[PATCH slab] slab: Clarify comments regarding pfmemalloc and NUMA preferences

[Alexei Starovoitov <alexei.starovoitov@gmail.com>]

From: Alexei Starovoitov <ast@kernel.org>

Clarify comments regarding pfmemalloc and NUMA preferences
when ___slab_alloc() operating in !allow_spin mode.

Signed-off-by: Alexei Starovoitov <ast@kernel.org>
---
 mm/slub.c | 19 +++++++++++--------
 1 file changed, 11 insertions(+), 8 deletions(-)

diff --git a/mm/slub.c b/mm/slub.c
index 83983de948f3..c995f3bec69d 100644
--- a/mm/slub.c
+++ b/mm/slub.c
@@ -4456,9 +4456,17 @@ static void *___slab_alloc(struct kmem_cache *s, gfp_t gfpflags, int node,
 		/*
 		 * same as above but node_match() being false already
 		 * implies node != NUMA_NO_NODE.
-		 * Reentrant slub cannot take locks necessary to
-		 * deactivate_slab, hence ignore node preference.
-		 * kmalloc_nolock() doesn't allow __GFP_THISNODE.
+		 *
+		 * We don't strictly honor pfmemalloc and NUMA preferences
+		 * when !allow_spin because:
+		 *
+		 * 1. Most kmalloc() users allocate objects on the local node,
+		 *    so kmalloc_nolock() tries not to interfere with them by
+		 *    deactivating the cpu slab.
+		 *
+		 * 2. Deactivating due to NUMA or pfmemalloc mismatch may cause
+		 *    unnecessary slab allocations even when n->partial list
+		 *    is not empty.
 		 */
 		if (!node_isset(node, slab_nodes) ||
 		    !allow_spin) {
@@ -4547,11 +4555,6 @@ static void *___slab_alloc(struct kmem_cache *s, gfp_t gfpflags, int node,
 		slab = slub_percpu_partial(c);
 		slub_set_percpu_partial(c, slab);
 
-		/*
-		 * Reentrant slub cannot take locks necessary for
-		 * __put_partials(), hence ignore node preference.
-		 * kmalloc_nolock() doesn't allow __GFP_THISNODE.
-		 */
 		if (likely(node_match(slab, node) &&
 			   pfmemalloc_match(slab, gfpflags)) ||
 		    !allow_spin) {
-- 
2.47.3