Re: [PATCH] RDMA/device: Fix a race between mad_client and cm_client init

[Shifeng Li]

On 2024/1/4 20:37, Jason Gunthorpe wrote:
> On Thu, Jan 04, 2024 at 02:48:14PM +0800, Shifeng Li wrote:
>
> > The root cause is that mad_client and cm_client may init concurrently
> > when devices_rwsem write semaphore is downgraded in enable_device_and_get() like:
>
> That can't be true, the module loader infrastructue ensures those two
> things are sequential.

I'm a bit confused how the module loader infrastructue ensures that mad_client.add() and
cm_client.add() are sequential. Could you explain in more detail please?

We know that the ib_cm driver and mlx5_ib driver can load concurrently.

Thanks.

> You are trying to say that the post-client fixup stuff will still see
> the DEVICE_REGISTERED before it reaches the clients_rwsem lock?
>
> That probably just says the clients_rwsem should be obtained before
> changing the DEVICE_STATE too :\
>
> Jason