在 2025/2/27 14:19, Oscar Salvador 写道:
On Thu, Feb 27, 2025 at 11:11:25AM +0800, Jinjiang Tu wrote:
The result with this patch:
Node 0 Node 1
Total 5 0
Free 0 0
Surp 5 0
Why would node0 have any surplus pages?
Also, this does not reproduce in my machine.
Sorry, I missed a step between step 2) and 3).
After step 2), echo 0 >
/sys/devices/system/node/node0/hugepages/hugepages-2048kB/nr_hugepages
So, the node0 has 5 surplus pages.
Are you overcommitting? Are you using the hvo optimization?
If so, it might be that we cannot allocate vmemmap pages and those
hugetlb pages stay in a surplus state?
>From bulk_vmemmap_restore_error():
* If unable to restore a hugetlb page, the hugetlb
* page is made a surplus page and removed from the list.
