lkml.org 
[lkml]   [2019]   [May]   [23]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
Date
SubjectRe: [PATCH -next v2] mm/hotplug: fix a null-ptr-deref during NUMA boot
On Thu, May 23, 2019 at 11:58 AM Pingfan Liu <kernelfans@gmail.com> wrote:
>
> On Wed, May 22, 2019 at 7:16 PM Michal Hocko <mhocko@kernel.org> wrote:
> >
> > On Wed 22-05-19 15:12:16, Pingfan Liu wrote:
> > > On Mon, May 13, 2019 at 11:31 PM Michal Hocko <mhocko@kernel.org> wrote:
> > > >
> > > > On Mon 13-05-19 11:20:46, Qian Cai wrote:
> > > > > On Mon, 2019-05-13 at 16:04 +0200, Michal Hocko wrote:
> > > > > > On Mon 13-05-19 09:43:59, Qian Cai wrote:
> > > > > > > On Mon, 2019-05-13 at 14:41 +0200, Michal Hocko wrote:
> > > > > > > > On Sun 12-05-19 01:48:29, Qian Cai wrote:
> > > > > > > > > The linux-next commit ("x86, numa: always initialize all possible
> > > > > > > > > nodes") introduced a crash below during boot for systems with a
> > > > > > > > > memory-less node. This is due to CPUs that get onlined during SMP boot,
> > > > > > > > > but that onlining triggers a page fault in bus_add_device() during
> > > > > > > > > device registration:
> > > > > > > > >
> > > > > > > > > error = sysfs_create_link(&bus->p->devices_kset->kobj,
> > > > > > > > >
> > > > > > > > > bus->p is NULL. That "p" is the subsys_private struct, and it should
> > > > > > > > > have been set in,
> > > > > > > > >
> > > > > > > > > postcore_initcall(register_node_type);
> > > > > > > > >
> > > > > > > > > but that happens in do_basic_setup() after smp_init().
> > > > > > > > >
> > > > > > > > > The old code had set this node online via alloc_node_data(), so when it
> > > > > > > > > came time to do_cpu_up() -> try_online_node(), the node was already up
> > > > > > > > > and nothing happened.
> > > > > > > > >
> > > > > > > > > Now, it attempts to online the node, which registers the node with
> > > > > > > > > sysfs, but that can't happen before the 'node' subsystem is registered.
> > > > > > > > >
> > > > > > > > > Since kernel_init() is running by a kernel thread that is in
> > > > > > > > > SYSTEM_SCHEDULING state, fixed this by skipping registering with sysfs
> > > > > > > > > during the early boot in __try_online_node().
> > > > > > > >
> > > > > > > > Relying on SYSTEM_SCHEDULING looks really hackish. Why cannot we simply
> > > > > > > > drop try_online_node from do_cpu_up? Your v2 remark below suggests that
> > > > > > > > we need to call node_set_online because something later on depends on
> > > > > > > > that. Btw. why do we even allocate a pgdat from this path? This looks
> > > > > > > > really messy.
> > > > > > >
> > > > > > > See the commit cf23422b9d76 ("cpu/mem hotplug: enable CPUs online before
> > > > > > > local
> > > > > > > memory online")
> > > > > > >
> > > > > > > It looks like try_online_node() in do_cpu_up() is needed for memory hotplug
> > > > > > > which is to put its node online if offlined and then hotadd_new_pgdat()
> > > > > > > calls
> > > > > > > build_all_zonelists() to initialize the zone list.
> > > > > >
> > > > > > Well, do we still have to followthe logic that the above (unreviewed)
> > > > > > commit has established? The hotplug code in general made a lot of ad-hoc
> > > > > > design decisions which had to be revisited over time. If we are not
> > > > > > allocating pgdats for newly added memory then we should really make sure
> > > > > > to do so at a proper time and hook. I am not sure about CPU vs. memory
> > > > > > init ordering but even then I would really prefer if we could make the
> > > > > > init less obscure and _documented_.
> > > > >
> > > > > I don't know, but I think it is a good idea to keep the existing logic rather
> > > > > than do a big surgery
> > > >
> > > > Adding more hacks just doesn't make the situation any better.
> > > >
> > > > > unless someone is able to confirm it is not breaking NUMA
> > > > > node physical hotplug.
> > > >
> > > > I have a machine to test whole node offline. I am just busy to prepare a
> > > > patch myself. I can have it tested though.
> > > >
> > > I think the definition of "node online" is worth of rethinking. Before
> > > patch "x86, numa: always initialize all possible nodes", online means
> > > either cpu or memory present. After this patch, only node owing memory
> > > as present.
> > >
> > > In the commit log, I think the change's motivation should be "Not to
> > > mention that it doesn't really make much sense to consider an empty
> > > node as online because we just consider this node whenever we want to
> > > iterate nodes to use and empty node is obviously not the best
> > > candidate."
> > >
> > > But in fact, we already have for_each_node_state(nid, N_MEMORY) to
> > > cover this purpose.
> >
> > I do not really think we want to spread N_MEMORY outside of the core MM.
> > It is quite confusing IMHO.
> > .
> But it has already like this. Just git grep N_MEMORY.
>
> > > Furthermore, changing the definition of online may
> > > break something in the scheduler, e.g. in task_numa_migrate(), where
> > > it calls for_each_online_node.
> >
> > Could you be more specific please? Why should numa balancing consider
> > nodes without any memory?
> >
> As my understanding, the destination cpu can be on a memory less node.
> BTW, there are several functions in the scheduler facing the same
> scenario, task_numa_migrate() is an example.
>
> > > By keeping the node owning cpu as online, Michal's patch can avoid
> > > such corner case and keep things easy. Furthermore, if needed, the
> > > other patch can use for_each_node_state(nid, N_MEMORY) to replace
> > > for_each_online_node is some space.
> >
> > Ideally no code outside of the core MM should care about what kind of
> > memory does the node really own. The external code should only care
> > whether the node is online and thus usable or offline and of no
> > interest.
> Yes, but maybe it will pay great effort on it.
>
And as a first step, we can find a way to fix the bug reported by me
and the one reported by Barret

> Regards,
> Pingfan
> > --
> > Michal Hocko
> > SUSE Labs

\
 
 \ /
  Last update: 2019-05-23 06:01    [W:0.124 / U:0.580 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site