On Tue, Jun 30, 2020 at 01:08:03AM -0400, Qian Cai wrote: > Even after applied the compling fix, > > https://lore.kernel.org/linux-mm/20200628065409.GA546944@u2004/ > > madvise(MADV_SOFT_OFFLINE) will fail with EIO with hugetlb where it > would succeed without this series. Steps: > > # git clone https://github.com/cailca/linux-mm > # cd linux-mm; make > # ./random 1 (Need at least two NUMA memory nodes) > start: migrate_huge_offline > - use NUMA nodes 0,4. > - mmap and free 8388608 bytes hugepages on node 0 > - mmap and free 8388608 bytes hugepages on node 4 > madvise: Input/output error Ok, sorry for the lateness, but I had to re-fetch the code on my brain again. I just finished v4 of this patchset and it seems this problem is gone: # ./random 1 - start: migrate_huge_offline - use NUMA nodes 0,1. - mmap and free 8388608 bytes hugepages on node 0 - mmap and free 8388608 bytes hugepages on node 1 - pass: mmap_offline_node_huge - start: hotplug_memory offline: Device or resource busy offline: Device or resource busy offline: Device or resource busy offline: Device or resource busy offline: Device or resource busy offline: Device or resource busy offline: Device or resource busy offline: Device or resource busy offline: Invalid argument offline: Device or resource busy offline: Invalid argument offline: Device or resource busy offline: Device or resource busy offline: Device or resource busy offline: Device or resource busy - pass: hotplug_memory The test seems to suceed and no crash on the kernel side. I will just run some more tests to make sure the thing is solid enough and then I will post v4. Thanks -- Oscar Salvador SUSE L3
