On Tue, Apr 02, 2019 at 11:11:08AM -0700, Bart Van Assche wrote:
> On Tue, 2019-04-02 at 10:55 +0800, Ming Lei wrote:
> > Also, isn't it the typical practice to release kobject related resources in
> > its release handler?
>
> A typical approach is to call kobject_del() before starting to clean up a
Yes.
> resource and to defer the final kfree() call to the release handler. I think
> that's how it works today in the block layer core. That is easy to see in the
> blk_mq_hw_sysfs_release() implementation:
>
> static void blk_mq_hw_sysfs_release(struct kobject *kobj)
> {
> struct blk_mq_hw_ctx *hctx = container_of(kobj, struct blk_mq_hw_ctx,
> kobj);
> free_cpumask_var(hctx->cpumask);
> kfree(hctx->ctxs);
> kfree(hctx);
> }
Right, then we should free other hctx fields here too, it is safe
and reliable.
Thanks,
Ming
