From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <axboe@kernel.dk>
Subject: Re: Linux OS killed fio process because fio invoked oom_killer
References: <CADvrUP8gH3S-fLnHU4QXJLO87YzUge8MxKt3WdffN+9uiVpH3w@mail.gmail.com>
 <56F17326.3000908@kernel.dk>
 <CADvrUP9duNxzfMuQum-e_YXqVUjG_K8any52e7sQn6ZcBy9o8Q@mail.gmail.com>
 <61A20AC4-3F8E-4B1E-A2C9-81BE827B188F@kernel.dk>
 <CADvrUP-C-Sm_kGbQTNZRitVMi8ory4GrMNy7+oXFeE=fYRJnvQ@mail.gmail.com>
 <BN3PR0401MB145719A81EC72FA99E7BB8A3F7810@BN3PR0401MB1457.namprd04.prod.outlook.com>
 <CADvrUP-Lz4WSzx6H+0bWpRtF5rp+FFSq0Atso8uaar=mt08zHQ@mail.gmail.com>
 <BN3PR0401MB145727B4551C1144668618CDF7810@BN3PR0401MB1457.namprd04.prod.outlook.com>
 <CADvrUP-x8Ur+K1WvUuFBr8P9og4183CRYj7+ihhWNrb9iPU3NA@mail.gmail.com>
 <56F3F481.6050508@kernel.dk> <56F44286.8000706@kernel.dk>
 <CADvrUP8WUwWUJYOr_8zdkgM3+k-JAm9oMVseHCs1DEW9z7x9QA@mail.gmail.com>
From: Jens Axboe <axboe@kernel.dk>
Message-ID: <56F554CA.7080205@kernel.dk>
Date: Fri, 25 Mar 2016 09:10:02 -0600
MIME-Version: 1.0
In-Reply-To: <CADvrUP8WUwWUJYOr_8zdkgM3+k-JAm9oMVseHCs1DEW9z7x9QA@mail.gmail.com>
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Transfer-Encoding: 7bit
To: flash yan <flashyan83@gmail.com>
Cc: Jeff Furlong <jeff.furlong@hgst.com>, "fio@vger.kernel.org" <fio@vger.kernel.org>
List-ID: <fio@vger.kernel.org>

On 03/25/2016 09:03 AM, flash yan wrote:
> I have tried your patch. It fixed the issue. But I don't understand
> "So we just kept piling on buffers to verify, but we never did".
> You mean that fio didn't run any verify job with verify_async enabled?

Right, it did not for your case. But that's because of the super short 
loop and the time_based setup, otherwise it would have. That was 
apparent while it was running too (not reads or V status), and in the 
final output (no reads there either). So it was a corner case where 
verify did not happen, since the referenced commit that broke it for 
that case.

-- 
Jens Axboe