Merge tag 'xfs-6.11-merge-3' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux

Pull xfs updates from Chandan Babu:
 "Major changes in this release are limited to enabling FITRIM on
  realtime devices and Byte-based grant head log reservation tracking.

  The remaining changes are limited to fixes and cleanups included in
  this pull request.

  Core:

   - Enable FITRIM on the realtime device

   - Introduce byte-based grant head log reservation tracking instead of
     physical log location tracking.

     This allows grant head to track a full 64 bit bytes space and hence
     overcome the limit of 4GB indexing that has been present until now

  Fixes:

   - xfs_flush_unmap_range() and xfs_prepare_shift() should consider RT
     extents in the flush unmap range

   - Implement bounds check when traversing log operations during log
     replay

   - Prevent out of bounds access when traversing a directory data block

   - Prevent incorrect ENOSPC when concurrently performing file creation
     and file writes

   - Fix rtalloc rotoring when delalloc is in use

  Cleanups:

   - Clean up I/O path inode locking helpers and the page fault handler

   - xfs: hoist inode operations to libxfs in anticipation of the
     metadata inode directory feature, which maintains a directory tree
     of metadata inodes. This will be necessary for further enhancements
     to the realtime feature, subvolume support

   - Clean up some warts in the extent freeing log intent code

   - Clean up the refcount and rmap intent code before adding support
     for realtime devices

   - Provide the correct email address for sysfs ABI documentation"

* tag 'xfs-6.11-merge-3' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux: (80 commits)
  xfs: fix rtalloc rotoring when delalloc is in use
  xfs: get rid of xfs_ag_resv_rmapbt_alloc
  xfs: skip flushing log items during push
  xfs: grant heads track byte counts, not LSNs
  xfs: pass the full grant head to accounting functions
  xfs: track log space pinned by the AIL
  xfs: collapse xlog_state_set_callback in caller
  xfs: l_last_sync_lsn is really AIL state
  xfs: ensure log tail is always up to date
  xfs: background AIL push should target physical space
  xfs: AIL doesn't need manual pushing
  xfs: move and rename xfs_trans_committed_bulk
  xfs: fix the contact address for the sysfs ABI documentation
  xfs: Avoid races with cnt_btree lastrec updates
  xfs: move xfs_refcount_update_defer_add to xfs_refcount_item.c
  xfs: simplify usage of the rcur local variable in xfs_refcount_finish_one
  xfs: don't bother calling xfs_refcount_finish_one_cleanup in xfs_refcount_finish_one
  xfs: reuse xfs_refcount_update_cancel_item
  xfs: add a ci_entry helper
  xfs: remove xfs_trans_set_refcount_flags
  ...

1 files changed, 76 insertions, 65 deletions

diff --git a/fs/xfs/xfs_file.c b/fs/xfs/xfs_file.c
index b240ea5241dc..4cdc54dc9686 100644
--- a/fs/xfs/xfs_file.c
+++ b/fs/xfs/xfs_file.c
@@ -213,29 +213,18 @@ xfs_ilock_iocb_for_write(
 	if (ret)
 		return ret;
 
-	if (*lock_mode == XFS_IOLOCK_EXCL)
-		return 0;
-	if (!xfs_iflags_test(ip, XFS_IREMAPPING))
-		return 0;
-
-	xfs_iunlock(ip, *lock_mode);
-	*lock_mode = XFS_IOLOCK_EXCL;
-	return xfs_ilock_iocb(iocb, *lock_mode);
-}
-
-static unsigned int
-xfs_ilock_for_write_fault(
-	struct xfs_inode	*ip)
-{
-	/* get a shared lock if no remapping in progress */
-	xfs_ilock(ip, XFS_MMAPLOCK_SHARED);
-	if (!xfs_iflags_test(ip, XFS_IREMAPPING))
-		return XFS_MMAPLOCK_SHARED;
+	/*
+	 * If a reflink remap is in progress we always need to take the iolock
+	 * exclusively to wait for it to finish.
+	 */
+	if (*lock_mode == XFS_IOLOCK_SHARED &&
+	    xfs_iflags_test(ip, XFS_IREMAPPING)) {
+		xfs_iunlock(ip, *lock_mode);
+		*lock_mode = XFS_IOLOCK_EXCL;
+		return xfs_ilock_iocb(iocb, *lock_mode);
+	}
 
-	/* wait for remapping to complete */
-	xfs_iunlock(ip, XFS_MMAPLOCK_SHARED);
-	xfs_ilock(ip, XFS_MMAPLOCK_EXCL);
-	return XFS_MMAPLOCK_EXCL;
+	return 0;
 }
 
 STATIC ssize_t
@@ -1247,31 +1236,77 @@ xfs_file_llseek(
 	return vfs_setpos(file, offset, inode->i_sb->s_maxbytes);
 }
 
