libbpf: Fixes incorrect rx_ring_setup_done
authorKev Jackson <foamdino@gmail.com>
Mon, 7 Jun 2021 13:08:35 +0000 (14:08 +0100)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Wed, 23 Jun 2021 12:42:41 +0000 (14:42 +0200)
[ Upstream commit 11fc79fc9f2e395aa39fa5baccae62767c5d8280 ]

When calling xsk_socket__create_shared(), the logic at line 1097 marks a
boolean flag true within the xsk_umem structure to track setup progress
in order to support multiple calls to the function.  However, instead of
marking umem->tx_ring_setup_done, the code incorrectly sets
umem->rx_ring_setup_done.  This leads to improper behaviour when
creating and destroying xsk and umem structures.

Multiple calls to this function is documented as supported.

Fixes: ca7a83e2487a ("libbpf: Only create rx and tx XDP rings when necessary")
Signed-off-by: Kev Jackson <foamdino@gmail.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Acked-by: Yonghong Song <yhs@fb.com>
Link: https://lore.kernel.org/bpf/YL4aU4f3Aaik7CN0@linux-dev
Signed-off-by: Sasha Levin <sashal@kernel.org>
tools/lib/bpf/xsk.c

index 7150e34..3028f93 100644 (file)
@@ -779,7 +779,7 @@ int xsk_socket__create_shared(struct xsk_socket **xsk_ptr,
                        goto out_put_ctx;
                }
                if (xsk->fd == umem->fd)
-                       umem->rx_ring_setup_done = true;
+                       umem->tx_ring_setup_done = true;
        }
 
        err = xsk_get_mmap_offsets(xsk->fd, &off);