Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1754879AbbLKW03 (ORCPT ); Fri, 11 Dec 2015 17:26:29 -0500 Received: from mail.kernel.org ([198.145.29.136]:33517 "EHLO mail.kernel.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753581AbbLKW01 (ORCPT ); Fri, 11 Dec 2015 17:26:27 -0500 Date: Fri, 11 Dec 2015 19:26:21 -0300 From: Arnaldo Carvalho de Melo To: Masami Hiramatsu Cc: Peter Zijlstra , Adrian Hunter , linux-kernel@vger.kernel.org, linux-perf-users@vger.kernel.org, Ingo Molnar , Namhyung Kim , Jiri Olsa , Wang Nan , Alexei Starovoitov Subject: Re: [PATCH perf/core 00/22] perf refcnt debugger API and fixes Message-ID: <20151211222621.GE6843@kernel.org> References: <20151209021047.10245.8918.stgit@localhost.localdomain> <20151209134138.GB15864@kernel.org> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20151209134138.GB15864@kernel.org> X-Url: http://acmel.wordpress.com User-Agent: Mutt/1.5.23 (2014-03-12) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 8202 Lines: 249 Em Wed, Dec 09, 2015 at 10:41:38AM -0300, Arnaldo Carvalho de Melo escreveu: > Em Wed, Dec 09, 2015 at 11:10:48AM +0900, Masami Hiramatsu escreveu: > > General refcnt miscodings > > ========================= > > > > BTW, while applying this change, I've found that there are refcnt > > coding mismatches in those code and most of the bugs come from those > > mismatches. > > > > - The reference counter can be initialized by 0 or 1. > > - If 0 is chosen, caller have to get it and free it if failed to > > register appropriately. > > - If 1 is chosen, caller doesn't need to get it, but when exits the > > caller, it has to put it. (except for returning the object itself) > > - The application should choose either one as its policy, to avoid > > confusion. > > > > perf tools mixes it up (moreover, it initializes 2 in a case) and > > caller usually forgets to put it (it is not surprising, because too > > many "put" usually cause SEGV by accessing freed object.) > > Well, we should fix the bugs and document why some initialization is > deemed better than a single initialization style. > For instance, if we know that we will keep multiple references straight > away, then we could init it with some value different that preferred, > that I aggee, is 1, i.e. the usual way for some constructor is: > > > struct foo *foo__new() > { > struct foo *f = malloc(sizeof (*f)); > > if (f) { > atomic_set(&f->refcnt, 1); > } > > return f; > } > > void *foo__delete(struct foo *f) > { > free(f); > } > > void foo__put(struct foo *f) > { > if (f && atomic_dec_and_test(f->refcnt)) > foo__delete(f); > } > > > Then, when using if, and before adding it to any other tree, list, i.e. > a container, we do: > > struct foo *f = foo__new(); > > /* > * assume f was allocated and then the function allocating it > * failed, when it bails out it should just do: > */ > out_error: > foo__put(f); > > And that will make it hit zero, which will call foo__delete(), case > closed. > > > As far as I can see, cgroup_sel, perf_mmap(this is initialized 0 or 2...), > > thread, and comm_str are initialized by 0. Others are initialized by 1. > > > > So, I'd like to suggest that we choose one policy and cleanup the code. > > I recommend to use init by 1 policy, because anyway caller has to get > > See above, if nothing else recommends using a different value, use 1. So, I think I fixed the thread->refcnt case, please take a look at my perf/core branch, more specifically this patch: >From a9a8442c64a86599600ecb2ae0f296b5c93e2a98 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Fri, 11 Dec 2015 19:11:23 -0300 Subject: [PATCH 1/1] perf thread: Fix reference count initial state We should always return from thread__new(), the constructor, with the object with a reference count of one, so that: struct thread *thread = thread__new(); thread__put(thread); Will call thread__delete(). If any reference is made to that 'thread' variable, it better use thread__get(thread) to hold a reference. We were returning with thread->refcnt set to zero, fix it and some cases where thread__delete() was being called, which were not a problem because just one reference was being used, now that we set it to 1, use thread__put() instead. Reported-by: Masami Hiramatsu Cc: Adrian