranges but contiguous ondisk
     regions.

   - Speedups in dirty buffer bitmap handling.

   - Make type verifier functions more inline-happy to reduce overhead.

   - Reduce debug overhead in directory checking code.

   - Many many typo fixes.

   - Begin to handle the permanent loss of the very end of a filesystem.

   - Fold struct xfs_icdinode into xfs_inode.

   - Deprecate the long defunct BMV_IF_NO_DMAPI_READ from the bmapx
     ioctl.

   - Remove a broken directory block format check from online scrub.

   - Fix a bug where we could produce an unnecessarily tall data fork
     btree when creating an attr fork.

   - Fix scrub and readonly remounts racing.

   - Fix a writeback ioend log deadlock problem by dropping the behavior
     where we could preallocate a setfilesize transaction.

   - Fix some bugs in the new extent count checking code.

   - Fix some bugs in the attr fork preallocation code.

   - Refactor if_flags out of the incore inode fork data structure"

* tag 'xfs-5.13-merge-3' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux: (77 commits)
  xfs: remove xfs_quiesce_attr declaration
  xfs: remove XFS_IFEXTENTS
  xfs: remove XFS_IFINLINE
  xfs: remove XFS_IFBROOT
  xfs: only look at the fork format in xfs_idestroy_fork
  xfs: simplify xfs_attr_remove_args
  xfs: rename and simplify xfs_bmap_one_block
  xfs: move the XFS_IFEXTENTS check into xfs_iread_extents
  xfs: drop unnecessary setfilesize helper
  xfs: drop unused ioend private merge and setfilesize code
  xfs: open code ioend needs workqueue helper
  xfs: drop submit side trans alloc for append ioends
  xfs: fix return of uninitialized value in variable error
  xfs: get rid of the ip parameter to xchk_setup_*
  xfs: fix scrub and remount-ro protection when running scrub
  xfs: move the check for post-EOF mappings into xfs_can_free_eofblocks
  xfs: move the xfs_can_free_eofblocks call under the IOLOCK
  xfs: precalculate default inode attribute offset
  xfs: default attr fork size does not handle device inodes
  xfs: inode fork allocation depends on XFS_IFEXTENT flag
  ...

1 files changed, 48 insertions, 17 deletions

diff --git a/fs/xfs/xfs_iops.c b/fs/xfs/xfs_iops.c
index e1f749b711f8..dfe24b7f26e5 100644
--- a/fs/xfs/xfs_iops.c
+++ b/fs/xfs/xfs_iops.c
@@ -127,6 +127,37 @@ xfs_cleanup_inode(
 	xfs_remove(XFS_I(dir), &teardown, XFS_I(inode));
 }
 
+/*
+ * Check to see if we are likely to need an extended attribute to be added to
+ * the inode we are about to allocate. This allows the attribute fork to be
+ * created during the inode allocation, reducing the number of transactions we
+ * need to do in this fast path.
+ *
+ * The security checks are optimistic, but not guaranteed. The two LSMs that
+ * require xattrs to be added here (selinux and smack) are also the only two
+ * LSMs that add a sb->s_security structure to the superblock. Hence if security
+ * is enabled and sb->s_security is set, we have a pretty good idea that we are
+ * going to be asked to add a security xattr immediately after allocating the
+ * xfs inode and instantiating the VFS inode.
+ */
+static inline bool
+xfs_create_need_xattr(
+	struct inode	*dir,
+	struct posix_acl *default_acl,
+	struct posix_acl *acl)
+{
+	if (acl)
+		return true;
+	if (default_acl)
+		return true;
+#if IS_ENABLED(CONFIG_SECURITY)
+	if (dir->i_sb->s_security)
+		return true;
+#endif
+	return false;
+}
+
+
 STATIC int
 xfs_generic_create(
 	struct user_namespace	*mnt_userns,
@@ -164,7 +195,8 @@ xfs_generic_create(
 
 	if (!tmpfile) {
 		error = xfs_create(mnt_userns, XFS_I(dir), &name, mode, rdev,
-				   &ip);
+				xfs_create_need_xattr(dir, default_acl, acl),
+				&ip);
 	} else {
 		error = xfs_create_tmpfile(mnt_userns, XFS_I(dir), mode, &ip);
 	}
@@ -488,7 +520,7 @@ xfs_vn_get_link_inline(
 	struct xfs_inode	*ip = XFS_I(inode);
 	char			*link;
 
-	ASSERT(ip->i_df.if_flags & XFS_IFINLINE);
+	ASSERT(ip->i_df.if_format == XFS_DINODE_FMT_LOCAL);
 
