From: Michal Hocko <mhocko@kernel.org>
To: Pingfan Liu <kernelfans@gmail.com>
Cc: linux-mm@kvack.org, linux-kernel@vger.kernel.org,
Andrew Morton <akpm@linux-foundation.org>,
Vlastimil Babka <vbabka@suse.cz>,
Mike Rapoport <rppt@linux.vnet.ibm.com>,
Bjorn Helgaas <bhelgaas@google.com>,
Jonathan Cameron <Jonathan.Cameron@huawei.com>
Subject: Re: [PATCH] mm/alloc: fallback to first node if the wanted node offline
Date: Tue, 4 Dec 2018 09:56:01 +0100 [thread overview]
Message-ID: <20181204085601.GC1286@dhcp22.suse.cz> (raw)
In-Reply-To: <CAFgQCTv56drDBx-sTr6KdeQNKJnojG3g_a-k8wKe_q2y9w9NtA@mail.gmail.com>
On Tue 04-12-18 16:20:32, Pingfan Liu wrote:
> On Tue, Dec 4, 2018 at 3:22 PM Michal Hocko <mhocko@kernel.org> wrote:
> >
> > On Tue 04-12-18 11:05:57, Pingfan Liu wrote:
> > > During my test on some AMD machine, with kexec -l nr_cpus=x option, the
> > > kernel failed to bootup, because some node's data struct can not be allocated,
> > > e.g, on x86, initialized by init_cpu_to_node()->init_memory_less_node(). But
> > > device->numa_node info is used as preferred_nid param for
> > > __alloc_pages_nodemask(), which causes NULL reference
> > > ac->zonelist = node_zonelist(preferred_nid, gfp_mask);
> > > This patch tries to fix the issue by falling back to the first online node,
> > > when encountering such corner case.
> >
> > We have seen similar issues already and the bug was usually that the
> > zonelists were not initialized yet or the node is completely bogus.
> > Zonelists should be initialized by build_all_zonelists quite early so I
> > am wondering whether the later is the case. What is the actual node
> > number the device is associated with?
> >
> The device's node num is 2. And in my case, I used nr_cpus param. Due
> to init_cpu_to_node() initialize all the possible node. It is hard
> for me to figure out without this param, how zonelists is accessed
> before page allocator works.
I believe we should focus on this. Why does the node have no zonelist
even though all zonelists should be initialized already? Maybe this is
nr_cpus pecularity and we do not initialize all the existing numa nodes.
Or maybe the device is associated to a non-existing node with that
setup. A full dmesg might help us here.
> > Your patch is not correct btw, because we want to fallback into the node in
> > the distance order rather into the first online node.
> > --
> What about this:
> +extern int find_next_best_node(int node, nodemask_t *used_node_mask);
> +
> /*
> * We get the zone list from the current node and the gfp_mask.
> * This zone list contains a maximum of MAXNODES*MAX_NR_ZONES zones.
> @@ -453,6 +455,11 @@ static inline int gfp_zonelist(gfp_t flags)
> */
> static inline struct zonelist *node_zonelist(int nid, gfp_t flags)
> {
> + if (unlikely(!node_online(nid))) {
> + nodemask_t used_mask;
> + nodes_complement(used_mask, node_online_map);
> + nid = find_next_best_node(nid, &used_mask);
> + }
> return NODE_DATA(nid)->node_zonelists + gfp_zonelist(flags);
> }
>
> I just finished the compiling, not test it yet, since the machine is
> not on hand yet. It needs some time to get it again.
This is clearly a no-go. nodemask_t can be giant and you cannot have it
on the stack for allocation paths which might be called from a deep
stack already. Also this is called from the allocator hot paths and each
branch counts.
--
Michal Hocko
SUSE Labs
next prev parent reply other threads:[~2018-12-04 8:56 UTC|newest]
Thread overview: 56+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-12-04 3:05 [PATCH] mm/alloc: fallback to first node if the wanted node offline Pingfan Liu
2018-12-04 3:53 ` David Rientjes
2018-12-04 7:16 ` Pingfan Liu
2018-12-05 5:49 ` Pingfan Liu
2018-12-05 19:00 ` David Rientjes
2018-12-04 6:54 ` Wei Yang
2018-12-04 7:20 ` Pingfan Liu
2018-12-04 8:34 ` Wei Yang
2018-12-04 8:52 ` Pingfan Liu
2018-12-04 9:09 ` Wei Yang
2018-12-05 5:50 ` Pingfan Liu
2018-12-04 7:22 ` Michal Hocko
2018-12-04 8:20 ` Pingfan Liu
2018-12-04 8:40 ` Wei Yang
2018-12-04 8:56 ` Pingfan Liu
2018-12-04 8:56 ` Michal Hocko [this message]
2018-12-04 14:42 ` Vlastimil Babka
2018-12-05 5:38 ` Pingfan Liu
2018-12-05 9:21 ` Michal Hocko
2018-12-05 9:29 ` Pingfan Liu
2018-12-05 9:40 ` Vlastimil Babka
2018-12-06 3:07 ` Pingfan Liu
2018-12-06 8:28 ` Michal Hocko
2018-12-06 10:03 ` Pingfan Liu
2018-12-06 10:44 ` Pingfan Liu
2018-12-06 12:11 ` Michal Hocko
2018-12-07 2:56 ` Pingfan Liu
2018-12-07 7:53 ` Michal Hocko
2018-12-07 9:40 ` Pingfan Liu
2018-12-07 11:30 ` Michal Hocko
2018-12-07 13:20 ` Pingfan Liu
2018-12-07 14:22 ` Michal Hocko
2018-12-07 14:27 ` Pingfan Liu
2018-12-07 14:50 ` Michal Hocko
2018-12-07 15:56 ` Michal Hocko
2018-12-10 4:00 ` Pingfan Liu
2018-12-10 7:57 ` Pingfan Liu
2018-12-10 12:37 ` Michal Hocko
2018-12-11 8:05 ` Pingfan Liu
2018-12-11 9:44 ` Michal Hocko
2018-12-12 8:33 ` Pingfan Liu
2018-12-12 8:31 ` Pingfan Liu
2018-12-12 11:53 ` Michal Hocko
2018-12-13 8:37 ` Pingfan Liu
2018-12-13 9:04 ` Pingfan Liu
2018-12-17 13:29 ` Michal Hocko
2018-12-20 7:19 ` Pingfan Liu
2018-12-20 9:19 ` Michal Hocko
2019-01-08 14:34 ` Michal Hocko
2019-01-09 3:13 ` Pingfan Liu
2019-01-11 3:12 ` Pingfan Liu
2019-01-11 9:23 ` Michal Hocko
2018-12-17 12:57 ` Michal Hocko
2018-12-05 9:43 ` Michal Hocko
2018-12-06 3:34 ` Pingfan Liu
2018-12-06 7:23 ` Michal Hocko
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=20181204085601.GC1286@dhcp22.suse.cz \
--to=mhocko@kernel.org \
--cc=Jonathan.Cameron@huawei.com \
--cc=akpm@linux-foundation.org \
--cc=bhelgaas@google.com \
--cc=kernelfans@gmail.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=rppt@linux.vnet.ibm.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;
as well as URLs for NNTP newsgroup(s).