On Wed, Jul 26, 2017 at 8:52 PM, Tom Rini trini@konsulko.com wrote:
On Wed, Jul 26, 2017 at 09:03:37AM -0500, Adam Ford wrote:
This converts the following to Kconfig: CONFIG_SYS_I2C_OMAP24XX CONFIG_SYS_I2C_OMAP34XX
Signed-off-by: Adam Ford aford173@gmail.com
This needs some manual attention. We should just drop CONFIG_SYS_I2C_OMAP24XX as it's meaningless now. Also:
I thought the same thing, but I looked at the driver and the driver has some explicit differences that are unique to the CONFIG_SYS_I2C_OMAP34XX.
As an example: #if defined(CONFIG_OMAP34XX) /* * Have to enable interrupts for OMAP2/3, these IPs don't have * an 'irqstatus_raw' register and we shall have to poll 'stat' */ writew(I2C_IE_XRDY_IE | I2C_IE_RRDY_IE | I2C_IE_ARDY_IE | I2C_IE_NACK_IE | I2C_IE_AL_IE, &i2c_base->ie); #endif
The comment in the code even states there are some minor differences: "Status functions now read irqstatus_raw as per TRM guidelines (except for OMAP243X and OMAP34XX)"
So I think we need both. Looking at the ti_omap4_common.h, it defines CONFIG_SYS_I2C_OMAP24XX, but not OMAP34XX, so it appears to me like we might want a naming convention change.
diff --git a/include/configs/devkit8000.h b/include/configs/devkit8000.h index 5476961..e15eaa6 100644 --- a/include/configs/devkit8000.h +++ b/include/configs/devkit8000.h @@ -65,8 +65,6 @@ #undef CONFIG_OMAP3_SPI
/* I2C */ -#undef CONFIG_SYS_I2C_OMAP24XX -#define CONFIG_SYS_I2C_OMAP34XX
When you see cases like this, you should delete the comment (and extra blank line) by hand as well, in the commit. There's a few instances of this in the patch. Thanks!
I can revisit this and fix it, but I'll wait to hear a reply to my above comment.
adam
-- Tom