Messages in this thread |  | | From | "Uladzislau Rezki (Sony)" <> | Subject | [PATCH 0/3] vmalloc() vs bulk allocator v2 | Date | Sun, 16 May 2021 22:20:53 +0200 |
| |
Hi.
There are three patches in this small series. This is a second iteration. First one was buggy and leaded to kernel panic due to passing NUMA_NO_NODE as "nid" to the bulk allocator(it assumes to have a valid node).
Therefore the patch number [1] adds an extra helper that guarantees a correct numa node ID if it is not specified. Basically saying when it is invoked with NUMA_NO_NODE.
A patch [2] has been slightly updated, the change-log is below.
V1 -> V2: - Switch to the alloc_pages_bulk_array_node() helper so the NUMA_NO_NODE is correctly handled(similar to alloc_pages_node() API func). - Use "while()" loop instead of "for()" for high-order pages and increase number of allocated pages after actual alloc.
Uladzislau Rezki (Sony) (3): [1] mm/page_alloc: Add an alloc_pages_bulk_array_node() helper [2] mm/vmalloc: Switch to bulk allocator in __vmalloc_area_node() [3] mm/vmalloc: Print a warning message first on failure
include/linux/gfp.h | 9 ++++++ mm/vmalloc.c | 78 +++++++++++++++++++++++++-------------------- 2 files changed, 52 insertions(+), 35 deletions(-)
-- 2.20.1
|  |