+ mm-mempolicy-recompute-automatic-interleave-weights-on-memory-hotplug.patch added to mm-hotfixes-unstable branch

+ mm-mempolicy-recompute-automatic-interleave-weights-on-memory-hotplug.patch added to mm-hotfixes-unstable branch

[Andrew Morton]

The patch titled
     Subject: mm/mempolicy: recompute automatic interleave weights on memory hotplug
has been added to the -mm mm-hotfixes-unstable branch.  Its filename is
     mm-mempolicy-recompute-automatic-interleave-weights-on-memory-hotplug.patch

This patch will shortly appear at
     https://git.kernel.org/pub/scm/linux/kernel/git/akpm/25-new.git/tree/patches/mm-mempolicy-recompute-automatic-interleave-weights-on-memory-hotplug.patch

This patch will later appear in the mm-hotfixes-unstable branch at
    git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm

Before you just go and hit "reply", please:
   a) Consider who else should be cc'ed
   b) Prefer to cc a suitable mailing list as well
   c) Ideally: find the original patch on the mailing list and do a
      reply-to-all to that, adding suitable additional cc's

*** Remember to use Documentation/process/submit-checklist.rst when testing your code ***

The -mm tree is included into linux-next via various
branches at git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm
and is updated there most days

------------------------------------------------------
From: "Gregory Price (Meta)" <gourry@gourry.net>
Subject: mm/mempolicy: recompute automatic interleave weights on memory hotplug
Date: Fri, 30 Oct 2026 12:12:34 -0400

In auto mode, weighted interleave derives each node's weight from its
bandwidth, reduced over the nodes in N_MEMORY.  HMAT and CXL report a
node's bandwidth once, when the node first gains memory, and nothing else
on the hotplug path recomputes the weights.

This can result in a major skew in distribution and overload tiers with
more data than their respective bandwidth can support, resulting in a
performance regression.

With HMAT bandwidths of 100G for nodes 0-3, and 50G and 25G for kmem nodes
4 and 5, weighted_interleave/nodeN reads:

  event             node 0-3   node 4   node 5
  node 4 online         7         3        -
  node 4 removed        7         -        -
  node 5 online         7         -        1
  node 4 back           7         1        1
  Expected:             6         3        1

Toggling auto off and on generates the correct result, but we should
detect this and fix it during hotplug.  Recompute the weights from the
recorded bandwidth whenever a node gains/loses its first/last memory.

The notifier cannot fail the hotplug, which has already happened, so warn
once if the new weights cannot be allocated.  The next recompute starts
from the full bandwidth table and corrects them.

mempolicy_set_node_perf() now only records the bandwidth and calls the
same helper, so the bandwidth is kept even if allocating the new weights
fails.  wi_state_lock is dropped between the two steps, which should be
safe because racing writer simply causes it to run with newer state.

Link: https://lore.kernel.org/20261030161234.1505367-1-gourry@gourry.net
Fixes: e341f9c3c841 ("mm/mempolicy: Weighted Interleave Auto-tuning")
Signed-off-by: Gregory Price (Meta) <gourry@gourry.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Joshua Hahn <joshua.hahnjy@gmail.com>
Cc: David Hildenbrand <david@kernel.org>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Rakie Kim <rakie.kim@sk.com>
Cc: Ying Huang <ying.huang@linux.alibaba.com>
Cc: Harry Yoo <harry@kernel.org>
Cc: <stable@vger.kernel.org>
---

 mm/mempolicy.c |   84 +++++++++++++++++++++++++++++++----------------
 1 file changed, 57 insertions(+), 27 deletions(-)

--- a/mm/mempolicy.c~mm-mempolicy-recompute-automatic-interleave-weights-on-memory-hotplug
+++ a/mm/mempolicy.c
@@ -245,27 +245,49 @@ static void reduce_interleave_weights(un
 		new_iw[nid] /= iw_gcd;
 }
 
+/* Recompute the weights from node_bw_table, unless they were set manually */
+static int wi_state_refresh(void)
+{
+	struct weighted_interleave_state *new_wi_state, *old_wi_state;
+	int i;
+
+	new_wi_state = kmalloc_flex(*new_wi_state, iw_table, nr_node_ids);
+	if (!new_wi_state)
+		return -ENOMEM;
+	new_wi_state->mode_auto = true;
+	for (i = 0; i < nr_node_ids; i++)
+		new_wi_state->iw_table[i] = 1;
+
+	mutex_lock(&wi_state_lock);
+	old_wi_state = rcu_dereference_protected(wi_state,
+					lockdep_is_held(&wi_state_lock));
+	/*
+	 * Hotplug events can arrive before any node reported its bandwidth,
+	 * leaving nothing to compute weights from. NULL wi_state assumes
+	 * auto=true.
+	 */
+	if (!node_bw_table || (old_wi_state && !old_wi_state->mode_auto)) {
+		mutex_unlock(&wi_state_lock);
+		kfree(new_wi_state);
+		return 0;
+	}
+	reduce_interleave_weights(node_bw_table, new_wi_state->iw_table);
+	rcu_assign_pointer(wi_state, new_wi_state);
+	mutex_unlock(&wi_state_lock);
+	wi_state_retire(old_wi_state);
+	return 0;
+}
+
 int mempolicy_set_node_perf(unsigned int node, struct access_coordinate *coords)
 {
-	struct weighted_interleave_state *new_wi_state, *old_wi_state = NULL;
 	unsigned int *old_bw, *new_bw;
 	unsigned int bw_val;
-	int i;
 
