On Wed, 5 Jul 2017 21:27:57 +0900 [off-list ref] wrote:

2017-07-05 21:20 GMT+09:00 Masahiro Yamada [off-list ref]:

quoted

2017-07-05 20:50 GMT+09:00 Kunihiko Hayashi [off-list ref]:

quoted

+
+#define TMOD                           0x0928
+#define TMOD_MASK                      GENMASK(9, 0)
+

 [ snip ]

quoted

+
+       /*
+        * The bit[8:0] of TMOD register represents 2's complement value
+        * of temperature in Celsius. Since bit8 of TMOD shows a sign bit,
+        * 32bit temperature value is obtained by sign extension.
+        */

Apparently, this comment does not match your code:

     #define TMOD_MASK                      GENMASK(9, 0)


TMOD_MASK is indicating bit[9:0].



Digging into the patch history, now I understood what happened.



In v1, you described
#define TMOD_MASK         0x1ff

This was correct.


In v2, you converted it into
#define TMOD_MASK          GENMASK(9, 0)

This was misconversion.  It should be GENMASK(8, 0)



Anyway, TMOD_MASK is not used any more.

That's true.
TMOD_MASK in v2/v3 is wrong, and not used no longer.

quoted

quoted

+       *out_temp = sign_extend32(temp, 8) * 1000;

Why magic number here?

   /* MSB of the TMOD field is a sign bit */
   *out_temp = sign_extend32(temp, TMOD_WIDTH) * 1000;

No.    sign_extend32(temp, TMOD_WIDTH - 1) or
       sign_extend32(temp, TMOD_MSB)  or whatever.

I see. I'll replace with the macro.

Best Regards,
Kunihiko Hayashi