From mboxrd@z Thu Jan 1 00:00:00 1970 From: Christian Ehrhardt Subject: Re: [PATCH v2] Memory leak when adding/removing vhost_user ports Date: Wed, 6 Jul 2016 14:37:43 +0200 Message-ID: References: <1467807898-27772-1-git-send-email-christian.ehrhardt@canonical.com> Mime-Version: 1.0 Content-Type: text/plain; charset=UTF-8 To: Christian Ehrhardt , Patrik Andersson R , Thomas Monjalon , dev , Yuanhan Liu , "Xie, Huawei" Return-path: Received: from mail-qk0-f182.google.com (mail-qk0-f182.google.com [209.85.220.182]) by dpdk.org (Postfix) with ESMTP id 098166CD4 for ; Wed, 6 Jul 2016 14:38:04 +0200 (CEST) Received: by mail-qk0-f182.google.com with SMTP id t127so313193792qkf.1 for ; Wed, 06 Jul 2016 05:38:03 -0700 (PDT) In-Reply-To: List-Id: patches and discussions about DPDK List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Sender: "dev" Now let us really get to this old discussion. Once again - sorry for mixing this up :-/ Found too much at once today. ## Ok - refocus ## Back then in http://dpdk.org/dev/patchwork/patch/12103/ http://dpdk.org/dev/patchwork/patch/12118/ We came up with a nice solution for the leak. But it was never picked up upstream. There were a lot of changes to all of that code, especially vhost client/server. Lacking an openvswitch for DPDK 16.07 I can't test if the issue still shows up, but looking at the code suggests it still does. Unfortunately the internal structures and scopes are slightly different so that I couldn't come up with a working forward port of the old patch. Attached is a patch as close as I got (obviously not working). I'm convinced that Yuanhan Liu and Xie Huawei with their deep context knowledge of dpdk's vhost_user will quickly know: - if the root cause still applies - what the best new way of fixing this would be As this makes long term usage of dpdk aborting by a leak I hope we have a chance to get this into 16.07 still. Kind Regards, Christian Christian Ehrhardt Software Engineer, Ubuntu Server Canonical Ltd On Wed, Jul 6, 2016 at 2:30 PM, Christian Ehrhardt < christian.ehrhardt@canonical.com> wrote: > This is the series this really belongs to > http://dpdk.org/dev/patchwork/patch/11581/ > Message ID < > 1458292380-9258-1-git-send-email-patrik.r.andersson@ericsson.com> > > Should I wait for a v2 to point the patch at the right ID or do you prefer > a fixed resubmit right away? > > Christian Ehrhardt > Software Engineer, Ubuntu Server > Canonical Ltd > > On Wed, Jul 6, 2016 at 2:26 PM, Christian Ehrhardt < > christian.ehrhardt@canonical.com> wrote: > >> Sorry, >> please ignore the two links, the cover letter has - they belong to a >> different issue I have to bring up again. >> Everything else still applies. >> >> Christian Ehrhardt >> Software Engineer, Ubuntu Server >> Canonical Ltd >> >> On Wed, Jul 6, 2016 at 2:24 PM, Christian Ehrhardt < >> christian.ehrhardt@canonical.com> wrote: >> >>> Hi, >>> while checking for dpdk 16.07 what backports are accepted in the >>> meantime so I >>> can drop them I found this particular discussion has been silently >>> forgotten by >>> all of us. >>> >>> Back then we had the patch and discussion first in >>> http://dpdk.org/dev/patchwork/patch/12103/ >>> and then >>> http://dpdk.org/dev/patchwork/patch/12118/ >>> >>> Things worked fine as I reported and I integrated the patch in our >>> packaging as >>> it fixed a severe issue. Since it was reported by someone else I do not >>> seem to >>> be the only one :-) >>> >>> So today I rebased the patch including my updates I made based on our >>> discussion >>> and I think it would make as much sense as it made back then to fix this. >>> >>> Christian Ehrhardt (1): >>> vhost_user: avoid crash when exeeding file descriptors >>> >>> lib/librte_vhost/vhost_user/fd_man.c | 11 ++++++----- >>> lib/librte_vhost/vhost_user/vhost-net-user.c | 19 +++++++++++++++++-- >>> 2 files changed, 23 insertions(+), 7 deletions(-) >>> >>> -- >>> 2.7.4 >>> >>> >> >