From: Muchun Song <songmuchun@bytedance.com>
To: Vlastimil Babka <vbabka@suse.cz>
Cc: cl@linux.com, penberg@kernel.org, rientjes@google.com,
iamjoonsoo.kim@lge.com,
Andrew Morton <akpm@linux-foundation.org>,
linux-mm@kvack.org, linux-kernel@vger.kernel.org,
Roman Gushchin <guro@fb.com>
Subject: Re: [External] Re: [PATCH] mm/slab: Add a __GFP_ACCOUNT GFP flag check for slab allocation
Date: Mon, 15 Jun 2020 21:32:50 +0800 [thread overview]
Message-ID: <CAMZfGtUeUyi5Tg0hv3j=-scFGiZJB5hQs8v5nOfYQ7xXsB2nsQ@mail.gmail.com> (raw)
In-Reply-To: <bca5f44c-5e30-17b1-09fc-6330d450433b@suse.cz>
On Mon, Jun 15, 2020 at 9:08 PM Vlastimil Babka <vbabka@suse.cz> wrote:
>
> On 6/14/20 8:38 AM, Muchun Song wrote:
> > When a kmem_cache is initialized with SLAB_ACCOUNT slab flag, we must
> > not call kmem_cache_alloc with __GFP_ACCOUNT GFP flag. In this case,
> > we can be accounted to kmemcg twice. This is not correct. So we add a
>
> Are you sure? How does that happen?
>
> The only place I see these evaluated is this condition in slab_pre_alloc_hook():
>
> if (memcg_kmem_enabled() &&
> ((flags & __GFP_ACCOUNT) || (s->flags & SLAB_ACCOUNT)))
> return memcg_kmem_get_cache(s);
>
> And it doesn't matter if one or both are set? Am I missing something?
>
> > __GFP_ACCOUNT GFP flag check for slab allocation.
> >
> > We also introduce a new helper named fixup_gfp_flags to do that check.
> > We can reuse the fixup_gfp_flags for SLAB/SLUB.
> >
> > Signed-off-by: Muchun Song <songmuchun@bytedance.com>
> > ---
> > mm/slab.c | 10 +---------
> > mm/slab.h | 21 +++++++++++++++++++++
> > mm/slub.c | 10 +---------
> > 3 files changed, 23 insertions(+), 18 deletions(-)
> >
> > diff --git a/mm/slab.c b/mm/slab.c
> > index 9350062ffc1a..6e0110bef2d6 100644
> > --- a/mm/slab.c
> > +++ b/mm/slab.c
> > @@ -126,8 +126,6 @@
> >
> > #include <trace/events/kmem.h>
> >
> > -#include "internal.h"
> > -
> > #include "slab.h"
> >
> > /*
> > @@ -2579,13 +2577,7 @@ static struct page *cache_grow_begin(struct kmem_cache *cachep,
> > * Be lazy and only check for valid flags here, keeping it out of the
> > * critical path in kmem_cache_alloc().
> > */
> > - if (unlikely(flags & GFP_SLAB_BUG_MASK)) {
> > - gfp_t invalid_mask = flags & GFP_SLAB_BUG_MASK;
> > - flags &= ~GFP_SLAB_BUG_MASK;
> > - pr_warn("Unexpected gfp: %#x (%pGg). Fixing up to gfp: %#x (%pGg). Fix your code!\n",
> > - invalid_mask, &invalid_mask, flags, &flags);
> > - dump_stack();
> > - }
> > + flags = fixup_gfp_flags(cachep, flags);
> > WARN_ON_ONCE(cachep->ctor && (flags & __GFP_ZERO));
> > local_flags = flags & (GFP_CONSTRAINT_MASK|GFP_RECLAIM_MASK);
> >
> > diff --git a/mm/slab.h b/mm/slab.h
> > index 815e4e9a94cd..0b91f2a7b033 100644
> > --- a/mm/slab.h
> > +++ b/mm/slab.h
> > @@ -109,6 +109,7 @@ struct memcg_cache_params {
> > #include <linux/kmemleak.h>
> > #include <linux/random.h>
> > #include <linux/sched/mm.h>
> > +#include "internal.h"
> >
> > /*
> > * State of the slab allocator.
