Khugepaged attempts to raise min_free_kbytes if its set too low. However, on boot khugepaged sets min_free_kbytes first from subsys_initcall(), and then the mm 'core' over-rides min_free_kbytes after from init_per_zone_wmark_min(), via a module_init() call. Khugepaged used to use a late_initcall() to set min_free_kbytes (such that it occurred after the core initialization), however this was removed when the initialization of min_free_kbytes was integrated into the starting of the khugepaged thread. The fix here is simply to invoke the core initialization using a core_initcall() instead of module_init(), such that the previous initialization ordering is restored. I didn't restore the late_initcall() since start_stop_khugepaged() already sets min_free_kbytes via set_recommended_min_free_kbytes(). This was noticed when we had a number of page allocation failures when moving a workload to a kernel with this new initialization ordering. On an 8GB system this restores min_free_kbytes back to 67584 from 11365 when CONFIG_TRANSPARENT_HUGEPAGE=y is set and either CONFIG_TRANSPARENT_HUGEPAGE_ALWAYS=y or CONFIG_TRANSPARENT_HUGEPAGE_MADVISE=y.
Unified · Split
| @@ -6485,7 +6485,7 @@ int __meminit init_per_zone_wmark_min(void) · 6484 unmodified lines | |||
| 6485 | setup_per_zone_inactive_ratio(); | 6485 | setup_per_zone_inactive_ratio(); |
| 6486 | return 0; | 6486 | return 0; |
| 6487 | } | 6487 | } |
| 6488 | module_init(init_per_zone_wmark_min) | 6488 | core_initcall(init_per_zone_wmark_min) |
| 6489 | 6489 | ||
| 6490 | /* | 6490 | /* |
| 6491 | * min_free_kbytes_sysctl_handler - just a wrapper around proc_dointvec() so | 6491 | * min_free_kbytes_sysctl_handler - just a wrapper around proc_dointvec() so |