On 02/01/2016 10:53 AM, Insu Yun wrote:
When kthread_run fails, it returns ERR, not NULL. Therefore, NULL checking is redundant. (https://www.kernel.org/doc/htmldocs/device-drivers/API-kthread-run.html)
Thanks, the change looks reasonable. ...
task = kthread_run(kthread, k, "%s", k->name); - if (task == NULL || IS_ERR(task)) + if (IS_ERR(task)) return -ENOMEM;
Interestingly, after this change, it's more clear that returning -ENOMEM is a bit misleading when the underlying kernel/kthread.c function, kthread_create_on_node returns ERR_PTR(-EINTR). But because that would happen when the process we need to have the driver working gets killed, I'm not sure it matters. -- Ed