Messages in this thread | | | Date | Wed, 20 Apr 2022 14:17:34 +0800 | Subject | Re: [PATCH v4] hugetlb: Fix wrong use of nr_online_nodes | From | "liupeng (DM)" <> |
| |
On 2022/4/19 22:07, Kefeng Wang wrote: > > On 2022/4/19 12:03, Andrew Morton wrote: >> On Sat, 16 Apr 2022 10:35:26 +0000 Peng Liu <liupeng256@huawei.com> >> wrote: >> >>> Certain systems are designed to have sparse/discontiguous nodes. In >>> this case, nr_online_nodes can not be used to walk through numa node. >>> Also, a valid node may be greater than nr_online_nodes. >>> >>> However, in hugetlb, it is assumed that nodes are contiguous. Recheck >>> all the places that use nr_online_nodes, and repair them one by one. >> oops. >> >> What are the user-visible runtime effects of this flaw? > > For example, there are four node=0,1,2,3, and nid = 1 is offline > node,nr_online_nodes = 3 > > 1) per-node alloc (hugepages=1:2) fails, > > 2) per-node alloc (hugepages=3:2) fails, but it could succeed. > > I assume that there is no user-visible runtime effects. > Thanks, you are right.
I have constructed node =0, 1, 3, 4, and requested huge pages as:
hugepagesz=1G hugepages=0:1,4:1 hugepagesz=2M hugepages=0:1024,4:1024
Without this patch:
HugeTLB: Invalid hugepages parameter 4:1 HugeTLB: Invalid hugepages parameter 4:1024 HugeTLB registered 1.00 GiB page size, pre-allocated 0 pages HugeTLB registered 2.00 MiB page size, pre-allocated 1024 pages
With this patch:
HugeTLB registered 1.00 GiB page size, pre-allocated 2 pages HugeTLB registered 2.00 MiB page size, pre-allocated 2048 pages
>> . > .
| |