pmdalinux: add mem.numa.hugepages metrics (per node, per hugepage size)

Resolves Red Hat bug RHEL-45876

qa/1788

@@ -2,7 +2,7 @@
 # PCP QA Test No. 1788
 # Exercise Linux kernel hugepage sysfs metrics.
 #
-# Copyright (c) 2024 Red Hat.  All Rights Reserved.
+# Copyright (c) 2024-2025 Red Hat.  All Rights Reserved.
 #
 
 seq=`basename $0`
@@ -18,16 +18,18 @@ echo "QA output created by $seq"
 _cleanup()
 {
     cd $here
-    $sudo rm -f $PCP_VAR_DIR/config/pmda/60.43
+    $sudo rm -f $PCP_VAR_DIR/config/pmda/60.{43,44}
     _restore_config $PCP_VAR_DIR/config/pmda/60.43
+    _restore_config $PCP_VAR_DIR/config/pmda/60.44
     $sudo rm -rf $tmp $tmp.*
 }
 
 status=1	# failure is the default!
 trap "_cleanup; exit \$status" 0 1 2 3 15
 
-# deterministic output - move aside any existing indom cache
+# deterministic output - move aside any existing indom caches
 _save_config $PCP_VAR_DIR/config/pmda/60.43
+_save_config $PCP_VAR_DIR/config/pmda/60.44
 
 _filter_instances()
 {
@@ -40,7 +42,7 @@ export LINUX_STATSPATH=$root
 pmda=$PCP_PMDAS_DIR/linux/pmda_linux.so,linux_init
 local="-L -K clear -K add,60,$pmda"
 
-metrics=`pminfo $local mem.hugepages | LC_COLLATE=POSIX sort`
+metrics=`pminfo $local mem.hugepages mem.numa.hugepages | LC_COLLATE=POSIX sort`
 for tgz in $here/linux/sysfs-hugepages-*.tgz
 do
     rm -fr $root

qa/1788.out

@@ -37,6 +37,18 @@ mem.hugepages.totalsize
     inst [N or "32768kB"] value 327680
     inst [N or "64kB"] value 6400
 
+mem.numa.hugepages.free
+No value(s) available!
+
+mem.numa.hugepages.pagesize
+No value(s) available!
+
+mem.numa.hugepages.surplus
+No value(s) available!
+
+mem.numa.hugepages.totalsize
+No value(s) available!
+
 == done
 
 == Checking hugepages metric values - sysfs-hugepages-002.tgz
@@ -65,5 +77,81 @@ mem.hugepages.totalsize
     inst [N or "1048576kB"] value 104857600
     inst [N or "2048kB"] value 4096000
 
+mem.numa.hugepages.free
+No value(s) available!
+
+mem.numa.hugepages.pagesize
+No value(s) available!
+
+mem.numa.hugepages.surplus
+No value(s) available!
+
+mem.numa.hugepages.totalsize
+No value(s) available!
+
+== done
+
+== Checking hugepages metric values - sysfs-hugepages-003.tgz
+
+mem.hugepages.free
+    inst [N or "1048576kB"] value 104857600
+    inst [N or "2048kB"] value 4096000
+    inst [N or "32768kB"] value 0
+    inst [N or "64kB"] value 0
+
+mem.hugepages.overcommit
+    inst [N or "1048576kB"] value 0
+    inst [N or "2048kB"] value 0
+    inst [N or "32768kB"] value 0
+    inst [N or "64kB"] value 0
+
+mem.hugepages.pagesize
+    inst [N or "1048576kB"] value 1048576
+    inst [N or "2048kB"] value 2048
+    inst [N or "32768kB"] value 32768
+    inst [N or "64kB"] value 64
+
+mem.hugepages.reserved
+    inst [N or "1048576kB"] value 0
+    inst [N or "2048kB"] value 16384
+    inst [N or "32768kB"] value 0
+    inst [N or "64kB"] value 0
+
+mem.hugepages.surplus
+    inst [N or "1048576kB"] value 0
+    inst [N or "2048kB"] value 8192
+    inst [N or "32768kB"] value 0
+    inst [N or "64kB"] value 0
+
+mem.hugepages.totalsize
+    inst [N or "1048576kB"] value 104857600
+    inst [N or "2048kB"] value 4096000
+    inst [N or "32768kB"] value 0
+    inst [N or "64kB"] value 0
+
+mem.numa.hugepages.free
+    inst [N or "node0::1048576kB"] value 104857600
+    inst [N or "node0::2048kB"] value 4096000
+    inst [N or "node0::32768kB"] value 0
+    inst [N or "node0::64kB"] value 0
+
+mem.numa.hugepages.pagesize
+    inst [N or "node0::1048576kB"] value 1048576
+    inst [N or "node0::2048kB"] value 2048
+    inst [N or "node0::32768kB"] value 32768
+    inst [N or "node0::64kB"] value 64
+
+mem.numa.hugepages.surplus
+    inst [N or "node0::1048576kB"] value 0
+    inst [N or "node0::2048kB"] value 8192
+    inst [N or "node0::32768kB"] value 0
+    inst [N or "node0::64kB"] value 0
+
+mem.numa.hugepages.totalsize
+    inst [N or "node0::1048576kB"] value 104857600
+    inst [N or "node0::2048kB"] value 4096000
+    inst [N or "node0::32768kB"] value 0
+    inst [N or "node0::64kB"] value 0
+
 == done
 

