Skip to contentSkip to conversation

regmap: Add support for continously numbered pages across regmap range

PATCH In review Mukund Navada 2 participants 8 messages Nov 28, 2012

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 regmap: Add support for continously numbered pages across regmap range

Page number in some devices increases across regmap ranges since the
selector register to choose the page remains the same. So in order to write correct page number, the start_page_num is stored in regmap_range_cfg.

---
 drivers/base/regmap/internal.h |    2 ++
 drivers/base/regmap/regmap.c   |    2 ++
 include/linux/regmap.h         |    5 +++++
 3 files changed, 9 insertions(+), 0 deletions(-)
drivers/base/regmap/internal.h +2 -0
index 288e135..fd31f5d 100644
@@ -134,6 +134,8 @@ struct regmap_range_node {
 
 	unsigned int window_start;
 	unsigned int window_len;
+
+	unsigned int start_page_num;
 };
 
 #ifdef CONFIG_DEBUG_FS
drivers/base/regmap/regmap.c +2 -0
index 42d5cb0..39e37a4 100644
@@ -645,6 +645,7 @@ struct regmap *regmap_init(struct device *dev,
 		new->selector_shift = range_cfg->selector_shift;
 		new->window_start = range_cfg->window_start;
 		new->window_len = range_cfg->window_len;
+		new->start_page_num = range_cfg->start_page_num;
 
 		if (_regmap_range_add(map, new) == false) {
 			dev_err(map->dev, "Failed to add range %d\n", i);
@@ -833,6 +834,7 @@ static int _regmap_select_page(struct regmap *map, unsigned int *reg,
 
 	win_offset = (*reg - range->range_min) % range->window_len;
 	win_page = (*reg - range->range_min) / range->window_len;
+	win_page += range->start_page_num;
 
 	if (val_num > 1) {
 		/* Bulk write shouldn't cross range boundary */
include/linux/regmap.h +5 -0
index b7e95bf..96f029e 100644
@@ -217,6 +217,8 @@ struct regmap_config {
  *
  * @window_start: Address of first (lowest) register in data window.
  * @window_len: Number of registers in data window.
+ *
+ * @start_page_num: Start page number for the range
  */
 struct regmap_range_cfg {
 	const char *name;
@@ -233,6 +235,9 @@ struct regmap_range_cfg {
 	/* Data window (per each page) */
 	unsigned int window_start;
 	unsigned int window_len;
+
+	/* Starting page number when it increases across range */
+	unsigned int start_page_num;
 };
 
 typedef int (*regmap_hw_write)(void *context, const void *data,
Signature · 1 line
1.7.1
#1 lore raw
Quoted 4 lines · Mukund Navada

Page number in some devices increases across regmap ranges since the
selector register to choose the page remains the same. So in order to write correct page number, the start_page_num is stored in regmap_range_cfg.

I'm sorry but I can't understand what this is supposed to do. The current paging code expects pages to be numbered continuously with no gaps. Can you please clarify?

#2 lore raw
Quoted 11 lines

-----Original Message-----
From: Mark Brown [mailto:[email protected]]
Sent: Wednesday, November 28, 2012 2:59 PM
To: Navada Kanyana, Mukund
Cc: Greg Kroah-Hartman; [email protected]
Subject: Re: [PATCH 1/1] regmap: Add support for continously numbered pages across regmap range

On Wed, Nov 28, 2012 at 02:52:57PM +0530, Mukund Navada wrote:

Quoted 8 lines
Quoted 4 lines

Page number in some devices increases across regmap ranges since the
selector register to choose the page remains the same. So in order to write correct page number, the start_page_num is stored in regmap_range_cfg.

I'm sorry but I can't understand what this is supposed to do. The current paging code expects pages to be numbered continuously with no gaps. Can you please clarify?

But the page number is expected to increase only till end of range. But in devices like TLV320AIC3256, the page number increases beyond range.

Example:

Pages 0 through Page 1 --> first register range for device configuration. Page 8 through Page 16 --> second range for DSP1 coefficient memory Page 26 through Page 34 --> third range for DSP2 coefficient memory

Window length of each page is 128. So if we need to correctly write the page number in selector register for second range, start_page_num would be 8. If we did not have start_page_num, it would write 0 to selector register.

#3 lore raw

On Wed, Nov 28, 2012 at 10:11:05AM +0000, Navada Kanyana, Mukund wrote:

Fix your mailer to word wrap within paragraphs.

Quoted 7 lines
Quoted 3 lines

I'm sorry but I can't understand what this is supposed to do. The current paging code expects pages to be numbered continuously with no gaps. Can you please clarify?

But the page number is expected to increase only till end of range. But in devices like TLV320AIC3256, the page number increases beyond range.

What does this mean? How can we have a page beyond the end of the range?

Quoted 9 lines

Pages 0 through Page 1 --> first register range for device configuration. Page 8 through Page 16 --> second range for DSP1 coefficient memory Page 26 through Page 34 --> third range for DSP2 coefficient memory

Window length of each page is 128. So if we need to correctly write the page number in selector register for second range, start_page_num would be 8. If we did not have start_page_num, it would write 0 to selector register.

I'm sorry but I don't understand how that would happen? What breaks accessing page 8? This sounds like you're doing something wrong when you're trying ton use the paging.