> > @@ -627,6 +628,26 @@ struct kmem_cache_node {
> >
> > };
> >
> > +static inline gfp_t fixup_gfp_flags(struct kmem_cache *s, gfp_t flags)
> > +{
> > + gfp_t invalid_mask = 0;
> > +
> > + if (unlikely(flags & GFP_SLAB_BUG_MASK))
> > + invalid_mask |= flags & GFP_SLAB_BUG_MASK;
> > +
> > + if (unlikely(flags & __GFP_ACCOUNT && s->flags & SLAB_ACCOUNT))
> > + invalid_mask |= __GFP_ACCOUNT;
> > +
> > + if (unlikely(invalid_mask)) {
> > + flags &= ~invalid_mask;
> > + pr_warn("Unexpected gfp: %#x (%pGg). Fixing up to gfp: %#x (%pGg). Fix your code!\n",
> > + invalid_mask, &invalid_mask, flags, &flags);
> > + dump_stack();
> > + }
> > +
> > + return flags;
> > +}
> > +
> > static inline struct kmem_cache_node *get_node(struct kmem_cache *s, int node)
> > {
> > return s->node[node];
> > diff --git a/mm/slub.c b/mm/slub.c
> > index b8f798b50d44..49b5cb7da318 100644
> > --- a/mm/slub.c
> > +++ b/mm/slub.c
> > @@ -37,8 +37,6 @@
> >
> > #include <trace/events/kmem.h>
> >
> > -#include "internal.h"
> > -
> > /*
> > * Lock order:
> > * 1. slab_mutex (Global Mutex)
> > @@ -1745,13 +1743,7 @@ static struct page *allocate_slab(struct kmem_cache *s, gfp_t flags, int node)
> >
> > static struct page *new_slab(struct kmem_cache *s, gfp_t flags, int node)
> > {
> > - if (unlikely(flags & GFP_SLAB_BUG_MASK)) {
> > - gfp_t invalid_mask = flags & GFP_SLAB_BUG_MASK;
> > - flags &= ~GFP_SLAB_BUG_MASK;
> > - pr_warn("Unexpected gfp: %#x (%pGg). Fixing up to gfp: %#x (%pGg). Fix your code!\n",
> > - invalid_mask, &invalid_mask, flags, &flags);
> > - dump_stack();
> > - }
> > + flags = fixup_gfp_flags(s, flags);
> >
> > return allocate_slab(s,
> > flags & (GFP_RECLAIM_MASK | GFP_CONSTRAINT_MASK), node);
> >
>
Yeah, you are right. I'm very sorry that I was not thoughtful before.
Please ignore
this patch. Thanks!
--
Yours,
Muchun
prev parent reply other threads:[~2020-06-15 13:33 UTC|newest]
Thread overview: 6+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-06-14 6:38 Muchun Song
2020-06-14 10:18 ` kernel test robot
2020-06-14 11:05 ` [External] " Muchun Song
2020-06-15 9:13 ` Michal Hocko
2020-06-15 13:08 ` Vlastimil Babka
2020-06-15 13:32 ` Muchun Song [this message]
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='CAMZfGtUeUyi5Tg0hv3j=-scFGiZJB5hQs8v5nOfYQ7xXsB2nsQ@mail.gmail.com' \
--to=songmuchun@bytedance.com \
--cc=akpm@linux-foundation.org \
--cc=cl@linux.com \
--cc=guro@fb.com \
--cc=iamjoonsoo.kim@lge.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=penberg@kernel.org \
--cc=rientjes@google.com \
--cc=vbabka@suse.cz \
/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