From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by smtp.lore.kernel.org (Postfix) with ESMTP id BAFE2C7EE22 for ; Tue, 2 May 2023 19:56:54 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S229722AbjEBT4w (ORCPT ); Tue, 2 May 2023 15:56:52 -0400 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:33782 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S229581AbjEBT4u (ORCPT ); Tue, 2 May 2023 15:56:50 -0400 Received: from us-smtp-delivery-124.mimecast.com (us-smtp-delivery-124.mimecast.com [170.10.129.124]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id 2973B19A7 for ; Tue, 2 May 2023 12:56:08 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=redhat.com; s=mimecast20190719; t=1683057367; h=from:from:reply-to:subject:subject:date:date:message-id:message-id: to:to:cc:cc:mime-version:mime-version:content-type:content-type: content-transfer-encoding:content-transfer-encoding: in-reply-to:in-reply-to:references:references; bh=iMs73NusrIc3WnWcdl7Fx0I1f6axuQqjUOBlJwm9l/c=; b=HHjTDNPbnDC4AFvS43aRgTee73bw2RWLhk5yANez+r0RAFMrlgRTFoTgDZamwISd5C3Cwc W0sZt2kjd8k7B3A1NHwWpbvoGiq+17dmvmc+309m/pHP0JFa24Ee+MvW3aoKIJ2/u0j1JV Z/qfvOxQgZwFKOhjmEUSzPAz0jvhG7E= Received: from mimecast-mx02.redhat.com (mx3-rdu2.redhat.com [66.187.233.73]) by relay.mimecast.com with ESMTP with STARTTLS (version=TLSv1.2, cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id us-mta-563-pHBg26pUOea2cWuITsTWXQ-1; Tue, 02 May 2023 15:56:03 -0400 X-MC-Unique: pHBg26pUOea2cWuITsTWXQ-1 Received: from smtp.corp.redhat.com (int-mx06.intmail.prod.int.rdu2.redhat.com [10.11.54.6]) (using TLSv1.2 with cipher AECDH-AES256-SHA (256/256 bits)) (No client certificate requested) by mimecast-mx02.redhat.com (Postfix) with ESMTPS id 65A3F3815F69; Tue, 2 May 2023 19:56:02 +0000 (UTC) Received: from [10.22.10.239] (unknown [10.22.10.239]) by smtp.corp.redhat.com (Postfix) with ESMTP id CEBEB2166B26; Tue, 2 May 2023 19:56:01 +0000 (UTC) Message-ID: <9e149289-c1c0-3297-145d-ad3a890056ac@redhat.com> Date: Tue, 2 May 2023 15:56:01 -0400 MIME-Version: 1.0 User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:102.0) Gecko/20100101 Thunderbird/102.7.1 Subject: Re: [RFC PATCH 3/3] cgroup: Do not take css_set_lock in cgroup_show_path Content-Language: en-US To: =?UTF-8?Q?Michal_Koutn=c3=bd?= , linux-fsdevel@vger.kernel.org, linux-kernel@vger.kernel.org, cgroups@vger.kernel.org Cc: Alexander Viro , Christian Brauner , Tejun Heo , Zefan Li , Johannes Weiner , Dave Chinner , Rik van Riel , Jiri Wiesner References: <20230502133847.14570-1-mkoutny@suse.com> <20230502133847.14570-4-mkoutny@suse.com> From: Waiman Long In-Reply-To: <20230502133847.14570-4-mkoutny@suse.com> Content-Type: text/plain; charset=UTF-8; format=flowed Content-Transfer-Encoding: 8bit X-Scanned-By: MIMEDefang 3.1 on 10.11.54.6 Precedence: bulk List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On 5/2/23 09:38, Michal Koutný wrote: > /proc/$pid/mountinfo may accumulate lots of entries (causing frequent > re-reads of whole file) or lots cgroupfs entries alone. > The cgroupfs entries rendered with cgroup_show_path() may increase/be > subject of css_set_lock contention causing further slowdown -- not only > mountinfo rendering but any other css_set_lock user. > > We leverage the fact that mountinfo reading happens with namespace_sem > taken and hierarchy roots thus cannot be freed concurrently. > > There are three relevant nodes for each cgroupfs entry: > > R ... cgroup hierarchy root > M ... mount root > C ... reader's cgroup NS root > > mountinfo is supposed to show path from C to M. > > Current's css_set (and linked root cgroups) are stable under > namespace_sem, therefore current_cgns_cgroup_from_root() doesn't need > css_set_lock. > > When the path is assembled in kernfs_path_from_node(), we know that: > - C kernfs_node is (transitively) pinned via current->nsproxy, > - M kernfs_node is pinned thanks to namespace_sem, > - path C to M is pinned via child->parent references (this holds also > when C and M are in distinct subtrees). > > Streamline mountinfo rendering a bit by relieving css_set_lock and add > careful notes about that. > > Signed-off-by: Michal Koutný > --- > kernel/cgroup/cgroup.c | 12 +++++++----- > 1 file changed, 7 insertions(+), 5 deletions(-) > > diff --git a/kernel/cgroup/cgroup.c b/kernel/cgroup/cgroup.c > index 32d693a797b9..e2ec6f0028be 100644 > --- a/kernel/cgroup/cgroup.c > +++ b/kernel/cgroup/cgroup.c > @@ -1407,12 +1407,18 @@ static inline struct cgroup *__cset_cgroup_from_root(struct css_set *cset, > struct cgroup *res_cgroup = NULL; > > if (cset == &init_css_set) { > + /* callers must ensure root stability */ > res_cgroup = &root->cgrp; > } else if (root == &cgrp_dfl_root) { > res_cgroup = cset->dfl_cgrp; > } else { > struct cgrp_cset_link *link; > - lockdep_assert_held(&css_set_lock); > + /* cset's cgroups are pinned unless they are root cgroups that > + * were unmounted. We look at links to !cgrp_dfl_root > + * cgroup_root, either lock ensures the list is not mutated > + */ > + lockdep_assert(lockdep_is_held(&css_set_lock) || > + lockdep_is_held_type(&namespace_sem, -1)); Again lockdep_is_held(&namespace_sem) is good enough. > > list_for_each_entry(link, &cset->cgrp_links, cgrp_link) { > struct cgroup *c = link->cgrp; > @@ -1438,8 +1444,6 @@ current_cgns_cgroup_from_root(struct cgroup_root *root) > struct cgroup *res = NULL; > struct css_set *cset; > > - lockdep_assert_held(&css_set_lock); > - > /* namespace_sem ensures `root` stability on unmount */ > lockdep_assert(lockdep_is_held_type(&namespace_sem, -1)); > > @@ -1905,10 +1909,8 @@ int cgroup_show_path(struct seq_file *sf, struct kernfs_node *kf_node, > if (!buf) > return -ENOMEM; > > - spin_lock_irq(&css_set_lock); > ns_cgroup = current_cgns_cgroup_from_root(kf_cgroot); > len = kernfs_path_from_node(kf_node, ns_cgroup->kn, buf, PATH_MAX); > - spin_unlock_irq(&css_set_lock); > > if (len >= PATH_MAX) > len = -ERANGE; Cheers, Longman From mboxrd@z Thu Jan 1 00:00:00 1970 From: Waiman Long Subject: Re: [RFC PATCH 3/3] cgroup: Do not take css_set_lock in cgroup_show_path Date: Tue, 2 May 2023 15:56:01 -0400 Message-ID: <9e149289-c1c0-3297-145d-ad3a890056ac@redhat.com> References: <20230502133847.14570-1-mkoutny@suse.com> <20230502133847.14570-4-mkoutny@suse.com> Mime-Version: 1.0 Content-Transfer-Encoding: 8bit Return-path: DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=redhat.com; s=mimecast20190719; t=1683057367; h=from:from:reply-to:subject:subject:date:date:message-id:message-id: to:to:cc:cc:mime-version:mime-version:content-type:content-type: content-transfer-encoding:content-transfer-encoding: in-reply-to:in-reply-to:references:references; bh=iMs73NusrIc3WnWcdl7Fx0I1f6axuQqjUOBlJwm9l/c=; b=HHjTDNPbnDC4AFvS43aRgTee73bw2RWLhk5yANez+r0RAFMrlgRTFoTgDZamwISd5C3Cwc W0sZt2kjd8k7B3A1NHwWpbvoGiq+17dmvmc+309m/pHP0JFa24Ee+MvW3aoKIJ2/u0j1JV Z/qfvOxQgZwFKOhjmEUSzPAz0jvhG7E= Content-Language: en-US In-Reply-To: <20230502133847.14570-4-mkoutny-IBi9RG/b67k@public.gmane.org> List-ID: Content-Type: text/plain; charset="utf-8"; format="flowed" To: =?UTF-8?Q?Michal_Koutn=c3=bd?