powerpc/numa: Perform full re-add of CPU for PRRN/VPHN topology update
diff mbox series

Message ID 7c3c840a-8a79-b639-185c-ce6d2e378850@linux.vnet.ibm.com
State New
Headers show
Series
  • powerpc/numa: Perform full re-add of CPU for PRRN/VPHN topology update
Related show

Checks

Context Check Description
snowpatch_ozlabs/snowpatch_job_snowpatch-linux-checkpatch warning total: 0 errors, 2 warnings, 4 checks, 253 lines checked
snowpatch_ozlabs/apply_patch success next/apply_patch Successfully applied

Commit Message

Michael Bringmann Nov. 12, 2018, 6:53 p.m. UTC
On pseries systems, performing changes to a partition's affinity
can result in altering the nodes a CPU is assigned to the
current system.  For example, some systems are subject to resource
balancing operations by the operator or control software.  In such
environments, system CPUs may be in node 1 and 3 at boot, and be
moved to nodes 2, 3, and 5, for better performance.

The current implementation attempts to recognize such changes within
the powerpc-specific version of arch_update_cpu_topology to modify a
range of system data structures directly.  However, some scheduler
data structures may be inaccessible, or the timing of a node change
may still lead to corruption or error in other modules (e.g. user
space) which do not receive notification of these changes.

This patch modifies the PRRN/VPHN topology update worker function to
recognize an affinity change for a CPU, and to perform a full DLPAR
remove and add of the CPU instead of dynamically changing its node
to resolve this issue.

[Based upon patch submission:
Subject: [PATCH] powerpc/pseries: Perform full re-add of CPU for topology update post-migration
From: Nathan Fontenot <nfont@linux.vnet.ibm.com>
Date: Tue Oct 30 05:43:36 AEDT 2018
]

[Replace patch submission:
Subject: [PATCH] powerpc/topology: Update numa mask when cpu node mapping changes
From: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
DAte: Wed Oct 10 15:24:46 AEDT 2018
]

Signed-off-by: Michael Bringmann <mwb@linux.vnet.ibm.com>
---
 arch/powerpc/include/asm/topology.h |    6 -
 arch/powerpc/kernel/rtasd.c         |    1 
 arch/powerpc/mm/numa.c              |  184 +++++------------------------------
 3 files changed, 27 insertions(+), 164 deletions(-)

Comments

Srikar Dronamraju Nov. 13, 2018, 8:39 a.m. UTC | #1
> -static void topology_work_fn(struct work_struct *work)
> -{
> -	rebuild_sched_domains();
> +	if (changed)
> +		rebuild_sched_domains();
>  }
>  static DECLARE_WORK(topology_work, topology_work_fn);
> 
> @@ -1553,7 +1424,6 @@ void __init shared_proc_topology_init(void)
>  	if (lppaca_shared_proc(get_lppaca())) {
>  		bitmap_fill(cpumask_bits(&cpu_associativity_changes_mask),
>  			    nr_cpumask_bits);
> -		numa_update_cpu_topology(false);

Shouldn't we be calling topology_schedule_update() here?

>  	}
>  }
>
Michael Bringmann Nov. 14, 2018, 2:13 p.m. UTC | #2
On 11/13/2018 02:39 AM, Srikar Dronamraju wrote:
>> -static void topology_work_fn(struct work_struct *work)
>> -{
>> -	rebuild_sched_domains();
>> +	if (changed)
>> +		rebuild_sched_domains();
>>  }
>>  static DECLARE_WORK(topology_work, topology_work_fn);
>>
>> @@ -1553,7 +1424,6 @@ void __init shared_proc_topology_init(void)
>>  	if (lppaca_shared_proc(get_lppaca())) {
>>  		bitmap_fill(cpumask_bits(&cpu_associativity_changes_mask),
>>  			    nr_cpumask_bits);
>> -		numa_update_cpu_topology(false);
> 
> Shouldn't we be calling topology_schedule_update() here?

Agreed.

> 
>>  	}
>>  }
>>
> 
>

Patch
diff mbox series

diff --git a/arch/powerpc/include/asm/topology.h b/arch/powerpc/include/asm/topology.h
index f85e2b0..9f85246 100644
--- a/arch/powerpc/include/asm/topology.h
+++ b/arch/powerpc/include/asm/topology.h
@@ -42,7 +42,6 @@  static inline int pcibus_to_node(struct pci_bus *bus)
 
 extern int sysfs_add_device_to_node(struct device *dev, int nid);
 extern void sysfs_remove_device_from_node(struct device *dev, int nid);
-extern int numa_update_cpu_topology(bool cpus_locked);
 
 static inline void update_numa_cpu_lookup_table(unsigned int cpu, int node)
 {
@@ -77,11 +76,6 @@  static inline void sysfs_remove_device_from_node(struct device *dev,
 {
 }
 
