From 9d33d1483d9cf99992ba1a5eae56838b3a02676f Mon Sep 17 00:00:00 2001
From: Dmitry Ilyin <6576495+widgetii@users.noreply.github.com>
Date: Sat, 10 Oct 2026 09:16:29 +0300
Subject: [PATCH] ingenic: restore the temperature HAL wiring, and report
 degrees not millidegrees

6e697a7 ("[INGENIC] Fix enable sensor clock") deleted the two lines in
setup_hal_ingenic() that assigned hal_temperature behind the same
access() guard the allwinner, rockchip, novatek, fh and gm HALs use.
Nothing was reported, because no Ingenic SoC up to and including T40 has
a die temperature sensor and so no Ingenic kernel registers a thermal
zone: the guard would have left hal_temperature NULL anyway, which is
also why ingenic_get_temp() has sat unreferenced ever since.

Being unreachable, it also kept two bugs. It returned the sysfs value
verbatim where every sibling HAL divides by 1000, so a 45 C chip would
have been reported as 45000, and on a failed read it returned -237.0,
which -t would have printed as a temperature instead of letting
gethwtemp()'s NAN produce the honest "Temperature cannot be retrieved".

Verified on the lab T31 (T31L, kernel 3.10.14 isvp_swan): /sys/class/
thermal does not exist there, so --temp still exits 1 with "Temperature
cannot be retrieved", unchanged from the stock build. The reader itself
was exercised against a real zone on the build host, 27800 -> 27.80, and
with the zone hidden in a private mount namespace -> NAN. Native, mips32
and arm32 builds clean; cYAML_test and reginfo_test pass.
---
 src/hal/ingenic.c | 22 ++++++++++++++++------
 1 file changed, 16 insertions(+), 6 deletions(-)

diff --git a/src/hal/ingenic.c b/src/hal/ingenic.c
index aad94600..492e5f96 100644
--- a/src/hal/ingenic.c
+++ b/src/hal/ingenic.c
@@ -1,5 +1,6 @@
 #include "hal/ingenic.h"
 
+#include <math.h>
 #include <stdlib.h>
 #include <string.h>
 
@@ -322,13 +323,13 @@ unsigned long ingenic_totalmem(unsigned long *media_mem) {
 }
 
 float ingenic_get_temp() {
-    float ret = -237.0;
     char buf[16];
-    if (line_from_file("/sys/class/thermal/thermal_zone0/temp", "(.+)", buf,
-                       sizeof(buf))) {
-        ret = strtof(buf, NULL);
-    }
-    return ret;
+    if (!line_from_file("/sys/class/thermal/thermal_zone0/temp", "(.+)", buf,
+                        sizeof(buf)))
+        return NAN;
+    /* Millidegrees, like everything under /sys/class/thermal. Returning the
+     * raw value would have reported a 45 C chip as 45000. */
+    return strtof(buf, NULL) / 1000;
 }
 
 /* Honours the adapter number instead of re-deriving it from the chip name on
@@ -378,6 +379,15 @@ void setup_hal_ingenic() {
     possible_i2c_addrs = ingenic_possible_i2c_addrs;
     i2c_adapter_nr = !strncmp(chip_name, "T40", 3) ? 1 : 0;
     open_i2c_sensor_fd = ingenic_open_i2c_fd;
+    /* Dropped by accident in 6e697a7, whose subject was the sensor clock. No
+     * Ingenic SoC up to and including T40 has a die temperature sensor, so no
+     * Ingenic kernel registers a thermal zone and the guard leaves
+     * hal_temperature NULL -- which is why nothing noticed. Kept in the shape
+     * the other sysfs-reading HALs use so that a kernel which ever grows a
+     * zone needs no change here, and so -t keeps answering honestly rather
+     * than through a function no caller can reach. */
+    if (!access("/sys/class/thermal/thermal_zone0/temp", R_OK))
+        hal_temperature = ingenic_get_temp;
     /* Also as a hook, because the call above only ever runs once: getchipname()
      * caches the chip id and returns before ever reaching here again. Anything
      * that gates the clock off afterwards — the vendor SDK does, on the way