Merge branch 'xfs-misc-fixes-for-3.18-2' into for-next

fs/xfs/libxfs/xfs_da_format.c

@@ -270,7 +270,6 @@ xfs_dir3_data_get_ftype(
 {
 	__uint8_t	ftype = dep->name[dep->namelen];
 
-	ASSERT(ftype < XFS_DIR3_FT_MAX);
 	if (ftype >= XFS_DIR3_FT_MAX)
 		return XFS_DIR3_FT_UNKNOWN;
 	return ftype;

fs/xfs/xfs_bmap_util.c

@@ -1646,7 +1646,7 @@ xfs_swap_extents_check_format(
 	return 0;
 }
 
-int
+static int
 xfs_swap_extent_flush(
 	struct xfs_inode	*ip)
 {

fs/xfs/xfs_buf_item.c

@@ -501,7 +501,7 @@ xfs_buf_item_unpin(
  * buffer being bad..
  */
 
-DEFINE_RATELIMIT_STATE(xfs_buf_write_fail_rl_state, 30 * HZ, 10);
+static DEFINE_RATELIMIT_STATE(xfs_buf_write_fail_rl_state, 30 * HZ, 10);
 
 STATIC uint
 xfs_buf_item_push(

fs/xfs/xfs_icache.c

@@ -33,7 +33,6 @@
 #include "xfs_trace.h"
 #include "xfs_icache.h"
 #include "xfs_bmap_util.h"
-#include "xfs_quota.h"
 #include "xfs_dquot_item.h"
 #include "xfs_dquot.h"
 

fs/xfs/xfs_iops.c

@@ -849,6 +849,36 @@ xfs_setattr_size(
 		return error;
 	truncate_setsize(inode, newsize);
 
+	/*
+	 * The "we can't serialise against page faults" pain gets worse.
+	 *
+	 * If the file is mapped then we have to clean the page at the old EOF
+	 * when extending the file. Extending the file can expose changes the
+	 * underlying page mapping (e.g. from beyond EOF to a hole or
+	 * unwritten), and so on the next attempt to write to that page we need
+	 * to remap it for write. i.e. we need .page_mkwrite() to be called.
+	 * Hence we need to clean the page to clean the pte and so a new write
+	 * fault will be triggered appropriately.
+	 *
+	 * If we do it before we change the inode size, then we can race with a
+	 * page fault that maps the page with exactly the same problem. If we do
+	 * it after we change the file size, then a new page fault can come in
+	 * and allocate space before we've run the rest of the truncate
+	 * transaction. That's kinda grotesque, but it's better than have data
+	 * over a hole, and so that's the lesser evil that has been chosen here.
+	 *
+	 * The real solution, however, is to have some mechanism for locking out
+	 * page faults while a truncate is in progress.
+	 */
+	if (newsize > oldsize && mapping_mapped(VFS_I(ip)->i_mapping)) {
+		error = filemap_write_and_wait_range(
+				VFS_I(ip)->i_mapping,
+				round_down(oldsize, PAGE_CACHE_SIZE),
+				round_up(oldsize, PAGE_CACHE_SIZE) - 1);
+		if (error)
+			return error;
+	}
+
 	tp = xfs_trans_alloc(mp, XFS_TRANS_SETATTR_SIZE);
 	error = xfs_trans_reserve(tp, &M_RES(mp)->tr_itruncate, 0, 0);
 	if (error)

fs/xfs/xfs_log_cil.c

@@ -463,12 +463,40 @@ xlog_cil_push(
 		spin_unlock(&cil->xc_push_lock);
 		goto out_skip;
 	}
-	spin_unlock(&cil->xc_push_lock);
 
 
 	/* check for a previously pushed seqeunce */
-	if (push_seq < cil->xc_ctx->sequence)
+	if (push_seq < cil->xc_ctx->sequence) {
+		spin_unlock(&cil->xc_push_lock);
 		goto out_skip;
+	}
+
+	/*
+	 * We are now going to push this context, so add it to the committing
+	 * list before we do anything else. This ensures that anyone waiting on
+	 * this push can easily detect the difference between a "push in
+	 * progress" and "CIL is empty, nothing to do".
+	 *
+	 * IOWs, a wait loop can now check for:
+	 *	the current sequence not being found on the committing list;
+	 *	an empty CIL; and
+	 *	an unchanged sequence number
+	 * to detect a push that had nothing to do and therefore does not need
+	 * waiting on. If the CIL is not empty, we get put on the committing
+	 * list before emptying the CIL and bumping the sequence number. Hence
+	 * an empty CIL and an unchanged sequence number means we jumped out
+	 * above after doing nothing.
+	 *
+	 * Hence the waiter will either find the commit sequence on the
+	 * committing list or the sequence number will be unchanged and the CIL
+	 * still dirty. In that latter case, the push has not yet started, and
+	 * so the waiter will have to continue trying to check the CIL
+	 * committing list until it is found. In extreme cases of delay, the
+	 * sequence may fully commit between the attempts the wait makes to wait
+	 * on the commit sequence.
+	 */
+	list_add(&ctx->committing, &cil->xc_committing);
+	spin_unlock(&cil->xc_push_lock);
 
 	/*
 	 * pull all the log vectors off the items in the CIL, and
@@ -532,7 +560,6 @@ xlog_cil_push(
 	 */
 	spin_lock(&cil->xc_push_lock);
 	cil->xc_current_sequence = new_ctx->sequence;
-	list_add(&ctx->committing, &cil->xc_committing);
 	spin_unlock(&cil->xc_push_lock);
 	up_write(&cil->xc_ctx_lock);
 
@@ -855,13 +882,15 @@ restart:
 	 * Hence by the time we have got here it our sequence may not have been
 	 * pushed yet. This is true if the current sequence still matches the
 	 * push sequence after the above wait loop and the CIL still contains
-	 * dirty objects.
+	 * dirty objects. This is guaranteed by the push code first adding the
+	 * context to the committing list before emptying the CIL.
 	 *
-	 * When the push occurs, it will empty the CIL and atomically increment
-	 * the currect sequence past the push sequence and move it into the
-	 * committing list. Of course, if the CIL is clean at the time of the
-	 * push, it won't have pushed the CIL at all, so in that case we should
-	 * try the push for this sequence again from the start just in case.
+	 * Hence if we don't find the context in the committing list and the
+	 * current sequence number is unchanged then the CIL contents are
+	 * significant.  If the CIL is empty, if means there was nothing to push
+	 * and that means there is nothing to wait for. If the CIL is not empty,
+	 * it means we haven't yet started the push, because if it had started
+	 * we would have found the context on the committing list.
 	 */
 	if (sequence == cil->xc_current_sequence &&
 	    !list_empty(&cil->xc_cil)) {

fs/xfs/xfs_rtalloc.c

@@ -46,7 +46,7 @@
  * Keeps track of a current summary block, so we don't keep reading
  * it from the buffer cache.
  */
-int
+static int
 xfs_rtget_summary(
 	xfs_mount_t	*mp,		/* file system mount structure */
 	xfs_trans_t	*tp,		/* transaction pointer */