Some limit values are available in q16.q16 format, referred to as 32-bit unsigned ireal in CCS. Read these correctly. Signed-off-by: Sakari Ailus <sakari.ailus@xxxxxxxxxxxxxxx> --- drivers/media/i2c/ccs/ccs-core.c | 10 ++-------- drivers/media/i2c/ccs/ccs-reg-access.c | 23 +++++++++++++++++++++-- drivers/media/i2c/ccs/ccs.h | 9 +++++++++ 3 files changed, 32 insertions(+), 10 deletions(-) diff --git a/drivers/media/i2c/ccs/ccs-core.c b/drivers/media/i2c/ccs/ccs-core.c index 5b7f142b2432..cd8978c085d4 100644 --- a/drivers/media/i2c/ccs/ccs-core.c +++ b/drivers/media/i2c/ccs/ccs-core.c @@ -120,8 +120,8 @@ void ccs_replace_limit(struct ccs_sensor *sensor, ccs_assign_limit(ptr, ccs_reg_width(linfo->reg), val); } -static u32 ccs_get_limit(struct ccs_sensor *sensor, - unsigned int limit, unsigned int offset) +u32 ccs_get_limit(struct ccs_sensor *sensor, unsigned int limit, + unsigned int offset) { void *ptr; u32 val; @@ -149,12 +149,6 @@ static u32 ccs_get_limit(struct ccs_sensor *sensor, return ccs_reg_conv(sensor, ccs_limits[limit].reg, val); } -#define CCS_LIM(sensor, limit) \ - ccs_get_limit(sensor, CCS_L_##limit, 0) - -#define CCS_LIM_AT(sensor, limit, offset) \ - ccs_get_limit(sensor, CCS_L_##limit, CCS_L_##limit##_OFFSET(offset)) - static int ccs_read_all_limits(struct ccs_sensor *sensor) { struct i2c_client *client = v4l2_get_subdevdata(&sensor->src->sd); diff --git a/drivers/media/i2c/ccs/ccs-reg-access.c b/drivers/media/i2c/ccs/ccs-reg-access.c index fe6112cba6be..91ccbca11577 100644 --- a/drivers/media/i2c/ccs/ccs-reg-access.c +++ b/drivers/media/i2c/ccs/ccs-reg-access.c @@ -15,6 +15,7 @@ #include <linux/i2c.h> #include "ccs.h" +#include "ccs-limits.h" static uint32_t float_to_u32_mul_1000000(struct i2c_client *client, uint32_t phloat) @@ -143,12 +144,30 @@ unsigned int ccs_reg_width(u32 reg) return sizeof(uint8_t); } +static u32 ireal32_to_u32_mul_1000000(struct i2c_client *client, u32 val) +{ + if (val >> 10 > U32_MAX / 15625) { + dev_warn(&client->dev, "value %u overflows!\n", val); + return U32_MAX; + } + + return ((val >> 10) * 15625) + + (val & GENMASK(9, 0)) * 15625 / 1024; +} + u32 ccs_reg_conv(struct ccs_sensor *sensor, u32 reg, u32 val) { struct i2c_client *client = v4l2_get_subdevdata(&sensor->src->sd); - if (reg & CCS_FL_FLOAT_IREAL) - val = float_to_u32_mul_1000000(client, val); + if (reg & CCS_FL_FLOAT_IREAL) { + if (CCS_LIM(sensor, CLOCK_CAPA_TYPE_CAPABILITY) & + CCS_CLOCK_CAPA_TYPE_CAPABILITY_IREAL) + val = ireal32_to_u32_mul_1000000(client, val); + else + val = float_to_u32_mul_1000000(client, val); + } else if (reg & CCS_FL_IREAL) { + val = ireal32_to_u32_mul_1000000(client, val); + } return val; } diff --git a/drivers/media/i2c/ccs/ccs.h b/drivers/media/i2c/ccs/ccs.h index cbcd93b519da..f60d1801c469 100644 --- a/drivers/media/i2c/ccs/ccs.h +++ b/drivers/media/i2c/ccs/ccs.h @@ -17,6 +17,7 @@ #include <media/v4l2-subdev.h> #include "ccs-data.h" +#include "ccs-limits.h" #include "ccs-quirk.h" #include "ccs-regs.h" #include "ccs-reg-access.h" @@ -50,6 +51,12 @@ #define CCS_DFL_I2C_ADDR (0x20 >> 1) /* Default I2C Address */ #define CCS_ALT_I2C_ADDR (0x6e >> 1) /* Alternate I2C Address */ +#define CCS_LIM(sensor, limit) \ + ccs_get_limit(sensor, CCS_L_##limit, 0) + +#define CCS_LIM_AT(sensor, limit, offset) \ + ccs_get_limit(sensor, CCS_L_##limit, CCS_L_##limit##_OFFSET(offset)) + /* * Sometimes due to board layout considerations the camera module can be * mounted rotated. The typical rotation used is 180 degrees which can be @@ -277,5 +284,7 @@ struct ccs_sensor { void ccs_replace_limit(struct ccs_sensor *sensor, unsigned int limit, unsigned int offset, u32 val); +u32 ccs_get_limit(struct ccs_sensor *sensor, unsigned int limit, + unsigned int offset); #endif /* __CCS_H__ */ -- 2.27.0