 	/*
 	 * The VFS crashes on a NULL pointer, so return -EFSCORRUPTED if
@@ -563,13 +595,12 @@ xfs_vn_getattr(
 	stat->atime = inode->i_atime;
 	stat->mtime = inode->i_mtime;
 	stat->ctime = inode->i_ctime;
-	stat->blocks =
-		XFS_FSB_TO_BB(mp, ip->i_d.di_nblocks + ip->i_delayed_blks);
+	stat->blocks = XFS_FSB_TO_BB(mp, ip->i_nblocks + ip->i_delayed_blks);
 
 	if (xfs_sb_version_has_v3inode(&mp->m_sb)) {
 		if (request_mask & STATX_BTIME) {
 			stat->result_mask |= STATX_BTIME;
-			stat->btime = ip->i_d.di_crtime;
+			stat->btime = ip->i_crtime;
 		}
 	}
 
@@ -577,11 +608,11 @@ xfs_vn_getattr(
 	 * Note: If you add another clause to set an attribute flag, please
 	 * update attributes_mask below.
 	 */
-	if (ip->i_d.di_flags & XFS_DIFLAG_IMMUTABLE)
+	if (ip->i_diflags & XFS_DIFLAG_IMMUTABLE)
 		stat->attributes |= STATX_ATTR_IMMUTABLE;
-	if (ip->i_d.di_flags & XFS_DIFLAG_APPEND)
+	if (ip->i_diflags & XFS_DIFLAG_APPEND)
 		stat->attributes |= STATX_ATTR_APPEND;
-	if (ip->i_d.di_flags & XFS_DIFLAG_NODUMP)
+	if (ip->i_diflags & XFS_DIFLAG_NODUMP)
 		stat->attributes |= STATX_ATTR_NODUMP;
 
 	stat->attributes_mask |= (STATX_ATTR_IMMUTABLE |
@@ -706,7 +737,7 @@ xfs_setattr_nonsize(
 		 */
 		ASSERT(udqp == NULL);
 		ASSERT(gdqp == NULL);
-		error = xfs_qm_vop_dqalloc(ip, uid, gid, ip->i_d.di_projid,
+		error = xfs_qm_vop_dqalloc(ip, uid, gid, ip->i_projid,
 					   qflags, &udqp, &gdqp, NULL);
 		if (error)
 			return error;
@@ -918,8 +949,8 @@ xfs_setattr_size(
 	 * operation.
 	 *
 	 * And we update in-core i_size and truncate page cache beyond newsize
-	 * before writeback the [di_size, newsize] range, so we're guaranteed
-	 * not to write stale data past the new EOF on truncate down.
+	 * before writeback the [i_disk_size, newsize] range, so we're
+	 * guaranteed not to write stale data past the new EOF on truncate down.
 	 */
 	truncate_setsize(inode, newsize);
 
@@ -932,9 +963,9 @@ xfs_setattr_size(
 	 * otherwise those blocks may not be zeroed after a crash.
 	 */
 	if (did_zeroing ||
-	    (newsize > ip->i_d.di_size && oldsize != ip->i_d.di_size)) {
+	    (newsize > ip->i_disk_size && oldsize != ip->i_disk_size)) {
 		error = filemap_write_and_wait_range(VFS_I(ip)->i_mapping,
-						ip->i_d.di_size, newsize - 1);
+						ip->i_disk_size, newsize - 1);
 		if (error)
 			return error;
 	}
@@ -976,7 +1007,7 @@ xfs_setattr_size(
 	 * permanent before actually freeing any blocks it doesn't matter if
 	 * they get written to.
 	 */
-	ip->i_d.di_size = newsize;
+	ip->i_disk_size = newsize;
 	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
 
 	if (newsize <= oldsize) {
@@ -1262,7 +1293,7 @@ xfs_inode_should_enable_dax(
 		return false;
 	if (ip->i_mount->m_flags & XFS_MOUNT_DAX_ALWAYS)
 		return true;
-	if (ip->i_d.di_flags2 & XFS_DIFLAG2_DAX)
+	if (ip->i_diflags2 & XFS_DIFLAG2_DAX)
 		return true;
 	return false;
 }
@@ -1319,7 +1350,7 @@ xfs_setup_inode(
 	/* make the inode look hashed for the writeback code */
 	inode_fake_hash(inode);
 
-	i_size_write(inode, ip->i_d.di_size);
+	i_size_write(inode, ip->i_disk_size);
 	xfs_diflags_to_iflags(ip, true);
 
 	if (S_ISDIR(inode->i_mode)) {
@@ -1377,7 +1408,7 @@ xfs_setup_iops(
 		inode->i_fop = &xfs_dir_file_operations;
 		break;
 	case S_IFLNK:
-		if (ip->i_df.if_flags & XFS_IFINLINE)
+		if (ip->i_df.if_format == XFS_DINODE_FMT_LOCAL)
 			inode->i_op = &xfs_inline_symlink_inode_operations;
 		else
 			inode->i_op = &xfs_symlink_inode_operations;