ARM: OMAP2+: Fix AM335x RTC External Abort

Hi,
With next-20160711 there is an external abort during boot on am335x platforms from the RTC which is caused by commit 183e2077d007 ("ARM:
AM33xx: fix module_wait_ready without clkctrl register"), log is here [1]. The RTC hwmod is unique in that it has a clkctrl_offs of 0 which is invalid normally, all other hwmods have non-zero offsets. Because of this, we need to flag the hwmod so 0 actually gets used for the offset instead of the just returning 0 and doing nothing.

This series adds a HWMOD_OMAP4_ZERO_CLKCTRL_OFFSET flag and moves the check for validity of the clkctrl_offs value to the omap_hwmod code so that we are able to check both the flag and clkctrl_offs values. This code is shared by all OMAP4+ platforms but shoudl have no affect for anything but RTC on am335x but still boot tested on the following platforms:

AM335x GP EVM: Pass http://pastebin.ubuntu.com/19186160/
Beaglebone Black: Pass http://pastebin.ubuntu.com/19186862/
AM437x GP EVM: Pass http://pastebin.ubuntu.com/19186555/
OMAP4 Panda ES: Pass http://pastebin.ubuntu.com/19186711/
OMAP5 uEVM: Pass http://pastebin.ubuntu.com/19186925/

Regards,
Dave

[1] https://storage.kernelci.org/next/next-20160711/arm-omap2plus_defconfig/lab-baylibre-seattle/boot-am335x-boneblack.html

Dave Gerlach (4):
ARM: OMAP4+: hwmod: Add hwmod flag for HWMOD_OMAP4_ZERO_CLKCTRL_OFFSET
ARM: OMAP2+: AM33XX: Add HWMOD_OMAP4_ZERO_CLKCTRL_OFFSET flag to rtc
hwmod
ARM: OMAP4+: Have _omap4_wait_target_* check for valid clkctrl_offs
ARM: OMAP4+: CM: Remove redundant checks for clkctrl_offs of zero

 arch/arm/mach-omap2/cm33xx.c                            | 6 ------
 arch/arm/mach-omap2/cminst44xx.c                        | 6 ------
 arch/arm/mach-omap2/omap_hwmod.c                        | 8 ++++++++
 arch/arm/mach-omap2/omap_hwmod.h                        | 4 ++++
 arch/arm/mach-omap2/omap_hwmod_33xx_43xx_ipblock_data.c | 2 ++
 5 files changed, 14 insertions(+), 12 deletions(-)

2.9.0

Nearly all modules on OMAP4 and newer platforms have clkctrl offsets that are non-zero except for the RTC on am335x. Because we rely on a clkctrl_offset of zero to indicate no clkctrl_offset being present, lets add a HWMOD_OMAP4_ZERO_CLKCTRL_OFFSET flag to use to indicate hwmods that have a valid clkctrl_offset of 0.

---
 arch/arm/mach-omap2/omap_hwmod.h | 4 ++++
 1 file changed, 4 insertions(+)

index 4041bad79a9a..78904017f18c 100644
@@ -443,8 +443,12 @@ struct omap_hwmod_omap2_prcm {
  * HWMOD_OMAP4_NO_CONTEXT_LOSS_BIT: Some IP blocks don't have a PRCM
  *     module-level context loss register associated with them; this
  *     flag bit should be set in those cases
+ * HWMOD_OMAP4_ZERO_CLKCTRL_OFFSET: Some IP blocks have a valid CLKCTRL
+ *	offset of zero; this flag bit should be set in those cases to
+ *	distinguish from hwmods that have no clkctrl offset.
  */
 #define HWMOD_OMAP4_NO_CONTEXT_LOSS_BIT		(1 << 0)
+#define HWMOD_OMAP4_ZERO_CLKCTRL_OFFSET		(1 << 1)
 
 /**
  * struct omap_hwmod_omap4_prcm - OMAP4-specific PRCM data

2.9.0

The RTC hwmod on AM335x family of SoCs is unique in that the clkctrl_offs in the PRCM is 0. We rely on a clkctrl_offs of zero as indicating no clkctrl is present so we must flag this hwmod with HWMOD_OMAP4_ZERO_CLKCTRL_OFFSET to indicate the 0 clkctrl_offs is in fact valid in this case.

---
 arch/arm/mach-omap2/omap_hwmod_33xx_43xx_ipblock_data.c | 2 ++
 1 file changed, 2 insertions(+)

index 55c5878577f4..e2d84aa7f595 100644
@@ -29,6 +29,7 @@
 #define CLKCTRL(oh, clkctrl) ((oh).prcm.omap4.clkctrl_offs = (clkctrl))
 #define RSTCTRL(oh, rstctrl) ((oh).prcm.omap4.rstctrl_offs = (rstctrl))
 #define RSTST(oh, rstst) ((oh).prcm.omap4.rstst_offs = (rstst))
+#define PRCM_FLAGS(oh, flag) ((oh).prcm.omap4.flags = (flag))
 