-#ifdef CONFIG_FS_DAX
 static inline vm_fault_t
-xfs_dax_fault(
+xfs_dax_fault_locked(
 	struct vm_fault		*vmf,
 	unsigned int		order,
-	bool			write_fault,
-	pfn_t			*pfn)
+	bool			write_fault)
 {
-	return dax_iomap_fault(vmf, order, pfn, NULL,
+	vm_fault_t		ret;
+	pfn_t			pfn;
+
+	if (!IS_ENABLED(CONFIG_FS_DAX)) {
+		ASSERT(0);
+		return VM_FAULT_SIGBUS;
+	}
+	ret = dax_iomap_fault(vmf, order, &pfn, NULL,
 			(write_fault && !vmf->cow_page) ?
 				&xfs_dax_write_iomap_ops :
 				&xfs_read_iomap_ops);
+	if (ret & VM_FAULT_NEEDDSYNC)
+		ret = dax_finish_sync_fault(vmf, order, pfn);
+	return ret;
 }
-#else
-static inline vm_fault_t
-xfs_dax_fault(
+
+static vm_fault_t
+xfs_dax_read_fault(
 	struct vm_fault		*vmf,
-	unsigned int		order,
-	bool			write_fault,
-	pfn_t			*pfn)
+	unsigned int		order)
 {
-	ASSERT(0);
-	return VM_FAULT_SIGBUS;
+	struct xfs_inode	*ip = XFS_I(file_inode(vmf->vma->vm_file));
+	vm_fault_t		ret;
+
+	xfs_ilock(ip, XFS_MMAPLOCK_SHARED);
+	ret = xfs_dax_fault_locked(vmf, order, false);
+	xfs_iunlock(ip, XFS_MMAPLOCK_SHARED);
+
+	return ret;
+}
+
+static vm_fault_t
+xfs_write_fault(
+	struct vm_fault		*vmf,
+	unsigned int		order)
+{
+	struct inode		*inode = file_inode(vmf->vma->vm_file);
+	struct xfs_inode	*ip = XFS_I(inode);
+	unsigned int		lock_mode = XFS_MMAPLOCK_SHARED;
+	vm_fault_t		ret;
+
+	sb_start_pagefault(inode->i_sb);
+	file_update_time(vmf->vma->vm_file);
+
+	/*
+	 * Normally we only need the shared mmaplock, but if a reflink remap is
+	 * in progress we take the exclusive lock to wait for the remap to
+	 * finish before taking a write fault.
+	 */
+	xfs_ilock(ip, XFS_MMAPLOCK_SHARED);
+	if (xfs_iflags_test(ip, XFS_IREMAPPING)) {
+		xfs_iunlock(ip, XFS_MMAPLOCK_SHARED);
+		xfs_ilock(ip, XFS_MMAPLOCK_EXCL);
+		lock_mode = XFS_MMAPLOCK_EXCL;
+	}
+
+	if (IS_DAX(inode))
+		ret = xfs_dax_fault_locked(vmf, order, true);
+	else
+		ret = iomap_page_mkwrite(vmf, &xfs_page_mkwrite_iomap_ops);
+	xfs_iunlock(ip, lock_mode);
+
+	sb_end_pagefault(inode->i_sb);
+	return ret;
 }
-#endif
 
 /*
  * Locking for serialisation of IO during page faults. This results in a lock
@@ -1290,38 +1325,14 @@ __xfs_filemap_fault(
 	bool			write_fault)
 {
 	struct inode		*inode = file_inode(vmf->vma->vm_file);
-	struct xfs_inode	*ip = XFS_I(inode);
-	vm_fault_t		ret;
-	unsigned int		lock_mode = 0;
-
-	trace_xfs_filemap_fault(ip, order, write_fault);
-
-	if (write_fault) {
-		sb_start_pagefault(inode->i_sb);
-		file_update_time(vmf->vma->vm_file);
-	}
-
-	if (IS_DAX(inode) || write_fault)
-		lock_mode = xfs_ilock_for_write_fault(XFS_I(inode));
 
-	if (IS_DAX(inode)) {
-		pfn_t pfn;
-
-		ret = xfs_dax_fault(vmf, order, write_fault, &pfn);
-		if (ret & VM_FAULT_NEEDDSYNC)
-			ret = dax_finish_sync_fault(vmf, order, pfn);
-	} else if (write_fault) {
-		ret = iomap_page_mkwrite(vmf, &xfs_page_mkwrite_iomap_ops);
-	} else {
-		ret = filemap_fault(vmf);
-	}
-
-	if (lock_mode)
-		xfs_iunlock(XFS_I(inode), lock_mode);
+	trace_xfs_filemap_fault(XFS_I(inode), order, write_fault);
 
 	if (write_fault)
-		sb_end_pagefault(inode->i_sb);
-	return ret;
+		return xfs_write_fault(vmf, order);
+	if (IS_DAX(inode))
+		return xfs_dax_read_fault(vmf, order);
+	return filemap_fault(vmf);
 }
 
 static inline bool