On 8/3/22 20:22, Alistair Popple wrote:
@@ -1970,29 +1968,25 @@ static long check_and_migrate_movable_pages(unsigned long nr_pages,
drain_allow = false;
}
- if (folio_isolate_lru(folio)) {
- isolation_error_count++;
+ if (folio_isolate_lru(folio))
continue;
- }
+
list_add_tail(&folio->lru, &movable_page_list);
node_stat_mod_folio(folio,
NR_ISOLATED_ANON + folio_is_file_lru(folio),
folio_nr_pages(folio));
}
- if (!list_empty(&movable_page_list) || isolation_error_count ||
- coherent_pages)
- goto unpin_pages;
-
/*
- * If list is empty, and no isolation errors, means that all pages are
- * in the correct zone.
+ * All pages are still pinned and in the correct zone.
*/
- return nr_pages;
+ if (list_empty(&movable_page_list) && !coherent_pages)
+ return 0;
At this point, a second part of the (rather long) routine begins. If I
were refactoring this, I'd split it into a top level routine that calls
the two parts:
check_and_migrate_movable_pages()
collect_migration_candidates()
unpin_and_migrate_pages()
...which I think would further help in working with this thing.
(I realize that this is additional work. And that it's best done as a
separate patch. And it's just a suggestion.)
-unpin_pages:
/*
- * pages[i] might be NULL if any device coherent pages were found.
+ * Unpin all pages. If device coherent pages were found
+ * migrate_deivce_coherent_page() will have already dropped the pin and
"migrate_device_coherent_page()"
thanks,
--
John Hubbard
NVIDIA