 /*
  * 'l3' class

@@ -1296,6 +1297,7 @@ static void omap_hwmod_am33xx_clkctrl(void)
 	CLKCTRL(am33xx_i2c1_hwmod, AM33XX_CM_WKUP_I2C0_CLKCTRL_OFFSET);
 	CLKCTRL(am33xx_wd_timer1_hwmod, AM33XX_CM_WKUP_WDT1_CLKCTRL_OFFSET);
 	CLKCTRL(am33xx_rtc_hwmod, AM33XX_CM_RTC_RTC_CLKCTRL_OFFSET);
+	PRCM_FLAGS(am33xx_rtc_hwmod, HWMOD_OMAP4_ZERO_CLKCTRL_OFFSET);
 	CLKCTRL(am33xx_mmc2_hwmod, AM33XX_CM_PER_MMC2_CLKCTRL_OFFSET);
 	CLKCTRL(am33xx_gpmc_hwmod, AM33XX_CM_PER_GPMC_CLKCTRL_OFFSET);
 	CLKCTRL(am33xx_l4_ls_hwmod, AM33XX_CM_PER_L4LS_CLKCTRL_OFFSET);

2.9.0

Previously the low-level CM call internal to the _omap4_wait_target_ready/disable calls was responsible for checking for a valid clkctrl_offs. Now we must also consider the value of the prcm.omap4.flags because if HWMOD_OMAP4_ZERO_CLKCTRL_OFFSET is set in the flags then clkctrl_offs of 0 is valid.

Let's move this check into the _omap4_wait_target_ready/disable where we have access to both the clkctrl_offs and the flags values and simply return 0 without calling the low level CM call at all, which would have returned 0 anyway if the clktrl_offs was zero.

---
 arch/arm/mach-omap2/omap_hwmod.c | 8 ++++++++
 1 file changed, 8 insertions(+)

index 5b709383381c..1052b29697b8 100644
@@ -1053,6 +1053,10 @@ static int _omap4_wait_target_disable(struct omap_hwmod *oh)
 	if (oh->flags & HWMOD_NO_IDLEST)
 		return 0;
 
+	if (!oh->prcm.omap4.clkctrl_offs &&
+	    !(oh->prcm.omap4.flags & HWMOD_OMAP4_ZERO_CLKCTRL_OFFSET))
+		return 0;
+
 	return omap_cm_wait_module_idle(oh->clkdm->prcm_partition,
 					oh->clkdm->cm_inst,
 					oh->prcm.omap4.clkctrl_offs, 0);

@@ -2971,6 +2975,10 @@ static int _omap4_wait_target_ready(struct omap_hwmod *oh)
 	if (!_find_mpu_rt_port(oh))
 		return 0;
 
+	if (!oh->prcm.omap4.clkctrl_offs &&
+	    !(oh->prcm.omap4.flags & HWMOD_OMAP4_ZERO_CLKCTRL_OFFSET))
+		return 0;
+
 	/* XXX check module SIDLEMODE, hardreset status */
 
 	return omap_cm_wait_module_ready(oh->clkdm->prcm_partition,

2.9.0

This series looks OK to me to fix the RTC regression. Paul and Tero, do you see any issues with this or care to ack?

I'm planning to s/Let's move this check/Let's add this check/ while committing to avoid confusion.

Regards,

Tony

Now that we have moved the check for valid clkctrl_offs to the caller of am33xx_cm_wait_module_ready/idle and omap4_cminst_wait_module_ready/idle let's remove the now redundant check for clkctrl_offs from these functions.

---
 arch/arm/mach-omap2/cm33xx.c     | 6 ------
 arch/arm/mach-omap2/cminst44xx.c | 6 ------
 2 files changed, 12 deletions(-)

index c073fb57dd13..6f2d0aec0513 100644
@@ -220,9 +220,6 @@ static int am33xx_cm_wait_module_ready(u8 part, s16 inst, u16 clkctrl_offs,
 {
 	int i = 0;
 
-	if (!clkctrl_offs)
-		return 0;
-
 	omap_test_timeout(_is_module_ready(inst, clkctrl_offs),
 			  MAX_MODULE_READY_TIME, i);
 

@@ -246,9 +243,6 @@ static int am33xx_cm_wait_module_idle(u8 part, s16 inst, u16 clkctrl_offs,
 {
 	int i = 0;
 
-	if (!clkctrl_offs)
-		return 0;
-
 	omap_test_timeout((_clkctrl_idlest(inst, clkctrl_offs) ==
 				CLKCTRL_IDLEST_DISABLED),
 				MAX_MODULE_READY_TIME, i);

index 2c0e07ed6b99..2ab27ade136a 100644
@@ -278,9 +278,6 @@ static int omap4_cminst_wait_module_ready(u8 part, s16 inst, u16 clkctrl_offs,
 {
 	int i = 0;
 
-	if (!clkctrl_offs)
-		return 0;
-
 	omap_test_timeout(_is_module_ready(part, inst, clkctrl_offs),
 			  MAX_MODULE_READY_TIME, i);
 

@@ -304,9 +301,6 @@ static int omap4_cminst_wait_module_idle(u8 part, s16 inst, u16 clkctrl_offs,
 {
 	int i = 0;
 
-	if (!clkctrl_offs)
-		return 0;
-
 	omap_test_timeout((_clkctrl_idlest(part, inst, clkctrl_offs) ==
 			   CLKCTRL_IDLEST_DISABLED),
 			  MAX_MODULE_DISABLE_TIME, i);

2.9.0

Series looks good to me, so:

-Tero

Applying these into omap-for-v4,8/soc.

Regards,

Tony