Re: for-next: KCSAN failures on 6130a25681d4 (kdave/for-next) Merge branch 'for-next-next-v5.20-20220804' into for-next-20220804

[David Sterba <dsterba@suse.cz>]

On Mon, Sep 05, 2022 at 05:42:03PM +0200, David Sterba wrote:
> On Tue, Aug 09, 2022 at 08:35:42AM +0100, Filipe Manana wrote:
> > On Tue, Aug 9, 2022 at 4:33 AM Zygo Blaxell
> > <ce3g8jdj@umail.furryterror.org> wrote:
> > >
> > > Some KCSAN complaints I found while testing for other things...
> > >
> > > Here's one related to extent refs:
> > 
> > It's about the block reserves, nothing to do with extents refs.
> > 
> > These get reported every now and then like here:
> > 
> > https://lore.kernel.org/linux-btrfs/CAAwBoOJDjei5Hnem155N_cJwiEkVwJYvgN-tQrwWbZQGhFU=cA@mail.gmail.com/
> > 
> > It's actually harmless, but if we keep it like this, we'll keep
> > getting reports in the future.
> 
> Can we add some kind of annotation so KCSAN understands that? The ->full
> member would be accessed using a helper when outside of the lock so the
> annotation can be there.

Something like (only compile-tested):

diff --git a/fs/btrfs/block-rsv.c b/fs/btrfs/block-rsv.c
index 6ce704d3bdd2..ec96285357e0 100644
--- a/fs/btrfs/block-rsv.c
+++ b/fs/btrfs/block-rsv.c
@@ -286,7 +286,7 @@ u64 btrfs_block_rsv_release(struct btrfs_fs_info *fs_info,
 	 */
 	if (block_rsv == delayed_rsv)
 		target = global_rsv;
-	else if (block_rsv != global_rsv && !delayed_rsv->full)
+	else if (block_rsv != global_rsv && !btrfs_block_rsv_full(delayed_rsv))
 		target = delayed_rsv;
 
 	if (target && block_rsv->space_info != target->space_info)
diff --git a/fs/btrfs/block-rsv.h b/fs/btrfs/block-rsv.h
index 0c183709be00..a1f40b88fa82 100644
--- a/fs/btrfs/block-rsv.h
+++ b/fs/btrfs/block-rsv.h
@@ -91,5 +91,9 @@ static inline void btrfs_unuse_block_rsv(struct btrfs_fs_info *fs_info,
 	btrfs_block_rsv_add_bytes(block_rsv, blocksize, false);
 	btrfs_block_rsv_release(fs_info, block_rsv, 0, NULL);
 }
+static inline bool btrfs_block_rsv_full(const struct btrfs_block_rsv *rsv)
+{
+	return data_race(rsv->full);
+}
 
 #endif /* BTRFS_BLOCK_RSV_H */
diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index a0a2652962ec..eb0ae7e396ef 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -2426,7 +2426,7 @@ static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
 	 * still can allocate chunks and thus are fine using the currently
 	 * calculated f_bavail.
 	 */
-	if (!mixed && btrfs_block_rsv_full(block_rsv->space_info) &&
+	if (!mixed && block_rsv->space_info->full &&
 	    total_free_meta - thresh < block_rsv->size)
 		buf->f_bavail = 0;
 
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index d9e608935e64..9d7563df81a0 100644
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -594,7 +594,7 @@ start_transaction(struct btrfs_root *root, unsigned int num_items,
 		 */
 		num_bytes = btrfs_calc_insert_metadata_size(fs_info, num_items);
 		if (flush == BTRFS_RESERVE_FLUSH_ALL &&
-		    delayed_refs_rsv->full == 0) {
+		    btrfs_block_rsv_full(delayed_refs_rsv) == 0) {
 			delayed_refs_bytes = num_bytes;
 			num_bytes <<= 1;
 		}
@@ -619,7 +619,7 @@ start_transaction(struct btrfs_root *root, unsigned int num_items,
 		if (rsv->space_info->force_alloc)
 			do_chunk_alloc = true;
 	} else if (num_items == 0 && flush == BTRFS_RESERVE_FLUSH_ALL &&
-		   !delayed_refs_rsv->full) {
+		   !btrfs_block_rsv_full(delayed_refs_rsv)) {
 		/*
 		 * Some people call with btrfs_start_transaction(root, 0)
 		 * because they can be throttled, but have some other mechanism