-static inline int numa_update_cpu_topology(bool cpus_locked)
-{
-	return 0;
-}
-
 static inline void update_numa_cpu_lookup_table(unsigned int cpu, int node) {}
 
 #endif /* CONFIG_NUMA */
diff --git a/arch/powerpc/kernel/rtasd.c b/arch/powerpc/kernel/rtasd.c
index 38cadae..c161d74 100644
--- a/arch/powerpc/kernel/rtasd.c
+++ b/arch/powerpc/kernel/rtasd.c
@@ -285,7 +285,6 @@  static void handle_prrn_event(s32 scope)
 	 * the RTAS event.
 	 */
 	pseries_devicetree_update(-scope);
-	numa_update_cpu_topology(false);
 }
 
 static void handle_rtas_event(const struct rtas_error_log *log)
diff --git a/arch/powerpc/mm/numa.c b/arch/powerpc/mm/numa.c
index be6216e..f79b65f 100644
--- a/arch/powerpc/mm/numa.c
+++ b/arch/powerpc/mm/numa.c
@@ -1236,96 +1236,25 @@  int find_and_online_cpu_nid(int cpu)
 	return new_nid;
 }
 
-/*
- * Update the CPU maps and sysfs entries for a single CPU when its NUMA
- * characteristics change. This function doesn't perform any locking and is
- * only safe to call from stop_machine().
- */
-static int update_cpu_topology(void *data)
-{
-	struct topology_update_data *update;
-	unsigned long cpu;
-
-	if (!data)
-		return -EINVAL;
-
-	cpu = smp_processor_id();
-
-	for (update = data; update; update = update->next) {
-		int new_nid = update->new_nid;
-		if (cpu != update->cpu)
-			continue;
-
-		unmap_cpu_from_node(cpu);
-		map_cpu_to_node(cpu, new_nid);
-		set_cpu_numa_node(cpu, new_nid);
-		set_cpu_numa_mem(cpu, local_memory_node(new_nid));
-		vdso_getcpu_init();
-	}
-
-	return 0;
-}
-
-static int update_lookup_table(void *data)
-{
-	struct topology_update_data *update;
-
-	if (!data)
-		return -EINVAL;
-
-	/*
-	 * Upon topology update, the numa-cpu lookup table needs to be updated
-	 * for all threads in the core, including offline CPUs, to ensure that
-	 * future hotplug operations respect the cpu-to-node associativity
-	 * properly.
-	 */
-	for (update = data; update; update = update->next) {
-		int nid, base, j;
-
-		nid = update->new_nid;
-		base = cpu_first_thread_sibling(update->cpu);
-
-		for (j = 0; j < threads_per_core; j++) {
-			update_numa_cpu_lookup_table(base + j, nid);
-		}
-	}
-
-	return 0;
-}
-
-/*
- * Update the node maps and sysfs entries for each cpu whose home node
- * has changed. Returns 1 when the topology has changed, and 0 otherwise.
- *
- * cpus_locked says whether we already hold cpu_hotplug_lock.
- */
-int numa_update_cpu_topology(bool cpus_locked)
+static void topology_work_fn(struct work_struct *work)
 {
 	unsigned int cpu, sibling, changed = 0;
-	struct topology_update_data *updates, *ud;
-	cpumask_t updated_cpus;
-	struct device *dev;
-	int weight, new_nid, i = 0;
+	int weight;
 
 	if (!prrn_enabled && !vphn_enabled && topology_inited)
-		return 0;
+		return;
 
 	weight = cpumask_weight(&cpu_associativity_changes_mask);
 	if (!weight)
-		return 0;
+		return;
 
-	updates = kcalloc(weight, sizeof(*updates), GFP_KERNEL);
-	if (!updates)
-		return 0;
+	pr_debug("Topology update CPUs beginning\n");
 
-	cpumask_clear(&updated_cpus);
+	lock_device_hotplug();
 
 	for_each_cpu(cpu, &cpu_associativity_changes_mask) {
-		/*
-		 * If siblings aren't flagged for changes, updates list
-		 * will be too short. Skip on this update and set for next
-		 * update.
-		 */
+		int new_nid;
+
 		if (!cpumask_subset(cpu_sibling_mask(cpu),
 					&cpu_associativity_changes_mask)) {
 			pr_info("Sibling bits not set for associativity "
@@ -1337,9 +1266,11 @@  int numa_update_cpu_topology(bool cpus_locked)
 			continue;
 		}
 
+		/* Use associativity from first thread for all siblings */
 		new_nid = find_and_online_cpu_nid(cpu);
 
