Re: [PATCH] xprtrdma: fix EP destruction logic

[Chuck Lever <chuck.lever@oracle.com>]

> On Jun 26, 2020, at 11:10 AM, Dan Aloni <dan@kernelim.com> wrote:
> 
> On Fri, Jun 26, 2020 at 08:56:41AM -0400, Chuck Lever wrote:
>>> On Jun 26, 2020, at 3:10 AM, Dan Aloni <dan@kernelim.com> wrote:
> [..]
>>> - Add a mutex in `rpcrdma_ep_destroy` to guard against concurrent calls
>>> to `rpcrdma_xprt_disconnect` coming from either `rpcrdma_xprt_connect`
>>> or `xprt_rdma_close`.
>> 
>> NAK. The RPC client provides appropriate exclusion, please let's not
>> add more serialization that can introduce further deadlocks.
> 
> It appeared to me that this exclusion does not works well. As for my
> considerations, if I am not mistaken from analyzing crashes I've
> seen:
> 
>   -> xprt_autoclose (running on xprtiod)
>     -> xprt->ops->close
>       -> xprt_rdma_close
>         -> rpcrdma_xprt_disconnect
> 
> and:
> 
>    -> xprt_rdma_connect_worker (running on xprtiod)
>      -> rpcrdma_xprt_connect
> 	-> rpcrdma_xprt_disconnect
> 
> I understand the rationale or at least the aim that `close` and
> `connect` ops should not be concurrent on the same `xprt`, however:
> 
> * `xprt_force_disconnect`, is called from various places, queues
>  a call to `xprt_autoclose` to the background on `xprtiod` workqueue item,
>  conditioned that `!XPRT_LOCKED` which is the case for connect that went
>  to the background.
> * `xprt_rdma_connect` also sends `xprt_rdma_connect_worker` as an `xprtiod`
>  workqueue item, unconditionally.
> 
> So we have two work items that can run in parallel, and I don't see
> clear gating on this from the code.

If close and connect are being called concurrently on the same xprt,
then there is a bug in the generic RPC xprt code. I don't believe
that to be the case here.

If xprtrdma invokes force_disconnect during connect processing,
XPRT_LOCKED should be held and the close should be delayed.


> Maybe there's a simpler fix for this. Perhaps a
> `cancel_delayed_work_sync(&r_xprt->rx_connect_worker);` is appropriate
> in `xprt_rdma_close`?

There are simpler, less deadlock-prone, and more justifiable fixes.
Please stand by, I will take care of this today.

--
Chuck Lever