MIME-Version: 1.0
References: <e52068a6e98031a386b5052a166a55c94fe940f6.camel@redhat.com> <ff89f867-0709-d8bb-b6f5-51b2be4cc2dd@oracle.com>
In-Reply-To: <ff89f867-0709-d8bb-b6f5-51b2be4cc2dd@oracle.com>
From:   Zhenguo Yao <yaozhenguo1@gmail.com>
Date:   Mon, 29 Nov 2021 20:26:02 +0800
Message-ID: <CA+WzARmTSD_S22xHSp2TinobzEXDwZzPU5vv7NX7-SqtUOtA5g@mail.gmail.com>
Subject: Re: Commit 'hugetlbfs: extend the definition of hugepages parameter
 to support node allocation' breaks old numa less syntax of reserving
 hugepages on boot.
To:     Mike Kravetz <mike.kravetz@oracle.com>
Cc:     Maxim Levitsky <mlevitsk@redhat.com>,
        Andrew Morton <akpm@linux-foundation.org>,
        Linux Memory Management List <linux-mm@kvack.org>,
        linux-kernel@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

Mike Kravetz <mike.kravetz@oracle.com> =E4=BA=8E2021=E5=B9=B411=E6=9C=8829=
=E6=97=A5=E5=91=A8=E4=B8=80 =E4=B8=8B=E5=8D=8812:31=E5=86=99=E9=81=93=EF=BC=
=9A
>
> On 11/28/21 03:18, Maxim Levitsky wrote:
> >
> > dmesg prints this:
> >
> > HugeTLB: allocating 64 of page size 1.00 GiB failed.  Only allocated 0 =
hugepages
> >
> > Huge pages were allocated on kernel command line (1/2 of 128GB system):
> >
> > 'default_hugepagesz=3D1G hugepagesz=3D1G hugepages=3D64'
> >
> > This is 3970X and no real support/need for NUMA, thus only fake NUMA no=
de 0 is present.
> >
> > Reverting the commit helps.
> >
> > New syntax also works ( hugepages=3D0:64 )
> >
> > I can test any patches for this bug.
>
> Argh!  I think preallocation of gigantic pages on all systems with only
> a single node is broken.  The issue is at the beginning of
> __alloc_bootmem_huge_page:
>
> int __alloc_bootmem_huge_page(struct hstate *h, int nid)
> {
>         struct huge_bootmem_page *m =3D NULL; /* initialize for clang */
>         int nr_nodes, node;
>
>         if (nid >=3D nr_online_nodes)
>                 return 0;
>
> Without using the node specific syntax, nid =3D=3D NUMA_NO_NODE =3D=3D -1=
.  For the
> comparison, nid will be converted to an unsigned into to match nr_online_=
nodes
> so we will immediately return 0 instead of doing the allocations.
>
> Zhenguo Yao,
> Can you verify and perhaps put together a patch?does
>
Preallocation of gigantic pages cant=E2=80=98 work in all the environment, =
not
only in single node.
I think the issue is because of the replacement
nodes_weight(node_states[N_MEMORY] with
nr_online_nodes in my patch of last version. Sorry for my careless. I
didn't notice that  parameter
nid is int =EF=BC=8Cbut nr_online_nodes is unsigned int. so, this if (nid >=
=3D
nr_online_nodes) is always
true when nid is NUMA_NO_NODE(-1).  I will send a fix  as soon as passible.
This is really a low-level mistake ^^
> >
> > Also unrelated, is there any progress on allocating 1GB pages on demand=
 so that I could
> > allocate them only when I run a VM?
>
> That should be possible.  Such support was added back in 2014 with commit
> 944d9fec8d7a "hugetlb: add support for gigantic page allocation at runtim=
e".
>
> >
> > i don't mind having these pages to be marked as to be used for userspac=
e only,
> > since as far as I remember its the kernel usage that makes some page un=
moveable.
> >
>
> Of course, finding 1GB of contiguous space for a gigantic page is often
> difficult at runtime.  So, allocations are likely to fail the longer the
> system is up and running and fragmentation increases.
>
> > Last time (many years ago) I tried to create a zone with only userspace=
 pages
> > (I don't remember what options I used) but it didn't work.
>
> Not too long ago, support was added to use CMA for gigantic page allocati=
on.
> See commit cf11e85fc08c "mm: hugetlb: optionally allocate gigantic hugepa=
ges
> using cma".  This sounds like something you might want to try.
> --
> Mike Kravetz
>
> >
> > Is there a way to debug what is causing unmoveable pages and doesn't le=
t
> > /proc/sys/vm/nr_hugepages work (I tried it today and as usual the numbe=
r
> > it can allocate steadly decreases over time).
>
>
>
>