-		if (new_nid == numa_cpu_lookup_table[cpu]) {
+		if ((new_nid == numa_cpu_lookup_table[cpu]) ||
+			!cpu_present(cpu)) {
 			cpumask_andnot(&cpu_associativity_changes_mask,
 					&cpu_associativity_changes_mask,
 					cpu_sibling_mask(cpu));
@@ -1349,89 +1280,29 @@  int numa_update_cpu_topology(bool cpus_locked)
 			continue;
 		}
 
-		for_each_cpu(sibling, cpu_sibling_mask(cpu)) {
-			ud = &updates[i++];
-			ud->next = &updates[i];
-			ud->cpu = sibling;
-			ud->new_nid = new_nid;
-			ud->old_nid = numa_cpu_lookup_table[sibling];
-			cpumask_set_cpu(sibling, &updated_cpus);
-		}
-		cpu = cpu_last_thread_sibling(cpu);
-	}
+		pr_debug("Topology update for cpu %d\n", cpu);
+		dlpar_cpu_readd(cpu);
+		changed++;
 
-	/*
-	 * Prevent processing of 'updates' from overflowing array
-	 * where last entry filled in a 'next' pointer.
-	 */
-	if (i)
-		updates[i-1].next = NULL;
-
-	pr_debug("Topology update for the following CPUs:\n");
-	if (cpumask_weight(&updated_cpus)) {
-		for (ud = &updates[0]; ud; ud = ud->next) {
-			pr_debug("cpu %d moving from node %d "
-					  "to %d\n", ud->cpu,
-					  ud->old_nid, ud->new_nid);
+		for_each_cpu(sibling, cpu_sibling_mask(cpu)) {
+			unmap_cpu_from_node(sibling);
+			map_cpu_to_node(sibling, new_nid);
+			set_cpu_numa_node(sibling, new_nid);
+			set_cpu_numa_mem(sibling,
+				local_memory_node(new_nid));
+			cpumask_clear_cpu(sibling,
+				&cpu_associativity_changes_mask);
 		}
-	}
-
-	/*
-	 * In cases where we have nothing to update (because the updates list
-	 * is too short or because the new topology is same as the old one),
-	 * skip invoking update_cpu_topology() via stop-machine(). This is
-	 * necessary (and not just a fast-path optimization) since stop-machine
-	 * can end up electing a random CPU to run update_cpu_topology(), and
-	 * thus trick us into setting up incorrect cpu-node mappings (since
-	 * 'updates' is kzalloc()'ed).
-	 *
-	 * And for the similar reason, we will skip all the following updating.
-	 */
-	if (!cpumask_weight(&updated_cpus))
-		goto out;
 
-	if (cpus_locked)
-		stop_machine_cpuslocked(update_cpu_topology, &updates[0],
-					&updated_cpus);
-	else
-		stop_machine(update_cpu_topology, &updates[0], &updated_cpus);
-
-	/*
-	 * Update the numa-cpu lookup table with the new mappings, even for
-	 * offline CPUs. It is best to perform this update from the stop-
-	 * machine context.
-	 */
-	if (cpus_locked)
-		stop_machine_cpuslocked(update_lookup_table, &updates[0],
-					cpumask_of(raw_smp_processor_id()));
-	else
-		stop_machine(update_lookup_table, &updates[0],
-			     cpumask_of(raw_smp_processor_id()));
-
-	for (ud = &updates[0]; ud; ud = ud->next) {
-		unregister_cpu_under_node(ud->cpu, ud->old_nid);
-		register_cpu_under_node(ud->cpu, ud->new_nid);
-
-		dev = get_cpu_device(ud->cpu);
-		if (dev)
-			kobject_uevent(&dev->kobj, KOBJ_CHANGE);
-		cpumask_clear_cpu(ud->cpu, &cpu_associativity_changes_mask);
-		changed = 1;
+		cpu = cpu_last_thread_sibling(cpu);
 	}
 
-out:
-	kfree(updates);
-	return changed;
-}
+	unlock_device_hotplug();
 
-int arch_update_cpu_topology(void)
-{
-	return numa_update_cpu_topology(true);
-}
+	pr_debug("Topology update CPUs ending\n");
 
-static void topology_work_fn(struct work_struct *work)
-{
-	rebuild_sched_domains();
+	if (changed)
+		rebuild_sched_domains();
 }
 static DECLARE_WORK(topology_work, topology_work_fn);
 
@@ -1553,7 +1424,6 @@  void __init shared_proc_topology_init(void)
 	if (lppaca_shared_proc(get_lppaca())) {
 		bitmap_fill(cpumask_bits(&cpu_associativity_changes_mask),
 			    nr_cpumask_bits);
-		numa_update_cpu_topology(false);
 	}
 }