Hunter Cc: David Ahern Cc: Jiri Olsa Cc: Namhyung Kim Cc: Wang Nan Link: http://lkml.kernel.org/n/tip-4b9mkuk66to4ecckpmpvqx6s@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/intel-pt.c | 4 ++-- tools/perf/util/machine.c | 19 ++++++++++++------- tools/perf/util/thread.c | 10 ++++++++-- 3 files changed, 22 insertions(+), 11 deletions(-) diff --git a/tools/perf/util/intel-pt.c b/tools/perf/util/intel-pt.c index 97f963a3dcb9..81a2eb77ba7f 100644 --- a/tools/perf/util/intel-pt.c +++ b/tools/perf/util/intel-pt.c @@ -1744,7 +1744,7 @@ static void intel_pt_free(struct perf_session *session) auxtrace_heap__free(&pt->heap); intel_pt_free_events(session); session->auxtrace = NULL; - thread__delete(pt->unknown_thread); + thread__put(pt->unknown_thread); free(pt); } @@ -2153,7 +2153,7 @@ int intel_pt_process_auxtrace_info(union perf_event *event, return 0; err_delete_thread: - thread__delete(pt->unknown_thread); + thread__zput(pt->unknown_thread); err_free_queues: intel_pt_log_disable(); auxtrace_queues__free(&pt->queues); diff --git a/tools/perf/util/machine.c b/tools/perf/util/machine.c index 1407d5107480..ad79297c76c8 100644 --- a/tools/perf/util/machine.c +++ b/tools/perf/util/machine.c @@ -352,13 +352,18 @@ static void machine__update_thread_pid(struct machine *machine, } th->mg = map_groups__get(leader->mg); - +out_put: + thread__put(leader); return; - out_err: pr_err("Failed to join map groups for %d:%d\n", th->pid_, th->tid); + goto out_put; } +/* + * Caller must eventually drop thread->refcnt returned with a successfull + * lookup/new thread inserted. + */ static struct thread *____machine__findnew_thread(struct machine *machine, pid_t pid, pid_t tid, bool create) @@ -376,7 +381,7 @@ static struct thread *____machine__findnew_thread(struct machine *machine, if (th != NULL) { if (th->tid == tid) { machine__update_thread_pid(machine, th, pid); - return th; + return thread__get(th); } machine->last_match = NULL; @@ -389,7 +394,7 @@ static struct thread *____machine__findnew_thread(struct machine *machine, if (th->tid == tid) { machine->last_match = th; machine__update_thread_pid(machine, th, pid); - return th; + return thread__get(th); } if (tid < th->tid) @@ -417,7 +422,7 @@ static struct thread *____machine__findnew_thread(struct machine *machine, if (thread__init_map_groups(th, machine)) { rb_erase_init(&th->rb_node, &machine->threads); RB_CLEAR_NODE(&th->rb_node); - thread__delete(th); + thread__put(th); return NULL; } /* @@ -441,7 +446,7 @@ struct thread *machine__findnew_thread(struct machine *machine, pid_t pid, struct thread *th; pthread_rwlock_wrlock(&machine->threads_lock); - th = thread__get(__machine__findnew_thread(machine, pid, tid)); + th = __machine__findnew_thread(machine, pid, tid); pthread_rwlock_unlock(&machine->threads_lock); return th; } @@ -451,7 +456,7 @@ struct thread *machine__find_thread(struct machine *machine, pid_t pid, { struct thread *th; pthread_rwlock_rdlock(&machine->threads_lock); - th = thread__get(____machine__findnew_thread(machine, pid, tid, false)); + th = ____machine__findnew_thread(machine, pid, tid, false); pthread_rwlock_unlock(&machine->threads_lock); return th; } diff --git a/tools/perf/util/thread.c b/tools/perf/util/thread.c index 0a9ae8014729..dfd00c6dad6e 100644 --- a/tools/perf/util/thread.c +++ b/tools/perf/util/thread.c @@ -19,8 +19,10 @@ int thread__init_map_groups(struct thread *thread, struct machine *machine) thread->mg = map_groups__new(machine); } else { leader = __machine__findnew_thread(machine, pid, pid); - if (leader) + if (leader) { thread->mg = map_groups__get(leader->mg); + thread__put(leader); + } } return thread->mg ? 0 : -1; @@ -53,7 +55,7 @@ struct thread *thread__new(pid_t pid, pid_t tid) goto err_thread; list_add(&comm->list, &thread->comm_list); - atomic_set(&thread->refcnt, 0); + atomic_set(&thread->refcnt, 1); RB_CLEAR_NODE(&thread->rb_node); } @@ -95,6 +97,10 @@ struct thread *thread__get(struct thread *thread) void thread__put(struct thread *thread) { if (thread && atomic_dec_and_test(&thread->refcnt)) { + /* + * Remove it from the dead_threads list, as last reference + * is gone. + */ list_del_init(&thread->node); thread__delete(thread); } -- 2.1.0 -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/