bcache: bugfix for race between moving_gc and bucket_invalidate (6d3d1a9c) · Commits · 戴 / test

There is a possibility for a bucket to be invalidated by the allocator while moving_gc was copying it's contents to another bucket, if the bucket only held cached data. To prevent this moving checks for a stale ptr (to an invalidated bucket), before and after reads. It it finds one, it simply ignores moving that data. This only affects bcache if the moving_gc was turned on, note that it's off by default. Signed-off-by:

Nicholas Swenson <nks@daterainc.com> Signed-off-by:

Kent Overstreet <kmo@daterainc.com>

drivers/md/bcache/movinggc.c

+10 −0

Original line number	Diff line number	Diff line
		@@ -64,11 +64,16 @@ static void write_moving_finish(struct closure *cl)

		static void read_moving_endio(struct bio *bio, int error)
		{
		struct bbio *b = container_of(bio, struct bbio, bio);
		struct moving_io *io = container_of(bio->bi_private,
		struct moving_io, cl);

		if (error)
		io->op.error = error;
		else if (!KEY_DIRTY(&b->key) &&
		ptr_stale(io->op.c, &b->key, 0)) {
		io->op.error = -EINTR;
		}

		bch_bbio_endio(io->op.c, bio, error, "reading data to move");
		}
		@@ -140,6 +145,11 @@ static void read_moving(struct cache_set *c)
		if (!w)
		break;

		if (ptr_stale(c, &w->key, 0)) {
		bch_keybuf_del(&c->moving_gc_keys, w);
		continue;
		}

		io = kzalloc(sizeof(struct moving_io) + sizeof(struct bio_vec)
		* DIV_ROUND_UP(KEY_SIZE(&w->key), PAGE_SECTORS),
		GFP_KERNEL);

Admin message