Hi Peng,
On 04/01/2016 06:16, Peng Fan wrote:
To TO1.0, we can not rely on finish bit to read temperature. But to TO1.1, the issue was fixed by IC, we can rely on finish bit for temperature reading for TO1.1.
Signed-off-by: Peng Fan peng.fan@nxp.com Cc: Stefano Babic sbabic@denx.de Cc: Tim Harvey tharvey@gateworks.com Cc: Fabio Estevam fabio.estevam@freescale.com Cc: Adrian Alonso aalonso@freescale.com
drivers/thermal/imx_thermal.c | 30 +++++++++++++++++++++--------- 1 file changed, 21 insertions(+), 9 deletions(-)
diff --git a/drivers/thermal/imx_thermal.c b/drivers/thermal/imx_thermal.c index 09a3c52..45d5953 100644 --- a/drivers/thermal/imx_thermal.c +++ b/drivers/thermal/imx_thermal.c @@ -169,18 +169,30 @@ static int read_cpu_temperature(struct udevice *dev) writel(TEMPMON_HW_ANADIG_TEMPSENSE1_FINISHED_MASK, &ccm_anatop->tempsense1_clr); writel(TEMPMON_HW_ANADIG_TEMPSENSE1_MEASURE_TEMP_MASK, &ccm_anatop->tempsense1_set);
- start = get_timer(0);
- /* Wait max 100ms */
- do {
/** Since we can not rely on finish bit, use 1ms delay to get* temperature. From RM, 17us is enough to get data, but* to gurantee to get the data, delay 100ms here.*/
- if (soc_rev() >= CHIP_REV_1_1) {
/* make sure that the latest temp is valid */while ((readl(&ccm_anatop->tempsense1) &TEMPMON_HW_ANADIG_TEMPSENSE1_FINISHED_MASK) == 0)udelay(10000);
If 10ms is the minimal time, you should poll with a timer < 10mS, else you go out after 20mS. Processor is not doing anything, you can simply poll the register more often.
reg = readl(&ccm_anatop->tempsense1); tmp = (reg & TEMPMON_HW_ANADIG_TEMPSENSE1_TEMP_VALUE_MASK) >> TEMPMON_HW_ANADIG_TEMPSENSE1_TEMP_VALUE_SHIFT;
- } while (get_timer(0) < (start + 100));
- } else {
start = get_timer(0);/* Wait max 100ms */do {/** Since we can not rely on finish bit, use 100ms* delay to get temperature. From RM, 17us is* enough to get data, but to gurantee to get* the data, delay 100ms here.*/reg = readl(&ccm_anatop->tempsense1);tmp = (reg &TEMPMON_HW_ANADIG_TEMPSENSE1_TEMP_VALUE_MASK)>> TEMPMON_HW_ANADIG_TEMPSENSE1_TEMP_VALUE_SHIFT;} while (get_timer(0) < (start + 100));- }
This is nasty. You read the register, but you do not evaluate the result and you wait always 100mS. This has drawbacks for the boottime, because it is not so negligible. What is the reason to poll the register if you simply wait for timeout ?
Best regards, Stefano Babic