From: Feng Tang <feng.tang@intel.com> To: "Huang, Ying" <ying.huang@intel.com> Cc: Yang Shi <shy828301@gmail.com>, "Hocko, Michal" <mhocko@suse.com>, "Aneesh Kumar K V" <aneesh.kumar@linux.ibm.com>, Andrew Morton <akpm@linux-foundation.org>, Johannes Weiner <hannes@cmpxchg.org>, Tejun Heo <tj@kernel.org>, Zefan Li <lizefan.x@bytedance.com>, Waiman Long <longman@redhat.com>, "linux-mm@kvack.org" <linux-mm@kvack.org>, "cgroups@vger.kernel.org" <cgroups@vger.kernel.org>, "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>, "Hansen, Dave" <dave.hansen@intel.com>, "Chen, Tim C" <tim.c.chen@intel.com>, "Yin, Fengwei" <fengwei.yin@intel.com> Subject: Re: [PATCH] mm/vmscan: respect cpuset policy during page demotion Date: Thu, 27 Oct 2022 15:51:07 +0800 [thread overview] Message-ID: <Y1o4a0HzYTZRArhU@feng-clx> (raw) In-Reply-To: <87k04lk8vr.fsf@yhuang6-desk2.ccr.corp.intel.com> On Thu, Oct 27, 2022 at 03:45:12PM +0800, Huang, Ying wrote: > Feng Tang <feng.tang@intel.com> writes: > > > On Thu, Oct 27, 2022 at 01:57:52AM +0800, Yang Shi wrote: > >> On Wed, Oct 26, 2022 at 8:59 AM Michal Hocko <mhocko@suse.com> wrote: > > [...] > >> > > > This all can get quite expensive so the primary question is, does the > >> > > > existing behavior generates any real issues or is this more of an > >> > > > correctness exercise? I mean it certainly is not great to demote to an > >> > > > incompatible numa node but are there any reasonable configurations when > >> > > > the demotion target node is explicitly excluded from memory > >> > > > policy/cpuset? > >> > > > >> > > We haven't got customer report on this, but there are quite some customers > >> > > use cpuset to bind some specific memory nodes to a docker (You've helped > >> > > us solve a OOM issue in such cases), so I think it's practical to respect > >> > > the cpuset semantics as much as we can. > >> > > >> > Yes, it is definitely better to respect cpusets and all local memory > >> > policies. There is no dispute there. The thing is whether this is really > >> > worth it. How often would cpusets (or policies in general) go actively > >> > against demotion nodes (i.e. exclude those nodes from their allowes node > >> > mask)? > >> > > >> > I can imagine workloads which wouldn't like to get their memory demoted > >> > for some reason but wouldn't it be more practical to tell that > >> > explicitly (e.g. via prctl) rather than configuring cpusets/memory > >> > policies explicitly? > >> > > >> > > Your concern about the expensive cost makes sense! Some raw ideas are: > >> > > * if the shrink_folio_list is called by kswapd, the folios come from > >> > > the same per-memcg lruvec, so only one check is enough > >> > > * if not from kswapd, like called form madvise or DAMON code, we can > >> > > save a memcg cache, and if the next folio's memcg is same as the > >> > > cache, we reuse its result. And due to the locality, the real > >> > > check is rarely performed. > >> > > >> > memcg is not the expensive part of the thing. You need to get from page > >> > -> all vmas::vm_policy -> mm -> task::mempolicy > >> > >> Yeah, on the same page with Michal. Figuring out mempolicy from page > >> seems quite expensive and the correctness can't be guranteed since the > >> mempolicy could be set per-thread and the mm->task depends on > >> CONFIG_MEMCG so it doesn't work for !CONFIG_MEMCG. > > > > Yes, you are right. Our "working" psudo code for mem policy looks like > > what Michal mentioned, and it can't work for all cases, but try to > > enforce it whenever possible: > > > > static bool __check_mpol_demotion(struct folio *folio, struct vm_area_struct *vma, > > unsigned long addr, void *arg) > > { > > bool *skip_demotion = arg; > > struct mempolicy *mpol; > > int nid, dnid; > > bool ret = true; > > > > mpol = __get_vma_policy(vma, addr); > > if (!mpol) { > > struct task_struct *task; > > task = NULL; > > > if (vma->vm_mm) > > task = vma->vm_mm->owner; > > > > if (task) { > > mpol = get_task_policy(task); > > if (mpol) > > mpol_get(mpol); > > } > > } > > > > if (!mpol) > > return ret; > > > > if (mpol->mode != MPOL_BIND) > > goto put_exit; > > > > nid = folio_nid(folio); > > dnid = next_demotion_node(nid); > > if (!node_isset(dnid, mpol->nodes)) { > > *skip_demotion = true; > > ret = false; > > } > > I think that you need to get a node mask instead. Even if > !node_isset(dnid, mpol->nodes), you may demote to other node in the node > mask. Yes, you are right. This code was written/tested about 2 months ago, before Aneesh's memory tiering interface patchset. It was listed to demonstrate idea of solution. Thanks, Feng > Best Regards, > Huang, Ying
WARNING: multiple messages have this Message-ID (diff)
From: Feng Tang <feng.tang-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org> To: "Huang, Ying" <ying.huang-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org> Cc: Yang Shi <shy828301-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>, "Hocko, Michal" <mhocko-IBi9RG/b67k@public.gmane.org>, Aneesh Kumar K V <aneesh.kumar-tEXmvtCZX7AybS5Ee8rs3A@public.gmane.org>, Andrew Morton <akpm-de/tnXTf+JLsfHDXvbKv3WD2FQJk+8+b@public.gmane.org>, Johannes Weiner <hannes-druUgvl0LCNAfugRpC6u6w@public.gmane.org>, Tejun Heo <tj-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org>, Zefan Li <lizefan.x-EC8Uxl6Npydl57MIdRCFDg@public.gmane.org>, Waiman Long <longman-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>, "linux-mm-Bw31MaZKKs3YtjvyW6yDsg@public.gmane.org" <linux-mm-Bw31MaZKKs3YtjvyW6yDsg@public.gmane.org>, "cgroups-u79uwXL29TY76Z2rM5mHXA@public.gmane.org" <cgroups-u79uwXL29TY76Z2rM5mHXA@public.gmane.org>, "linux-kernel-u79uwXL29TY76Z2rM5mHXA@public.gmane.org" <linux-kernel-u79uwXL29TY76Z2rM5mHXA@public.gmane.org>, "Hansen, Dave" <dave.hansen-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>, "Chen, Tim C" <tim.c.chen-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>, "Yin, Fengwei" <fengwei.yin-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org> Subject: Re: [PATCH] mm/vmscan: respect cpuset policy during page demotion Date: Thu, 27 Oct 2022 15:51:07 +0800 [thread overview] Message-ID: <Y1o4a0HzYTZRArhU@feng-clx> (raw) In-Reply-To: <87k04lk8vr.fsf-fFUE1NP8JkzwuUmzmnQr+vooFf0ArEBIu+b9c/7xato@public.gmane.org> On Thu, Oct 27, 2022 at 03:45:12PM +0800, Huang, Ying wrote: > Feng Tang <feng.tang-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org> writes: > > > On Thu, Oct 27, 2022 at 01:57:52AM +0800, Yang Shi wrote: > >> On Wed, Oct 26, 2022 at 8:59 AM Michal Hocko <mhocko-IBi9RG/b67k@public.gmane.org> wrote: > > [...] > >> > > > This all can get quite expensive so the primary question is, does the > >> > > > existing behavior generates any real issues or is this more of an > >> > > > correctness exercise? I mean it certainly is not great to demote to an > >> > > > incompatible numa node but are there any reasonable configurations when > >> > > > the demotion target node is explicitly excluded from memory > >> > > > policy/cpuset? > >> > > > >> > > We haven't got customer report on this, but there are quite some customers > >> > > use cpuset to bind some specific memory nodes to a docker (You've helped > >> > > us solve a OOM issue in such cases), so I think it's practical to respect > >> > > the cpuset