On 4/1/21 3:35 PM, Wei Xu wrote: > A small suggestion: Given that migrate_pages() requires that > *nr_succeeded should be initialized to 0 when it is called due to its > use of *nr_succeeded in count_vm_events() and trace_mm_migrate_pages(), > it would be less error-prone if migrate_pages() initializes > *nr_succeeded itself. That's a good point, especially if a caller made multiple migrate_pages() calls without resetting it, which a number of callers do. That could really have caused some interesting problems. Thanks for catching that! I'll do what you suggested.
