From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 36124 invoked by alias); 9 Nov 2016 12:18:56 -0000 Mailing-List: contact libc-alpha-help@sourceware.org; run by ezmlm Precedence: bulk List-Id: List-Subscribe: List-Archive: List-Post: List-Help: , Sender: libc-alpha-owner@sourceware.org Received: (qmail 36105 invoked by uid 89); 9 Nov 2016 12:18:54 -0000 Authentication-Results: sourceware.org; auth=none X-Virus-Found: No X-Spam-SWARE-Status: No, score=-4.8 required=5.0 tests=BAYES_00,RP_MATCHES_RCVD,SPF_HELO_PASS autolearn=ham version=3.3.2 spammy=structs.def, start_routine, set_tid_address, UD:getpid.c X-HELO: mx1.redhat.com Subject: Re: [PATCH] Remove cached PID/TID in clone To: Adhemerval Zanella , libc-alpha@sourceware.org References: <1476387924-4509-1-git-send-email-adhemerval.zanella@linaro.org> From: Florian Weimer Message-ID: <58c8bc8e-e643-2d3a-5e57-55a3457ca70f@redhat.com> Date: Wed, 09 Nov 2016 12:18:00 -0000 User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:45.0) Gecko/20100101 Thunderbird/45.4.0 MIME-Version: 1.0 In-Reply-To: Content-Type: text/plain; charset=UTF-8; format=flowed Content-Transfer-Encoding: 8bit X-SW-Source: 2016-11/txt/msg00325.txt.bz2 On 11/08/2016 08:58 PM, Adhemerval Zanella wrote: > The tid fields is basically used internally on pthread implementations > (including getpid) and since correct usage means thread *must* be > created using pthread_create we are sure the tid field will be > correctly set due 'set_tid_address' from __pthread_initialize_pids. Thanks for the explanation. I really think we should document the clone system call wrapper and spell out these requirements, but that's a separate matter. >> > Please rename to “pid_unused” or something like that, to make sure it's no longer referenced. > I renamed it on my local branch and I also updated the change spot > that it incur: > > diff --git a/nptl_db/structs.def b/nptl_db/structs.def > index a9b621b..1cb6a46 100644 > --- a/nptl_db/structs.def > +++ b/nptl_db/structs.def > @@ -48,7 +48,6 @@ DB_STRUCT (pthread) > DB_STRUCT_FIELD (pthread, list) > DB_STRUCT_FIELD (pthread, report_events) > DB_STRUCT_FIELD (pthread, tid) > -DB_STRUCT_FIELD (pthread, pid) > DB_STRUCT_FIELD (pthread, start_routine) > DB_STRUCT_FIELD (pthread, cancelhandling) > DB_STRUCT_FIELD (pthread, schedpolicy) Have you tested that thread debugging still works after these changes (at least on one architecture)? > The patch also removes the TID caching in clone. My understanding for > such semantic is try provide some pthread usage after a user program > issue clone directly (as done by thread creation with CLONE_PARENT_SETTID > and pthread tid member). However, as stated before in multiple threads, “discussion threads”? ☹ > GLIBC provides clone syscalls without further supporting all this > semantics. It means that, although GLIBC currently tries a better effort, > since it does not make any more guarantees, specially for newer and newer > clone flags. I don't quite understand the above part. > * sysdeps/unix/sysv/linux/getpid.c: Likewise. This needs updating (file was removed). I do not have further comments, but I have not reviewed the assembler language implementations (only i386/x86_64). I support the removal of PID caching, though. Thanks, Florian