#4 lore raw
Quoted 30 lines

-----Original Message-----
From: Mark Brown [mailto:[email protected]]
Sent: Wednesday, November 28, 2012 5:05 PM
To: Navada Kanyana, Mukund
Cc: Greg Kroah-Hartman; [email protected]
Subject: Re: [PATCH 1/1] regmap: Add support for continously numbered pages across regmap range

On Wed, Nov 28, 2012 at 10:11:05AM +0000, Navada Kanyana, Mukund wrote:

Fix your mailer to word wrap within paragraphs.

Quoted 7 lines
Quoted 3 lines

I'm sorry but I can't understand what this is supposed to do. The current paging code expects pages to be numbered continuously with no gaps. Can you please clarify?

But the page number is expected to increase only till end of range. But in devices like TLV320AIC3256, the page number increases beyond range.

What does this mean? How can we have a page beyond the end of the range?

Quoted 4 lines

Pages 0 through Page 1 --> first register range for device configuration. Page 8 through Page 16 --> second range for DSP1 coefficient memory Page 26 through Page 34 --> third range for DSP2 coefficient memory

Quoted 8 lines
Quoted 4 lines

Window length of each page is 128. So if we need to correctly write the page number in selector register for second range, start_page_num would be 8. If we did not have start_page_num, it would write 0 to selector register.

I'm sorry but I don't understand how that would happen? What breaks accessing page 8? This sounds like you're doing something wrong when you're trying ton use the paging.

I have my regmap_range_cfg as below

#define AIC325X_DSP_WINDOW_LEN 128
#define AIC325X_ADSP_CM_A_BASE (9 * AIC325X_DSP_WINDOW_LEN)

static const struct regmap_range_cfg aic325x_ranges[] = {
        { .name = "CONF", .range_min = 128,
          .range_max = 383,
          .selector_reg = 0x0,
          .selector_mask = 0xff,
          .selector_shift = 0,
          .window_start = 0, .window_len =128,
        },
        { .name = "MINIDSP_A_CM_A", .range_min = AIC325X_ADSP_CM_A_BASE,
          .range_max = AIC325X_ADSP_CM_A_BASE + 1152,
          .selector_reg = 0x0,
          .selector_mask = 0xff,
          .selector_shift = 0,
          .window_start = 0, .window_len =128,
        },
};

In _regmap_select_page function that calculates win_page

win_page = (*reg - range->range_min) / range->window_len;

So when my *reg is within MINIDSP_A_CM_A (that corresponds to Page 8), the above win_page calculation would be 0 instead of 8 for *reg value of AIC325X_ADSP_CM_A_BASE.

#5 lore raw
Quoted 8 lines · Mukund
static const struct regmap_range_cfg aic325x_ranges[] = {
        { .name = "CONF", .range_min = 128,
          .range_max = 383,
          .selector_reg = 0x0,
          .selector_mask = 0xff,
          .selector_shift = 0,
          .window_start = 0, .window_len =128,
        },
Quoted 8 lines
        { .name = "MINIDSP_A_CM_A", .range_min = AIC325X_ADSP_CM_A_BASE,
          .range_max = AIC325X_ADSP_CM_A_BASE + 1152,
          .selector_reg = 0x0,
          .selector_mask = 0xff,
          .selector_shift = 0,
          .window_start = 0, .window_len =128,
        },
};

You appear to have two different ranges with the same selector and window. I'm really not sure why you expect this to work... what happens when someone reads from the two different ranges simultaneously?

#6 lore raw
Quoted 29 lines

-----Original Message-----
From: Mark Brown [mailto:[email protected]]
Sent: Wednesday, November 28, 2012 5:27 PM
To: Navada Kanyana, Mukund
Cc: Greg Kroah-Hartman; [email protected]
Subject: Re: [PATCH 1/1] regmap: Add support for continously numbered pages across regmap range

Quoted 8 lines · Mukund
static const struct regmap_range_cfg aic325x_ranges[] = {
        { .name = "CONF", .range_min = 128,
          .range_max = 383,
          .selector_reg = 0x0,
          .selector_mask = 0xff,
          .selector_shift = 0,
          .window_start = 0, .window_len =128,
        },
Quoted 8 lines
        { .name = "MINIDSP_A_CM_A", .range_min = AIC325X_ADSP_CM_A_BASE,
          .range_max = AIC325X_ADSP_CM_A_BASE + 1152,
          .selector_reg = 0x0,
          .selector_mask = 0xff,
          .selector_shift = 0,
          .window_start = 0, .window_len =128,
        },
};
Quoted 3 lines

You appear to have two different ranges with the same selector and window. I'm really not sure why you expect this to work... what happens when someone reads from the two different ranges simultaneously?

The selector register has to written appropriate page number. Two different ranges cannot be accessed simultaneously without changing the page number in the selector register.

#7 lore raw

On Wed, Nov 28, 2012 at 12:10:21PM +0000, Navada Kanyana, Mukund wrote:

To repeat, fix your mailer to word wrap within paragraphs.

Quoted 7 lines
Quoted 3 lines

You appear to have two different ranges with the same selector and window. I'm really not sure why you expect this to work... what happens when someone reads from the two different ranges simultaneously?

The selector register has to written appropriate page number. Two different ranges cannot be accessed simultaneously without changing the page number in the selector register.

This would be exactly the problem, yes. In any case, this looks like user error. You shouldn't be using the same window twice.