Return-Path: Received: from mx1.redhat.com ([209.132.183.28]:51716 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753610AbcD0Rud (ORCPT ); Wed, 27 Apr 2016 13:50:33 -0400 Subject: Re: [PATCH v3 1/3] gssd: use pthreads to handle upcalls To: "Kornievskaia, Olga" References: <1461603513-67523-1-git-send-email-kolga@netapp.com> <1461603513-67523-2-git-send-email-kolga@netapp.com> <5720D2BA.5010400@RedHat.com> <59AF3631-819C-42E5-AB0D-3483590E279D@netapp.com> <5720D977.10008@RedHat.com> <57DB10EC-2538-4191-B5D7-03D53FD1F9C9@netapp.com> Cc: "linux-nfs@vger.kernel.org" From: Steve Dickson Message-ID: <5720FBE7.2000300@RedHat.com> Date: Wed, 27 Apr 2016 13:50:31 -0400 MIME-Version: 1.0 In-Reply-To: Content-Type: text/plain; charset=windows-1252 Sender: linux-nfs-owner@vger.kernel.org List-ID: On 04/27/2016 12:33 PM, Kornievskaia, Olga wrote: >>>> I think the biggest saving would be to take the next step >>>> and do the pool of threads that are pre-created (and thus >>>> not taking time to be allocated for each upcall). Then we >>>> need a way to make sure we cleanup any threads that might >>>> be hanging for whatever reason. >>> This will be an excellent next step! >> >> This patch series is suppose to show ?proof-of-concept? and >> functionally equivalent but better to what we have now. >> Old: upcall -> new process. New: upcall -> new thread. >> Old: process hangs -> resources not cleaned and no new upcalls. >> New: thread hangs -> resources not cleaned and new upcalls can proceed. Well I think you proved it... Nice work! >> >> I will submit the new version with: >> ? make function inline >> ? change setresuid/setresgid to set all three uids: real, effective, saved. git blame tells me Jeff you wrote the comment about attacks from ptrace(). > > Sorry I unintentionally hit send. I wanted to say that then I?ll work on > getting the pool of thread but working on top of these patches. > Or do you want me to keep working and then submit that version > as if this one wasn?t written? Let's get these in now so they will get some real world testing... So lets build on these... steved.