 	bw_val = min(coords->read_bandwidth, coords->write_bandwidth);
 	new_bw = kcalloc(nr_node_ids, sizeof(unsigned int), GFP_KERNEL);
 	if (!new_bw)
 		return -ENOMEM;
 
-	new_wi_state = kmalloc_flex(*new_wi_state, iw_table, nr_node_ids);
-	if (!new_wi_state) {
-		kfree(new_bw);
-		return -ENOMEM;
-	}
-	new_wi_state->mode_auto = true;
-	for (i = 0; i < nr_node_ids; i++)
-		new_wi_state->iw_table[i] = 1;
-
 	/*
 	 * Update bandwidth info, even in manual mode. That way, when switching
 	 * to auto mode in the future, iw_table can be overwritten using
@@ -278,25 +300,32 @@ int mempolicy_set_node_perf(unsigned int
 		memcpy(new_bw, old_bw, nr_node_ids * sizeof(*old_bw));
 	new_bw[node] = bw_val;
 	node_bw_table = new_bw;
+	mutex_unlock(&wi_state_lock);
 
-	old_wi_state = rcu_dereference_protected(wi_state,
-					lockdep_is_held(&wi_state_lock));
-	if (old_wi_state && !old_wi_state->mode_auto) {
-		/* Manual mode; skip reducing weights and updating wi_state */
-		mutex_unlock(&wi_state_lock);
-		kfree(new_wi_state);
-		goto out;
-	}
+	kfree(old_bw);
+	return wi_state_refresh();
+}
 
-	/* NULL wi_state assumes auto=true; reduce weights and update wi_state*/
-	reduce_interleave_weights(new_bw, new_wi_state->iw_table);
-	rcu_assign_pointer(wi_state, new_wi_state);
+/*
+ * HMAT and CXL report a node's bandwidth only when it first gains memory, so
+ * recompute the weights whenever N_MEMORY changes.
+ */
+static int wi_state_node_notifier(struct notifier_block *nb,
+				  unsigned long action, void *data)
+{
+	struct node_notify *nn = data;
+	int err;
 
-	mutex_unlock(&wi_state_lock);
-	wi_state_retire(old_wi_state);
-out:
-	kfree(old_bw);
-	return 0;
+	if (action != NODE_ADDED_FIRST_MEMORY &&
+	    action != NODE_REMOVED_LAST_MEMORY)
+		return NOTIFY_OK;
+
+	/* The hotplug is done, failing the notification would not undo it */
+	err = wi_state_refresh();
+	if (err)
+		pr_warn_once("failed to recompute weighted interleave weights for node%d: %d\n",
+			     nn->nid, err);
+	return NOTIFY_OK;
 }
 
 /**
@@ -3421,6 +3450,7 @@ void __init numa_policy_init(void)
 	if (do_set_mempolicy(MPOL_INTERLEAVE, 0, &interleave_nodes))
 		pr_err("%s: interleaving failed\n", __func__);
 
+	hotplug_node_notifier(wi_state_node_notifier, DEFAULT_CALLBACK_PRI);
 	check_numabalancing_enable();
 }
 
_

Patches currently in -mm which might be from gourry@gourry.net are

mm-page_alloc-let-the-highest-memory-node-recover-from-hopeless-state.patch
mm-mempolicy-recompute-automatic-interleave-weights-on-memory-hotplug.patch
mm-memory-tiers-reset-demotion-targets-of-nodes-without-memory.patch
mm-mempolicy-convert-mempolicy-refcnt-to-refcount_t.patch
selftests-mm-add-madv_cold-and-madv_pageout-contract-tests.patch
selftests-mm-check-madv_cold-and-madv_pageout-range-handling.patch
selftests-mm-check-that-madv_cold-and-madv_pageout-reject-hugetlb.patch
selftests-mm-check-that-madv_cold-and-madv_pageout-preserve-shared-memory.patch
selftests-mm-check-madv_cold-and-madv_pageout-on-partly-advised-thps.patch
selftests-mm-check-madv_pageout-file-permissions.patch
selftests-mm-check-that-unauthorized-madv_pageout-skips-large-file-folios.patch