From: Mel Gorman <mgorman@techsingularity.net>
To: Vasily Averin <vvs@virtuozzo.com>
Cc: Michal Hocko <mhocko@kernel.org>,
Johannes Weiner <hannes@cmpxchg.org>,
Vladimir Davydov <vdavydov.dev@gmail.com>,
Andrew Morton <akpm@linux-foundation.org>,
Shakeel Butt <shakeelb@google.com>, Roman Gushchin <guro@fb.com>,
Uladzislau Rezki <urezki@gmail.com>,
Vlastimil Babka <vbabka@suse.cz>,
cgroups@vger.kernel.org, linux-mm@kvack.org,
linux-kernel@vger.kernel.org, kernel@openvz.org
Subject: Re: [PATCH mm v2] memcg: enable memory accounting in __alloc_pages_bulk
Date: Tue, 12 Oct 2021 14:10:19 +0100 [thread overview]
Message-ID: <20211012131019.GV3959@techsingularity.net> (raw)
In-Reply-To: <2410e99a-087c-3f89-9bdf-b62a7d5df725@virtuozzo.com>
On Tue, Oct 12, 2021 at 01:18:39PM +0300, Vasily Averin wrote:
> Enable memory accounting for bulk page allocator.
>
> Fixes: 387ba26fb1cb ("mm/page_alloc: add a bulk page allocator")
> Cc: <stable@vger.kernel.org>
> Signed-off-by: Vasily Averin <vvs@virtuozzo.com>
> ---
> v2: modified according to Shakeel Butt's remarks
> ---
> include/linux/memcontrol.h | 11 +++++++++
> mm/memcontrol.c | 48 +++++++++++++++++++++++++++++++++++++-
> mm/page_alloc.c | 14 ++++++++++-
> 3 files changed, 71 insertions(+), 2 deletions(-)
>
> diff --git a/include/linux/memcontrol.h b/include/linux/memcontrol.h
> index 3096c9a0ee01..990acd70c846 100644
> --- a/include/linux/memcontrol.h
> +++ b/include/linux/memcontrol.h
> @@ -810,6 +810,12 @@ static inline void obj_cgroup_put(struct obj_cgroup *objcg)
> percpu_ref_put(&objcg->refcnt);
> }
>
> +static inline void obj_cgroup_put_many(struct obj_cgroup *objcg,
> + unsigned long nr)
> +{
> + percpu_ref_put_many(&objcg->refcnt, nr);
> +}
> +
> static inline void mem_cgroup_put(struct mem_cgroup *memcg)
> {
> if (memcg)
> @@ -1746,4 +1752,9 @@ static inline struct mem_cgroup *mem_cgroup_from_obj(void *p)
>
> #endif /* CONFIG_MEMCG_KMEM */
>
> +bool memcg_bulk_pre_charge_hook(struct obj_cgroup **objcgp, gfp_t gfp,
> + unsigned int nr_pages);
> +void memcg_bulk_charge_hook(struct obj_cgroup *objcgp, struct page *page);
> +void memcg_bulk_post_charge_hook(struct obj_cgroup *objcg,
> + unsigned int nr_pages);
> #endif /* _LINUX_MEMCONTROL_H */
> diff --git a/mm/memcontrol.c b/mm/memcontrol.c
> index 87e41c3cac10..16fe3384c12c 100644
> --- a/mm/memcontrol.c
> +++ b/mm/memcontrol.c
> @@ -3239,7 +3239,53 @@ void obj_cgroup_uncharge(struct obj_cgroup *objcg, size_t size)
> refill_obj_stock(objcg, size, true);
> }
>
> -#endif /* CONFIG_MEMCG_KMEM */
> +bool memcg_bulk_pre_charge_hook(struct obj_cgroup **objcgp, gfp_t gfp,
> + unsigned int nr_pages)
> +{
> + struct obj_cgroup *objcg = NULL;
> +
> + if (!memcg_kmem_enabled() || !(gfp & __GFP_ACCOUNT))
> + return true;
> +
> + objcg = get_obj_cgroup_from_current();
> +
> + if (objcg && obj_cgroup_charge_pages(objcg, gfp, nr_pages)) {
> + obj_cgroup_put(objcg);
> + return false;
> + }
> + obj_cgroup_get_many(objcg, nr_pages - 1);
> + *objcgp = objcg;
> + return true;
> +}
> +
This is probably a stupid question but why is it necessary to get many
references instead of taking one reference here and dropping it in
memcg_bulk_post_charge_hook?
> +void memcg_bulk_charge_hook(struct obj_cgroup *objcg, struct page *page)
> +{
> + page->memcg_data = (unsigned long)objcg | MEMCG_DATA_KMEM;
> +}
> +
> +void memcg_bulk_post_charge_hook(struct obj_cgroup *objcg,
> + unsigned int nr_pages)
> +{
> + obj_cgroup_uncharge_pages(objcg, nr_pages);
> + obj_cgroup_put_many(objcg, nr_pages);
> +}
And are you sure obj_cgroup_uncharge_pages should be called here? I
thought the pages get uncharged on free.
