On Sat, Feb 27, 2016 at 1:21 AM, David Daney [off-list ref] wrote:

On 02/26/2016 10:53 AM, Will Deacon wrote:
[...]

quoted

quoted

diff --git a/arch/arm64/mm/numa.c b/arch/arm64/mm/numa.c
new file mode 100644
index 0000000..604e886
--- /dev/null
+++ b/arch/arm64/mm/numa.c

@@ -0,0 +1,403 @@

[...]

quoted

quoted

+
+static int numa_off;
+static int numa_distance_cnt;
+static u8 *numa_distance;
+
+static __init int numa_parse_early_param(char *opt)
+{
+       if (!opt)
+               return -EINVAL;
+       if (!strncmp(opt, "off", 3)) {
+               pr_info("%s\n", "NUMA turned off");
+               numa_off = 1;
+       }
+       return 0;
+}
+early_param("numa", numa_parse_early_param);

Curious, but when is this option actually useful?

Good point.  I will remove that bit, it was used as an aid in debugging
while bringing up the patch set.

this is handy in debugging new platforms.
this boot argument option forces to boot as single node dummy system
adding all resources to node0.

quoted

quoted

+
+cpumask_var_t node_to_cpumask_map[MAX_NUMNODES];
+EXPORT_SYMBOL(node_to_cpumask_map);
+
+#ifdef CONFIG_DEBUG_PER_CPU_MAPS
+
+/*
+ * Returns a pointer to the bitmask of CPUs on Node 'node'.
+ */
+const struct cpumask *cpumask_of_node(int node)
+{
+       if (WARN_ON(node >= nr_node_ids))
+               return cpu_none_mask;
+
+       if (WARN_ON(node_to_cpumask_map[node] == NULL))
+               return cpu_online_mask;
+
+       return node_to_cpumask_map[node];
+}
+EXPORT_SYMBOL(cpumask_of_node);
+
+#endif
+
+static void map_cpu_to_node(unsigned int cpu, int nid)
+{
+       set_cpu_numa_node(cpu, nid);
+       if (nid >= 0)
+               cpumask_set_cpu(cpu, node_to_cpumask_map[nid]);
+}
+
+static void unmap_cpu_to_node(unsigned int cpu)
+{
+       int nid = cpu_to_node(cpu);
+
+       if (nid >= 0)
+               cpumask_clear_cpu(cpu, node_to_cpumask_map[nid]);
+       set_cpu_numa_node(cpu, NUMA_NO_NODE);
+}

How do you end up with negative nids this late in the game?

It might be possible with some of the hot plugging code.  It is a little
paranoia programming.

If you really don't like it, we can remove it.

quoted

quoted

+
+void numa_clear_node(unsigned int cpu)
+{
+       unmap_cpu_to_node(cpu);

Why don't you just inline this function?

Good point, I will do that.

[...]

quoted

quoted

+int __init numa_add_memblk(int nid, u64 start, u64 size)
+{
+       int ret;
+
+       ret = memblock_set_node(start, size, &memblock.memory, nid);
+       if (ret < 0) {
+               pr_err("NUMA: memblock [0x%llx - 0x%llx] failed to add on
node %d\n",
+                       start, (start + size - 1), nid);
+               return ret;
+       }
+
+       node_set(nid, numa_nodes_parsed);
+       pr_info("NUMA: Adding memblock [0x%llx - 0x%llx] on node %d\n",
+                       start, (start + size - 1), nid);
+       return ret;
+}
+EXPORT_SYMBOL(numa_add_memblk);

But this is marked __init... (and you've done this elsewhere in the patch
too).

I will fix these.

quoted

Will

Ganapat

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r@public.gmane.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel