My local syzbot instance hit memory leak in copy_process(). The problem was in unputted task struct in _destroy_id().
Simple reproducer:
int main(void) { struct rdma_ucm_cmd_hdr *hdr; struct rdma_ucm_create_id *cmd_id; char cmd[sizeof(*hdr) + sizeof(*cmd_id)] = {0}; int fd;
hdr = (struct rdma_ucm_cmd_hdr *)cmd; cmd_id = (struct rdma_ucm_create_id *) (cmd + sizeof(*hdr));
hdr->cmd = 0; hdr->in = 0x18; hdr->out = 0xfa00;
cmd_id->uid = 0x3; cmd_id->response = 0x0; cmd_id->ps = 0x106;
fd = open("/dev/infiniband/rdma_cm", O_WRONLY); write(fd, cmd, sizeof(cmd)); }
Ftrace log:
ucma_open(); ucma_write() { ucma_create_id() { ucma_alloc_ctx(); rdma_create_user_id() { rdma_restrack_new(); rdma_restrack_set_name() { rdma_restrack_attach_task.part.0(); <--- task_struct getted } } ucma_destroy_private_ctx() { ucma_put_ctx(); rdma_destroy_id() { _destroy_id() <--- id_priv freed } } } } ucma_close();
From previous log it's easy to undertand that
_destroy_id() is the last place, where task_struct can be putted, because at the end of this function id_priv is freed.
With this patch applied, above reproducer doesn't hit memory leak anymore.
Fixes: e51060f08a61 ("IB: IP address based RDMA connection manager") Cc: stable@vger.kernel.org Signed-off-by: Pavel Skripkin paskripkin@gmail.com --- drivers/infiniband/core/cma.c | 1 + 1 file changed, 1 insertion(+)
diff --git a/drivers/infiniband/core/cma.c b/drivers/infiniband/core/cma.c index ab148a696c0c..2760352261b3 100644 --- a/drivers/infiniband/core/cma.c +++ b/drivers/infiniband/core/cma.c @@ -1874,6 +1874,7 @@ static void _destroy_id(struct rdma_id_private *id_priv,
kfree(id_priv->id.route.path_rec);
+ rdma_restrack_put(&id_priv->res); put_net(id_priv->id.route.addr.dev_addr.net); kfree(id_priv); }