2011-03-05 16:06:52

by Guenter Roeck

[permalink] [raw]
Subject: [PATCH] hwmon: (pmbus) Improve support for paged fans

So far, it seemed like fans would either all be in page 0, or that
there would be one page per fan.

Turns out this was a wrong assumption. There is at least one PMBus
fan controller which supports three pages with four fans each.
Update code to handle this situation.

Reported-by: Greg Schnorr <[email protected]>
Signed-off-by: Guenter Roeck <[email protected]>
---
drivers/hwmon/pmbus_core.c | 70 +++++++++++++++++++++++++-------------------
1 files changed, 40 insertions(+), 30 deletions(-)

diff --git a/drivers/hwmon/pmbus_core.c b/drivers/hwmon/pmbus_core.c
index b7c64ba..d025a11 100644
--- a/drivers/hwmon/pmbus_core.c
+++ b/drivers/hwmon/pmbus_core.c
@@ -58,12 +58,10 @@
#define PMBUS_MAX_INPUT_LABELS 4 /* vin, vcap, iin, pin */

/*
- * status, status_vout, status_iout, status_fans, and status_temp
- * are paged. status_input and status_fan34 are unpaged.
- * status_fan34 is a special case to handle a second set of fans
- * on page 0.
+ * status, status_vout, status_iout, status_fans, status_fan34, and status_temp
+ * are paged. status_input is unpaged.
*/
-#define PB_NUM_STATUS_REG (PMBUS_PAGES * 5 + 2)
+#define PB_NUM_STATUS_REG (PMBUS_PAGES * 6 + 1)

/*
* Index into status register array, per status register group
@@ -73,7 +71,7 @@
#define PB_STATUS_IOUT_BASE (PB_STATUS_VOUT_BASE + PMBUS_PAGES)
#define PB_STATUS_FAN_BASE (PB_STATUS_IOUT_BASE + PMBUS_PAGES)
#define PB_STATUS_FAN34_BASE (PB_STATUS_FAN_BASE + PMBUS_PAGES)
-#define PB_STATUS_INPUT_BASE (PB_STATUS_FAN34_BASE + 1)
+#define PB_STATUS_INPUT_BASE (PB_STATUS_FAN34_BASE + PMBUS_PAGES)
#define PB_STATUS_TEMP_BASE (PB_STATUS_INPUT_BASE + 1)

struct pmbus_sensor {
@@ -327,14 +325,17 @@ static struct pmbus_data *pmbus_update_device(struct device *dev)
= pmbus_get_status(client, i, PMBUS_STATUS_FAN_12);
}

+ for (i = 0; i < info->pages; i++) {
+ if (!(info->func[i] & PMBUS_HAVE_STATUS_FAN34))
+ continue;
+ data->status[PB_STATUS_FAN34_BASE + i]
+ = pmbus_get_status(client, i, PMBUS_STATUS_FAN_34);
+ }
+
if (info->func[0] & PMBUS_HAVE_STATUS_INPUT)
data->status[PB_STATUS_INPUT_BASE]
= pmbus_get_status(client, 0, PMBUS_STATUS_INPUT);

- if (info->func[0] & PMBUS_HAVE_STATUS_FAN34)
- data->status[PB_STATUS_FAN34_BASE]
- = pmbus_get_status(client, 0, PMBUS_STATUS_FAN_34);
-
for (i = 0; i < data->num_sensors; i++) {
struct pmbus_sensor *sensor = &data->sensors[i];

@@ -817,6 +818,20 @@ static const int pmbus_fan_status_registers[] = {
PMBUS_STATUS_FAN_34
};

+static const u32 pmbus_fan_flags[] = {
+ PMBUS_HAVE_FAN12,
+ PMBUS_HAVE_FAN12,
+ PMBUS_HAVE_FAN34,
+ PMBUS_HAVE_FAN34
+};
+
+static const u32 pmbus_fan_status_flags[] = {
+ PMBUS_HAVE_STATUS_FAN12,
+ PMBUS_HAVE_STATUS_FAN12,
+ PMBUS_HAVE_STATUS_FAN34,
+ PMBUS_HAVE_STATUS_FAN34
+};
+
/*
* Determine maximum number of sensors, booleans, and labels.
* To keep things simple, only make a rough high estimate.
@@ -848,17 +863,12 @@ static void pmbus_find_max_attr(struct i2c_client *client,
max_labels++;
}
if (info->func[page] & PMBUS_HAVE_FAN12) {
- if (page == 0) {
- max_sensors +=
- ARRAY_SIZE(pmbus_fan_registers) *
- PMBUS_MAX_SENSORS_PER_FAN;
- max_booleans +=
- ARRAY_SIZE(pmbus_fan_registers) *
- PMBUS_MAX_BOOLEANS_PER_FAN;
- } else {
- max_sensors += PMBUS_MAX_SENSORS_PER_FAN;
- max_booleans += PMBUS_MAX_BOOLEANS_PER_FAN;
- }
+ max_sensors += 2 * PMBUS_MAX_SENSORS_PER_FAN;
+ max_booleans += 2 * PMBUS_MAX_BOOLEANS_PER_FAN;
+ }
+ if (info->func[page] & PMBUS_HAVE_FAN34) {
+ max_sensors += 2 * PMBUS_MAX_SENSORS_PER_FAN;
+ max_booleans += 2 * PMBUS_MAX_BOOLEANS_PER_FAN;
}
if (info->func[page] & PMBUS_HAVE_TEMP) {
if (page == 0) {
@@ -1365,15 +1375,14 @@ static void pmbus_find_attributes(struct i2c_client *client,
*/
in_index = 1;
for (page = 0; page < info->pages; page++) {
- int fans, f;
+ int f;

- if (!(info->func[page] & PMBUS_HAVE_FAN12))
- continue;
-
- fans = page ? 1 : ARRAY_SIZE(pmbus_fan_registers);
- for (f = 0; f < fans; f++) {
+ for (f = 0; f < ARRAY_SIZE(pmbus_fan_registers); f++) {
int regval;

+ if (!(info->func[page] & pmbus_fan_flags[f]))
+ break;
+
if (!pmbus_check_word_register(client, page,
pmbus_fan_registers[f])
|| !pmbus_check_byte_register(client, page,
@@ -1399,12 +1408,13 @@ static void pmbus_find_attributes(struct i2c_client *client,
* Each fan status register covers multiple fans,
* so we have to do some magic.
*/
- if (pmbus_check_byte_register
- (client, page, pmbus_fan_status_registers[f])) {
+ if ((info->func[page] & pmbus_fan_status_flags[f]) &&
+ pmbus_check_byte_register(client,
+ page, pmbus_fan_status_registers[f])) {
int base;

if (f > 1) /* fan 3, 4 */
- base = PB_STATUS_FAN34_BASE;
+ base = PB_STATUS_FAN34_BASE + page;
else
base = PB_STATUS_FAN_BASE + page;
pmbus_add_boolean_reg(data, "fan", "alarm",
--
1.7.3.1


2011-03-07 21:42:37

by Greg Schnorr

[permalink] [raw]
Subject: Re: [PATCH] hwmon: (pmbus) Improve support for paged fans

Verified this patch works as expected on such "fan rich" hardware...

On 3/5/11 8:08 AM, "Guenter Roeck" <[email protected]> wrote:

> So far, it seemed like fans would either all be in page 0, or that
> there would be one page per fan.
>
> Turns out this was a wrong assumption. There is at least one PMBus
> fan controller which supports three pages with four fans each.
> Update code to handle this situation.
>
> Reported-by: Greg Schnorr <[email protected]>
> Signed-off-by: Guenter Roeck <[email protected]>
Acked-by: Greg Schnorr <[email protected]>

2011-03-07 22:10:28

by Guenter Roeck

[permalink] [raw]
Subject: Re: [PATCH] hwmon: (pmbus) Improve support for paged fans

On Mon, Mar 07, 2011 at 04:42:37PM -0500, Greg Schnorr wrote:
> Verified this patch works as expected on such "fan rich" hardware...
>
> On 3/5/11 8:08 AM, "Guenter Roeck" <[email protected]> wrote:
>
> > So far, it seemed like fans would either all be in page 0, or that
> > there would be one page per fan.
> >
> > Turns out this was a wrong assumption. There is at least one PMBus
> > fan controller which supports three pages with four fans each.
> > Update code to handle this situation.
> >
> > Reported-by: Greg Schnorr <[email protected]>
> > Signed-off-by: Guenter Roeck <[email protected]>
> Acked-by: Greg Schnorr <[email protected]>
>
Thanks!

Guenter