> +#else /* !CONFIG_MEMCG_KMEM */
> +bool memcg_bulk_pre_charge_hook(struct obj_cgroup **objcgp, gfp_t gfp,
> + unsigned int nr_pages)
> +{
> + return true;
> +}
> +
> +void memcg_bulk_charge_hook(struct obj_cgroup *objcgp, struct page *page)
> +{
> +}
> +
> +void memcg_bulk_post_charge_hook(struct obj_cgroup *objcg,
> + unsigned int nr_pages)
> +{
> +}
> +#endif
> +
>
> /*
> * Because page_memcg(head) is not set on tails, set it now.
> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> index b37435c274cf..eb37177bf507 100644
> --- a/mm/page_alloc.c
> +++ b/mm/page_alloc.c
> @@ -5207,6 +5207,8 @@ unsigned long __alloc_pages_bulk(gfp_t gfp, int preferred_nid,
> gfp_t alloc_gfp;
> unsigned int alloc_flags = ALLOC_WMARK_LOW;
> int nr_populated = 0, nr_account = 0;
> + unsigned int nr_pre_charge = 0;
> + struct obj_cgroup *objcg = NULL;
>
> /*
> * Skip populated array elements to determine if any pages need
> @@ -5275,6 +5277,10 @@ unsigned long __alloc_pages_bulk(gfp_t gfp, int preferred_nid,
> if (unlikely(!zone))
> goto failed;
>
> + nr_pre_charge = nr_pages - nr_populated;
> + if (!memcg_bulk_pre_charge_hook(&objcg, gfp, nr_pre_charge))
> + goto failed;
> +
> /* Attempt the batch allocation */
> local_lock_irqsave(&pagesets.lock, flags);
> pcp = this_cpu_ptr(zone->per_cpu_pageset);
> @@ -5299,6 +5305,9 @@ unsigned long __alloc_pages_bulk(gfp_t gfp, int preferred_nid,
> nr_account++;
>
> prep_new_page(page, 0, gfp, 0);
> + if (objcg)
> + memcg_bulk_charge_hook(objcg, page);
> +
> if (page_list)
> list_add(&page->lru, page_list);
> else
> @@ -5310,13 +5319,16 @@ unsigned long __alloc_pages_bulk(gfp_t gfp, int preferred_nid,
>
> __count_zid_vm_events(PGALLOC, zone_idx(zone), nr_account);
> zone_statistics(ac.preferred_zoneref->zone, zone, nr_account);
> + if (objcg)
> + memcg_bulk_post_charge_hook(objcg, nr_pre_charge - nr_account);
>
> out:
> return nr_populated;
>
> failed_irq:
> local_unlock_irqrestore(&pagesets.lock, flags);
> -
> + if (objcg)
> + memcg_bulk_post_charge_hook(objcg, nr_pre_charge);
> failed:
> page = __alloc_pages(gfp, 0, preferred_nid, nodemask);
> if (page) {
> --
> 2.31.1
>
--
Mel Gorman
SUSE Labs
next prev parent reply other threads:[~2021-10-12 13:10 UTC|newest]
Thread overview: 28+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-10-07 8:04 memcg memory accounting in vmalloc is broken Vasily Averin
2021-10-07 8:13 ` Michal Hocko
2021-10-07 8:16 ` Michal Hocko
2021-10-07 8:50 ` Vasily Averin
2021-10-07 10:08 ` Michal Hocko
2021-10-07 10:20 ` Mel Gorman
2021-10-07 14:02 ` Vlastimil Babka
2021-10-07 14:00 ` Vlastimil Babka
2021-10-07 14:09 ` Michal Hocko
2021-10-07 19:33 ` Vasily Averin
2021-10-08 9:23 ` [PATCH memcg] memcg: enable memory accounting in __alloc_pages_bulk Vasily Averin
2021-10-08 17:35 ` Shakeel Butt
2021-10-12 10:18 ` [PATCH mm v2] " Vasily Averin
2021-10-12 13:10 ` Mel Gorman [this message]
2021-10-12 13:40 ` Vasily Averin
2021-10-12 14:58 ` [PATCH mm v3] " Vasily Averin
2021-10-12 15:19 ` Shakeel Butt
2021-10-12 15:20 ` Mel Gorman
2021-10-12 15:36 ` Michal Hocko
2021-10-12 16:08 ` Shakeel Butt
2021-10-12 18:24 ` Michal Hocko
2021-10-13 16:41 ` Shakeel Butt
2021-10-13 17:16 ` Michal Hocko
2021-10-13 17:30 ` Shakeel Butt
2021-10-12 18:45 ` Vasily Averin
2021-10-14 8:02 ` [PATCH mm v5] " Vasily Averin
2021-10-15 21:34 ` Andrew Morton
2021-10-16 6:04 ` Vasily Averin
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=20211012131019.GV3959@techsingularity.net \
--to=mgorman@techsingularity.net \
--cc=akpm@linux-foundation.org \
--cc=cgroups@vger.kernel.org \
--cc=guro@fb.com \
--cc=hannes@cmpxchg.org \
--cc=kernel@openvz.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=mhocko@kernel.org \
--cc=shakeelb@google.com \
--cc=urezki@gmail.com \
--cc=vbabka@suse.cz \
--cc=vdavydov.dev@gmail.com \
--cc=vvs@virtuozzo.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: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).