= , linux-fsdevel-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, linux-kernel-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, cgroups-u79uwXL29TY76Z2rM5mHXA@public.gmane.org Cc: Alexander Viro , Christian Brauner , Tejun Heo , Zefan Li , Johannes Weiner , Dave Chinner , Rik van Riel , Jiri Wiesner On 5/2/23 09:38, Michal Koutný wrote: > /proc/$pid/mountinfo may accumulate lots of entries (causing frequent > re-reads of whole file) or lots cgroupfs entries alone. > The cgroupfs entries rendered with cgroup_show_path() may increase/be > subject of css_set_lock contention causing further slowdown -- not only > mountinfo rendering but any other css_set_lock user. > > We leverage the fact that mountinfo reading happens with namespace_sem > taken and hierarchy roots thus cannot be freed concurrently. > > There are three relevant nodes for each cgroupfs entry: > > R ... cgroup hierarchy root > M ... mount root > C ... reader's cgroup NS root > > mountinfo is supposed to show path from C to M. > > Current's css_set (and linked root cgroups) are stable under > namespace_sem, therefore current_cgns_cgroup_from_root() doesn't need > css_set_lock. > > When the path is assembled in kernfs_path_from_node(), we know that: > - C kernfs_node is (transitively) pinned via current->nsproxy, > - M kernfs_node is pinned thanks to namespace_sem, > - path C to M is pinned via child->parent references (this holds also > when C and M are in distinct subtrees). > > Streamline mountinfo rendering a bit by relieving css_set_lock and add > careful notes about that. > > Signed-off-by: Michal Koutný > --- > kernel/cgroup/cgroup.c | 12 +++++++----- > 1 file changed, 7 insertions(+), 5 deletions(-) > > diff --git a/kernel/cgroup/cgroup.c b/kernel/cgroup/cgroup.c > index 32d693a797b9..e2ec6f0028be 100644 > --- a/kernel/cgroup/cgroup.c > +++ b/kernel/cgroup/cgroup.c > @@ -1407,12 +1407,18 @@ static inline struct cgroup *__cset_cgroup_from_root(struct css_set *cset, > struct cgroup *res_cgroup = NULL; > > if (cset == &init_css_set) { > + /* callers must ensure root stability */ > res_cgroup = &root->cgrp; > } else if (root == &cgrp_dfl_root) { > res_cgroup = cset->dfl_cgrp; > } else { > struct cgrp_cset_link *link; > - lockdep_assert_held(&css_set_lock); > + /* cset's cgroups are pinned unless they are root cgroups that > + * were unmounted. We look at links to !cgrp_dfl_root > + * cgroup_root, either lock ensures the list is not mutated > + */ > + lockdep_assert(lockdep_is_held(&css_set_lock) || > + lockdep_is_held_type(&namespace_sem, -1)); Again lockdep_is_held(&namespace_sem) is good enough. > > list_for_each_entry(link, &cset->cgrp_links, cgrp_link) { > struct cgroup *c = link->cgrp; > @@ -1438,8 +1444,6 @@ current_cgns_cgroup_from_root(struct cgroup_root *root) > struct cgroup *res = NULL; > struct css_set *cset; > > - lockdep_assert_held(&css_set_lock); > - > /* namespace_sem ensures `root` stability on unmount */ > lockdep_assert(lockdep_is_held_type(&namespace_sem, -1)); > > @@ -1905,10 +1909,8 @@ int cgroup_show_path(struct seq_file *sf, struct kernfs_node *kf_node, > if (!buf) > return -ENOMEM; > > - spin_lock_irq(&css_set_lock); > ns_cgroup = current_cgns_cgroup_from_root(kf_cgroot); > len = kernfs_path_from_node(kf_node, ns_cgroup->kn, buf, PATH_MAX); > - spin_unlock_irq(&css_set_lock); > > if (len >= PATH_MAX) > len = -ERANGE; Cheers, Longman