From mboxrd@z Thu Jan 1 00:00:00 1970 From: Quanyang Wang Subject: Re: [V2][PATCH] cgroup: fix memory leak caused by missing cgroup_bpf_offline Date: Tue, 19 Oct 2021 18:41:14 +0800 Message-ID: References: <20211018075623.26884-1-quanyang.wang@windriver.com> <8fdcaded-474e-139b-a9bc-5ab6f91fbd4f@windriver.com> Mime-Version: 1.0 Content-Transfer-Encoding: 7bit Return-path: DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=windriver.com; h=subject : to : cc : references : from : message-id : date : in-reply-to : content-type : content-transfer-encoding : mime-version; s=PPS06212021; bh=R9WzwAsfHJdXo2agfuKGPT29kGt1urdRgBEDxIyXQEg=; b=hJ1OC15jYdezq5aK/K9oYrIdRaUjQose+7JFuKLgKmuOhuyerZQ9UQK+UsCP3ZgVJ2I7 HdLaA+m00IzjL2DwsPEeW5pLQ9EYldO3G5GOH7QUJp++6Li1fYNsOQ5hiIjiMDAQoM0g zxITxhKRmcB4vHtVPBwdbXeGAmZZNo+pXw8eu8r6EjPaIbRnqn5gCgYz3R3fF9vCQP+S cT4jwcRDrJjxazozDXphPMegIZ3+7AG3/WW5NWc+mwkbXVaKSKADr+eWuPAQKuyP5quz dZY/uIuZzeiuDHxiZNdzY09iF3LWWe/m2tYqWH5Sb+8jCDAW/NBKBCncDqDegAwoGfhq lg== In-Reply-To: Content-Language: en-US List-ID: Content-Type: text/plain; charset="us-ascii"; format="flowed" To: Ming Lei Cc: Tejun Heo , Zefan Li , Johannes Weiner , Alexei Starovoitov , Daniel Borkmann , Andrii Nakryiko , Jens Axboe , Roman Gushchin , mkoutny-IBi9RG/b67k@public.gmane.org, cgroups-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, linux-kernel-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, netdev-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, bpf-u79uwXL29TY76Z2rM5mHXA@public.gmane.org Hi Ming, On 10/18/21 8:59 PM, Ming Lei wrote: > On Mon, Oct 18, 2021 at 06:06:28PM +0800, Quanyang Wang wrote: >> Hi Ming, >> >> On 10/18/21 5:02 PM, Ming Lei wrote: >>> On Mon, Oct 18, 2021 at 03:56:23PM +0800, quanyang.wang-CWA4WttNNZF54TAoqtyWWQ@public.gmane.org wrote: >>>> From: Quanyang Wang >>>> >>>> When enabling CONFIG_CGROUP_BPF, kmemleak can be observed by running >>>> the command as below: >>>> >>>> $mount -t cgroup -o none,name=foo cgroup cgroup/ >>>> $umount cgroup/ >>>> >>>> unreferenced object 0xc3585c40 (size 64): >>>> comm "mount", pid 425, jiffies 4294959825 (age 31.990s) >>>> hex dump (first 32 bytes): >>>> 01 00 00 80 84 8c 28 c0 00 00 00 00 00 00 00 00 ......(......... >>>> 00 00 00 00 00 00 00 00 6c 43 a0 c3 00 00 00 00 ........lC...... >>>> backtrace: >>>> [] cgroup_bpf_inherit+0x44/0x24c >>>> [<1f03679c>] cgroup_setup_root+0x174/0x37c >>>> [] cgroup1_get_tree+0x2c0/0x4a0 >>>> [] vfs_get_tree+0x24/0x108 >>>> [] path_mount+0x384/0x988 >>>> [] do_mount+0x64/0x9c >>>> [<208c9cfe>] sys_mount+0xfc/0x1f4 >>>> [<06dd06e0>] ret_fast_syscall+0x0/0x48 >>>> [] 0xbeb4daa8 >>>> >>>> This is because that since the commit 2b0d3d3e4fcf ("percpu_ref: reduce >>>> memory footprint of percpu_ref in fast path") root_cgrp->bpf.refcnt.data >>>> is allocated by the function percpu_ref_init in cgroup_bpf_inherit which >>>> is called by cgroup_setup_root when mounting, but not freed along with >>>> root_cgrp when umounting. Adding cgroup_bpf_offline which calls >>>> percpu_ref_kill to cgroup_kill_sb can free root_cgrp->bpf.refcnt.data in >>>> umount path. >>>> >>>> This patch also fixes the commit 4bfc0bb2c60e ("bpf: decouple the lifetime >>>> of cgroup_bpf from cgroup itself"). A cgroup_bpf_offline is needed to do a >>>> cleanup that frees the resources which are allocated by cgroup_bpf_inherit >>>> in cgroup_setup_root. >>>> >>>> And inside cgroup_bpf_offline, cgroup_get() is at the beginning and >>>> cgroup_put is at the end of cgroup_bpf_release which is called by >>>> cgroup_bpf_offline. So cgroup_bpf_offline can keep the balance of >>>> cgroup's refcount. >>>> >>>> Fixes: 2b0d3d3e4fcf ("percpu_ref: reduce memory footprint of percpu_ref in fast path") >>> >>> If I understand correctly, cgroup_bpf_release() won't be called without >>> your patch. So anything allocated in cgroup_bpf_inherit() will be >>> leaked? >> No, for now cgroup_bpf_release is called to release bpf.refcnt.data of the >> cgroup which is not root_cgroup. Only root_cgroup's bpf data is leaked. > > You mean that cgroup_bpf_inherit() allocates nothing for root_cgroup? Yes, cgroup_bpf_inherit allocates something for root_cgroup. The earlier commit 4bfc0bb2c60e ("bpf: decouple the lifetime of cgroup_bpf from cgroup itself") introduces an imbalance that call cgroup_bpf_inherit(&root_cgroup) but not call cgroup_bpf_offline(&root_cgroup). But there was no memory leak here. When the commit 2b0d3d3e4fcf ("percpu_ref: reduce memory footprint of percpu_ref in fast path") applies, some data is allocated for root_cgroup and not released with root_cgroup, and memory leak is observed. So I add 2 "Fixes tags" here to indicate that 2 commits introduce two different issues. But it seems that 2 "Fixes tags" is misleading now. So maybe just fix earlier commit 4bfc0bb2c60e which introduces imbalance? Thanks, Quanyang > > If yes, I agree you can add 'Fixes: 2b0d3d3e4fcf', otherwise please > remove it. > > > Thanks, > Ming >