qa/821.out

@@ -137,6 +137,18 @@ mem.numa.alloc.miss
 mem.numa.alloc.other_node
     inst [0 or "node0"] value 0
 
+mem.numa.hugepages.free
+No value(s) available!
+
+mem.numa.hugepages.pagesize
+No value(s) available!
+
+mem.numa.hugepages.surplus
+No value(s) available!
+
+mem.numa.hugepages.totalsize
+No value(s) available!
+
 mem.numa.max_bandwidth
 No value(s) available!
 
@@ -2933,6 +2945,18 @@ mem.numa.alloc.other_node
     inst [0 or "node0"] value 11838
     inst [1 or "node1"] value 53630
 
+mem.numa.hugepages.free
+No value(s) available!
+
+mem.numa.hugepages.pagesize
+No value(s) available!
+
+mem.numa.hugepages.surplus
+No value(s) available!
+
+mem.numa.hugepages.totalsize
+No value(s) available!
+
 mem.numa.max_bandwidth
 No value(s) available!
 
@@ -5738,6 +5762,18 @@ mem.numa.alloc.other_node
     inst [0 or "node0"] value 11838
     inst [1 or "node1"] value 53630
 
+mem.numa.hugepages.free
+No value(s) available!
+
+mem.numa.hugepages.pagesize
+No value(s) available!
+
+mem.numa.hugepages.surplus
+No value(s) available!
+
+mem.numa.hugepages.totalsize
+No value(s) available!
+
 mem.numa.max_bandwidth
 No value(s) available!
 
@@ -8537,6 +8573,18 @@ mem.numa.alloc.miss
 mem.numa.alloc.other_node
     inst [0 or "node0"] value 0
 
+mem.numa.hugepages.free
+No value(s) available!
+
+mem.numa.hugepages.pagesize
+No value(s) available!
+
+mem.numa.hugepages.surplus
+No value(s) available!
+
+mem.numa.hugepages.totalsize
+No value(s) available!
+
 mem.numa.max_bandwidth
 No value(s) available!
 
@@ -11073,6 +11121,18 @@ mem.numa.alloc.miss
 mem.numa.alloc.other_node
     inst [0 or "node0"] value 0
 
+mem.numa.hugepages.free
+No value(s) available!
+
+mem.numa.hugepages.pagesize
+No value(s) available!
+
+mem.numa.hugepages.surplus
+No value(s) available!
+
+mem.numa.hugepages.totalsize
+No value(s) available!
+
 mem.numa.max_bandwidth
 No value(s) available!
 
@@ -13673,6 +13733,18 @@ mem.numa.alloc.miss
 mem.numa.alloc.other_node
     inst [0 or "node0"] value 0
 
+mem.numa.hugepages.free
+No value(s) available!
+
+mem.numa.hugepages.pagesize
+No value(s) available!
+
+mem.numa.hugepages.surplus
+No value(s) available!
+
+mem.numa.hugepages.totalsize
+No value(s) available!
+
 mem.numa.max_bandwidth
 No value(s) available!
 
