Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:18 as permitted sender) client-ip=2620:137:e000::1:18;
Message-ID: <b32cf3e2-a68c-b1b0-f3da-72e5f0b9d86c@kernel.dk>
Date:   Thu, 21 Apr 2022 17:33:04 -0600
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux aarch64; rv:91.0) Gecko/20100101
 Thunderbird/91.8.0
From:   Jens Axboe <axboe@kernel.dk>
Subject: Re: [PATCH 6/6] io_uring: allow NOP opcode in IOPOLL mode
To:     Dylan Yudaken <dylany@fb.com>
Cc:     io-uring <io-uring@vger.kernel.org>,
        "Pavel Begunkov (Silence)" <asml.silence@gmail.com>,
        linux-kernel <linux-kernel@vger.kernel.org>,
        FB Kernel Team <kernel-team@fb.com>,
        Dylan Yudaken <dylany@fb.com>
References: <20220421091345.2115755-1-dylany@fb.com>
 <20220421091345.2115755-7-dylany@fb.com>
Content-Language: en-US
In-Reply-To: <20220421091345.2115755-7-dylany@fb.com>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 7bit
Precedence: bulk

On Thu, Apr 21, 2022 at 3:17 AM Dylan Yudaken <dylany@fb.com> wrote:
>
> This is useful for tests so that IOPOLL can be tested without requiring
> files. NOP is acceptable in IOPOLL as it always completes immediately.

This one actually breaks two liburing test cases (link and defer) that
assume NOP on IOPOLL will return -EINVAL. Not a huge deal, but we do
need to figure out how to make them reliably -EINVAL in a different
way then.

Maybe add a nop_flags to the usual flags spot in the sqe, and define
a flag that says NOP_IOPOLL or something. Require this flag set for
allowing NOP on iopoll. That'd allow testing, but still retain the
-EINVAL behavior if not set.

Alternatively, modify test cases...

I'll drop this one for now, just because it fails the regression
tests.

-- 
Jens Axboe