Re: [PATCH v8 5/8] arm64: Use of_cpu_node_to_id helper for CPU topology parsing

On 17/10/17 17:20, Suzuki K Poulose wrote:

On 17/10/17 17:11, Will Deacon wrote:

quoted

On Tue, Oct 17, 2017 at 04:24:23PM +0100, Mark Rutland wrote:

quoted

On Tue, Oct 10, 2017 at 11:33:00AM +0100, Suzuki K Poulose wrote:

quoted

Make use of the new generic helper to convert an of_node of a CPU
to the logical CPU id in parsing the topology.

Cc: Catalin Marinas <redacted>
Cc: Leo Yan <redacted>
Cc: Will Deacon <redacted>
Cc: Mark Rutland <redacted>
Signed-off-by: Suzuki K Poulose <redacted>

This looks sane to me, but it will need an ack from Will or Catalin.

FWIW:

Acked-by: Mark Rutland <redacted>

Thanks,
Mark.

quoted

---
  arch/arm64/kernel/topology.c | 16 ++++++----------
  1 file changed, 6 insertions(+), 10 deletions(-)

diff --git a/arch/arm64/kernel/topology.c b/arch/arm64/kernel/topology.c
index 8d48b233e6ce..21868530018e 100644
--- a/arch/arm64/kernel/topology.c
+++ b/arch/arm64/kernel/topology.c

@@ -37,18 +37,14 @@ static int __init get_cpu_for_node(struct device_node *node)

      if (!cpu_node)
          return -1;
-    for_each_possible_cpu(cpu) {
-        if (of_get_cpu_node(cpu, NULL) == cpu_node) {
-            topology_parse_cpu_capacity(cpu_node, cpu);
-            of_node_put(cpu_node);
-            return cpu;
-        }
-    }
-
-    pr_crit("Unable to find CPU node for %pOF\n", cpu_node);
+    cpu = of_cpu_node_to_id(cpu_node);
+    if (cpu >= 0)
+        topology_parse_cpu_capacity(cpu_node, cpu);
+    else
+        pr_crit("Unable to find CPU node for %pOF\n", cpu_node);
      of_node_put(cpu_node);

This of_node_put is confusing me. Since of_cpu_node_to_id appears to be
balanced with its use of the node refcount, is this one intended to pair
with the earlier call to of_parse_phandle?

Yes.

  If so, does that mainline is

quoted

currently broken here because it doesn't drop the refcount twice for the
matching node?

No. This of_node_put is for the failure case where we couldn't match a CPU.
In the success case, it is dropped just before we return the result within
the loop.

As we discussed offline, there is indeed a missing of_node_put() for all the
nodes we loop through to match. But with this change, we have fixed that.
And I don't think it is worth sending to stable.

Cheers
Suzuki
--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html