Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
References: <20220831180950.76907-1-schspa@gmail.com>
 <Yw/HmHcmXBVIg/SW@codewreck.org>
User-agent: mu4e 1.6.10; emacs 27.2
From:   Schspa Shi <schspa@gmail.com>
To:     asmadeus@codewreck.org
Cc:     ericvh@gmail.com, lucho@ionkov.net, linux_oss@crudebyte.com,
        davem@davemloft.net, edumazet@google.com, kuba@kernel.org,
        pabeni@redhat.com, v9fs-developer@lists.sourceforge.net,
        netdev@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH] p9: trans_fd: Fix deadlock when connection cancel
Date:   Thu, 01 Sep 2022 10:55:36 +0800
In-reply-to: <Yw/HmHcmXBVIg/SW@codewreck.org>
Message-ID: <m2bkrz7qc8.fsf@gmail.com>
MIME-Version: 1.0
Content-Type: text/plain
Precedence: bulk


asmadeus@codewreck.org writes:

> Schspa Shi wrote on Thu, Sep 01, 2022 at 02:09:50AM +0800:
>> To fix it, we can add extra reference counter to avoid deadlock, and
>> decrease it after we unlock the client->lock.
>
> Thanks for the patch!
>
> Unfortunately I already sent a slightly different version to the list,
> hidden in another syzbot thread, here:
> https://lkml.kernel.org/r/YvyD053bdbGE9xoo@codewreck.org
>
> (yes, sorry, not exactly somewhere I'd expect someone to find it... 9p
> hasn't had many contributors recently)
>
>
> Basically instead of taking an extra lock I just released the client
> lock before calling p9_client_cb, so it shouldn't hang anymore.
>
> We don't need the lock to call the cb as in p9_conn_cancel we already
> won't accept any new request and by this point the requests are in a
> local list that isn't shared anywhere.
>

Ok, thank you for pointing that out.

> If you have a test setup, would you mind testing my patch?
> That's the main reason I was delaying pushing it.
>

I have test it with my enviroment, it not hang anymore.

> Since you went out of your way to make this patch if you agree with my
> approach I don't mind adding your sign off or another mark of having
> worked on it.
>
> Thank you,

-- 
BRs
Schspa Shi