On Wed, Mar 18, 2020 at 08:50:44AM +0530, Srikar Dronamraju wrote:
> * Vlastimil Babka <vbabka@xxxxxxx> [2020-03-17 17:45:15]:
>
> > On 3/17/20 5:25 PM, Srikar Dronamraju wrote:
> > > * Vlastimil Babka <vbabka@xxxxxxx> [2020-03-17 16:56:04]:
> > >
> > >>
> > >> I wonder why do you get a memory leak while Sachin in the same situation [1]
> > >> gets a crash? I don't understand anything anymore.
> > >
> > > Sachin was testing on linux-next which has Kirill's patch which modifies
> > > slub to use kmalloc_node instead of kmalloc. While Bharata is testing on
> > > upstream, which doesn't have this.
> >
> > Yes, that Kirill's patch was about the memcg shrinker map allocation. But the
> > patch hunk that Bharata posted as a "hack" that fixes the problem, it follows
> > that there has to be something else that calls kmalloc_node(node) where node is
> > one that doesn't have present pages.
> >
> > He mentions alloc_fair_sched_group() which has:
> >
> > for_each_possible_cpu(i) {
> > cfs_rq = kzalloc_node(sizeof(struct cfs_rq),
> > GFP_KERNEL, cpu_to_node(i));
> > ...
> > se = kzalloc_node(sizeof(struct sched_entity),
> > GFP_KERNEL, cpu_to_node(i));
> >
>
>
> Sachin's experiment.
> Upstream-next/ memcg /
> possible nodes were 0-31
> online nodes were 0-1
> kmalloc_node called for_each_node / for_each_possible_node.
> This would crash while allocating slab from !N_ONLINE nodes.
>
> Bharata's experiment.
> Upstream
> possible nodes were 0-1
> online nodes were 0-1
> kmalloc_node called for_each_online_node/ for_each_possible_cpu
> i.e kmalloc is called for N_ONLINE nodes.
> So wouldn't crash
>
> Even if his possible nodes were 0-256. I don't think we have kmalloc_node
> being called in !N_ONLINE nodes. Hence its not crashing.
> If we see the above code that you quote, kzalloc_node is using cpu_to_node
> which in Bharata's case will always return 1.
>
>
> > I assume one of these structs is 1k and other 512 bytes (rounded) and that for
> > some possible cpu's cpu_to_node(i) will be 0, which has no present pages. And as
> > Bharata pasted, node_to_mem_node(0) = 0
Correct, these two kazalloc_node() calls for all possible cpus are
causing increased slab memory consumption in my case.
> > So this looks like the same scenario, but it doesn't crash? Is the node 0
> > actually online here, and/or does it have N_NORMAL_MEMORY state?
>
Node 0 is online, but N_NORMAL_MEMORY state is empty. In fact memory
leak goes away if I insert the below check/assignment in the slab
alloc code path:
+ if (!node_isset(node, node_states[N_NORMAL_MEMORY]))
+ node = NUMA_NO_NODE;
Regards,
Bharata.