semantics as much as we can. > >> > > >> > Yes, it is definitely better to respect cpusets and all local memory > >> > policies. There is no dispute there. The thing is whether this is really > >> > worth it. How often would cpusets (or policies in general) go actively > >> > against demotion nodes (i.e. exclude those nodes from their allowes node > >> > mask)? > >> > > >> > I can imagine workloads which wouldn't like to get their memory demoted > >> > for some reason but wouldn't it be more practical to tell that > >> > explicitly (e.g. via prctl) rather than configuring cpusets/memory > >> > policies explicitly? > >> > > >> > > Your concern about the expensive cost makes sense! Some raw ideas are: > >> > > * if the shrink_folio_list is called by kswapd, the folios come from > >> > > the same per-memcg lruvec, so only one check is enough > >> > > * if not from kswapd, like called form madvise or DAMON code, we can > >> > > save a memcg cache, and if the next folio's memcg is same as the > >> > > cache, we reuse its result. And due to the locality, the real > >> > > check is rarely performed. > >> > > >> > memcg is not the expensive part of the thing. You need to get from page > >> > -> all vmas::vm_policy -> mm -> task::mempolicy > >> > >> Yeah, on the same page with Michal. Figuring out mempolicy from page > >> seems quite expensive and the correctness can't be guranteed since the > >> mempolicy could be set per-thread and the mm->task depends on > >> CONFIG_MEMCG so it doesn't work for !CONFIG_MEMCG. > > > > Yes, you are right. Our "working" psudo code for mem policy looks like > > what Michal mentioned, and it can't work for all cases, but try to > > enforce it whenever possible: > > > > static bool __check_mpol_demotion(struct folio *folio, struct vm_area_struct *vma, > > unsigned long addr, void *arg) > > { > > bool *skip_demotion = arg; > > struct mempolicy *mpol; > > int nid, dnid; > > bool ret = true; > > > > mpol = __get_vma_policy(vma, addr); > > if (!mpol) { > > struct task_struct *task; > > task = NULL; > > > if (vma->vm_mm) > > task = vma->vm_mm->owner; > > > > if (task) { > > mpol = get_task_policy(task); > > if (mpol) > > mpol_get(mpol); > > } > > } > > > > if (!mpol) > > return ret; > > > > if (mpol->mode != MPOL_BIND) > > goto put_exit; > > > > nid = folio_nid(folio); > > dnid = next_demotion_node(nid); > > if (!node_isset(dnid, mpol->nodes)) { > > *skip_demotion = true; > > ret = false; > > } > > I think that you need to get a node mask instead. Even if > !node_isset(dnid, mpol->nodes), you may demote to other node in the node > mask. Yes, you are right. This code was written/tested about 2 months ago, before Aneesh's memory tiering interface patchset. It was listed to demonstrate idea of solution. Thanks, Feng > Best Regards, > Huang, Ying
next prev parent reply other threads:[~2022-10-27 7:51 UTC|newest] Thread overview: 98+ messages / expand[flat|nested] mbox.gz Atom feed top 2022-10-26 7:43 [PATCH] mm/vmscan: respect cpuset policy during page demotion Feng Tang 2022-10-26 7:43 ` Feng Tang 2022-10-26 7:49 ` Aneesh Kumar K V 2022-10-26 7:49 ` Aneesh Kumar K V 2022-10-26 8:00 ` Feng Tang 2022-10-26 8:00 ` Feng Tang 2022-10-26 9:19 ` Michal Hocko 2022-10-26 9:19 ` Michal Hocko 2022-10-26 10:42 ` Aneesh Kumar K V 2022-10-26 10:42 ` Aneesh Kumar K V 2022-10-26 11:02 ` Michal Hocko 2022-10-26 11:02 ` Michal Hocko 2022-10-26 12:08 ` Aneesh Kumar K V 2022-10-26 12:08 ` Aneesh Kumar K V 2022-10-26 12:21 ` Michal Hocko 2022-10-26 12:21 ` Michal Hocko 2022-10-26 12:35 ` Aneesh Kumar K V 2022-10-26 12:35 ` Aneesh Kumar K V 2022-10-27 9:02 ` Michal Hocko 2022-10-27 9:02 ` Michal Hocko 2022-10-27 10:16 ` Aneesh Kumar K V 2022-10-27 10:16 ` Aneesh Kumar K V 2022-10-27 13:05 ` Michal Hocko 2022-10-27 13:05 ` Michal Hocko 2022-10-26 12:20 ` Feng Tang 2022-10-26 12:20 ` Feng Tang 2022-10-26 15:59 ` Michal Hocko 2022-10-26 15:59 ` Michal Hocko 2022-10-26 17:57 ` Yang Shi 2022-10-26 17:57 ` Yang Shi 2022-10-27 7:11 ` Feng Tang 2022-10-27 7:11 ` Feng Tang 2022-10-27 7:45 ` Huang, Ying 2022-10-27 7:45 ` Huang, Ying 2022-10-27 7:51 ` Feng Tang [this message] 2022-10-27 7:51 ` Feng Tang 2022-10-27 17:55 ` Yang Shi 2022-10-27 17:55 ` Yang Shi 2022-10-28 3:37 ` Feng Tang 2022-10-28 3:37 ` Feng Tang 2022-10-28 5:54 ` Huang, Ying 2022-10-28 5:54 ` Huang, Ying 2022-10-28 17:23 ` Yang Shi 2022-10-28 17:23 ` Yang Shi 2022-10-31 1:56 ` Huang, Ying 2022-10-31 1:56 ` Huang, Ying 2022-10-31 2:19 ` Feng Tang 2022-10-31 2:19 ` Feng Tang 2022-10-28 5:09 ` Aneesh Kumar K V 2022-10-28 5:09 ` Aneesh Kumar K V 2022-10-28 17:16 ` Yang Shi 2022-10-28 17:16 ` Yang Shi 2022-10-31 1:53 ` Huang, Ying 2022-10-31 1:53 ` Huang, Ying 2022-10-27 6:47 ` Huang, Ying 2022-10-27 6:47 ` Huang, Ying 2022-10-27 7:10 ` Michal Hocko 2022-10-27 7:10 ` Michal Hocko 2022-10-27 7:39 ` Huang, Ying 2022-10-27 7:39 ` Huang, Ying 2022-10-27 8:01 ` Michal Hocko 2022-10-27 8:01 ` Michal Hocko 2022-10-27 9:31 ` Huang, Ying 2022-10-27 9:31 ` Huang, Ying 2022-10-27 12:29 ` Michal Hocko 2022-10-27 12:29 ` Michal Hocko 2022-10-27 23:22 ` Huang, Ying 2022-10-27 23:22 ` Huang, Ying 2022-10-31 8:40 ` Michal Hocko 2022-10-31 8:40 ` Michal Hocko 2022-10-31 8:51 ` Huang, Ying 2022-10-31 8:51 ` Huang, Ying 2022-10-31 9:18 ` Michal Hocko 2022-10-31 9:18 ` Michal Hocko 2022-10-31 14:09 ` Feng Tang 2022-10-31 14:09 ` Feng Tang 2022-10-31 14:32 ` Michal Hocko 2022-10-31 14:32 ` Michal Hocko 2022-11-07 8:05 ` Feng Tang 2022-11-07 8:05 ` Feng Tang 2022-11-07 8:17 ` Michal Hocko 2022-11-07 8:17 ` Michal Hocko 2022-11-01 3:17 ` Huang, Ying 2022-11-01 3:17 ` Huang, Ying 2022-10-26 8:26 ` Yin, Fengwei 2022-10-26 8:26 ` Yin, Fengwei 2022-10-26 8:37 ` Feng Tang 2022-10-26 8:37 ` Feng Tang 2022-10-26 14:36 ` Waiman Long 2022-10-26 14:36 ` Waiman Long 2022-10-27 5:57 ` Feng Tang 2022-10-27 5:57 ` Feng Tang 2022-10-27 5:13 ` Huang, Ying 2022-10-27 5:13 ` Huang, Ying 2022-10-27 5:49 ` Feng Tang 2022-10-27 5:49 ` Feng Tang 2022-10-27 6:05 ` Huang, Ying 2022-10-27 6:05 ` Huang, Ying
Reply instructions: You may reply publicly to this message via plain-text email using any one of the following methods: * Save the following mbox file, import it into your mail client, and reply-to-all from there: mbox Avoid top-posting and favor interleaved quoting: https://en.wikipedia.org/wiki/Posting_style#Interleaved_style * Reply using the --to, --cc, and --in-reply-to switches of git-send-email(1): git send-email \ --in-reply-to=Y1o4a0HzYTZRArhU@feng-clx \ --to=feng.tang@intel.com \ --cc=akpm@linux-foundation.org \ --cc=aneesh.kumar@linux.ibm.com \ --cc=cgroups@vger.kernel.org \ --cc=dave.hansen@intel.com \ --cc=fengwei.yin@intel.com \ --cc=hannes@cmpxchg.org \ --cc=linux-kernel@vger.kernel.org \ --cc=linux-mm@kvack.org \ --cc=lizefan.x@bytedance.com \ --cc=longman@redhat.com \ --cc=mhocko@suse.com \ --cc=shy828301@gmail.com \ --cc=tim.c.chen@intel.com \ --cc=tj@kernel.org \ --cc=ying.huang@intel.com \ /path/to/YOUR_REPLY https://kernel.org/pub/software/scm/git/docs/git-send-email.html * If your mail client supports setting the In-Reply-To header via mailto: links, try the mailto: linkBe sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes, see mirroring instructions on how to clone and mirror all data and code used by this external index.