@@ -15551,6 +15623,18 @@ mem.numa.alloc.other_node
     inst [0 or "node0"] value 847
     inst [1 or "node1"] value 492232
 
+mem.numa.hugepages.free
+No value(s) available!
+
+mem.numa.hugepages.pagesize
+No value(s) available!
+
+mem.numa.hugepages.surplus
+No value(s) available!
+
+mem.numa.hugepages.totalsize
+No value(s) available!
+
 mem.numa.max_bandwidth
 No value(s) available!
 

src/pmdas/linux/help

@@ -3,7 +3,7 @@
 # Portions Copyright (c) International Business Machines Corp., 2002
 # Portions Copyright (c) 2007-2009 Aconex.  All Rights Reserved.
 # Portions Copyright (c) 2016-2017 Fujitsu.
-# Portions Copyright (c) 2013-2021,2023-2024 Red Hat.
+# Portions Copyright (c) 2013-2021,2023-2025 Red Hat.
 #
 # This program is free software; you can redistribute it and/or modify it
 # under the terms of the GNU General Public License as published by the
@@ -1041,6 +1041,11 @@ filled up manually after running some bandwidth saturation benchmark tools.
 @ mem.numa.util.hugepagesTotalBytes per-node total amount of hugepages memory
 @ mem.numa.util.hugepagesFreeBytes per-node amount of free hugepages memory
 @ mem.numa.util.hugepagesSurpBytes per-node amount of surplus hugepages memory
+@ mem.numa.hugepages.pagesize per-node available huge page sizes
+@ mem.numa.hugepages.free per-node huge page space not yet allocated
+@ mem.numa.hugepages.surplus per-node huge page space in the pool above total size
+@ mem.numa.hugepages.totalsize per-node memory consumed by each huge page size
+
 @ mem.vmstat.nr_dirty number of pages in dirty state
 Instantaneous number of pages in dirty state, from /proc/vmstat
 @ mem.vmstat.nr_dirty_background_threshold background writeback threshold

src/pmdas/linux/linux.h

@@ -116,6 +116,7 @@ enum {
 	CLUSTER_WWID,		/* 92 multipath aggregated stats */
 	CLUSTER_PRESSURE_IRQ,	/* 93 /proc/pressure/irq metrics */
 	CLUSTER_HUGEPAGES,	/* 94 /sys/kernel/mm/hugepages metrics */
+	CLUSTER_NUMA_HUGEPAGES,	/* 95 /sys/devices/system/node/nodeN/hugepages metrics */
 
 	NUM_CLUSTERS		/* one more than highest numbered cluster */
 };
@@ -197,6 +198,7 @@ enum {
 	SOFTIRQ_CPU_INDOM,	/* 41 - per-CPU soft IRQs */
 	WWID_INDOM,		/* 42 - per-WWID multipath device */
 	HUGEPAGES_INDOM,	/* 43 - hugepages (fixed sizes) */
+	NUMA_HUGEPAGES_INDOM,	/* 44 - NUMA hugepages (fixed sizes) */
 
 	NUM_INDOMS		/* one more than highest numbered cluster */
 };

src/pmdas/linux/linux_table.c

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2012 Red Hat.
+ * Copyright (c) 2012,2025 Red Hat.
  * Copyright (c) 2004 Silicon Graphics, Inc.  All Rights Reserved.
  *
  * This program is free software; you can redistribute it and/or modify it
@@ -46,16 +46,17 @@ linux_table_clone(struct linux_table *table)
 {
     struct linux_table *ret;
     struct linux_table *t;
-    int len;
+    size_t len, bytes;
 
     if (!table)
 	return NULL;
     for (len=1, t=table; t->field; t++)
     	len++;
-    ret = (struct linux_table *)malloc(len * sizeof(struct linux_table));
+    bytes = len * sizeof(struct linux_table);
+    ret = (struct linux_table *)malloc(bytes);
     if (!ret)
 	return NULL;
-    memcpy(ret, table, len * sizeof(struct linux_table));
+    memcpy(ret, table, bytes);
 
     /* Initialize the table */
     for (t=ret; t && t->field; t++) {
@@ -75,7 +76,7 @@ linux_table_scan(FILE *fp, struct linux_table *table)
     char buf[1024];
     int ret = 0;
 
-    while(fgets(buf, sizeof(buf), fp) != NULL) {
+    while (fgets(buf, sizeof(buf), fp) != NULL) {
 	for (t=table; t && t->field; t++) {
 	    if ((p = strstr(buf, t->field)) != NULL) {
 		/* first digit after the matched field */
@@ -95,7 +96,7 @@ linux_table_scan(FILE *fp, struct linux_table *table)
 
     /* calculate current value, accounting for counter wrap */
     for (t=table; t && t->field; t++) {
-    	if (t->maxval == 0)
+	if (t->maxval == 0)
 	    /* instantaneous value */
 	    t->val = t->this;
 	else {