OMAP2+: voltage: reorganize, split code from data

This is a first pass at reorganizing mach-omap2/voltage.c:

- Separate almost all of the data from the code of mach-omap2/voltage.c.
  The code remains in mach-omap2/voltage.c.  The data goes into one
  of several places, depending on what type of data it is:

  - Silicon process/validation data: mach-omap2/opp*_data.c
  - VC (Voltage Controller) data: mach-omap2/vc*_data.c
  - VP (Voltage Processor) data: mach-omap2/vp*_data.c
  - Voltage domain data: mach-omap2/voltagedomains*_data.c

  The ultimate goal is for all this data to be autogenerated, the same
  way we autogenerate the rest of our data.

- Separate VC and VP common data from VDD-specific VC and VP data.

- Separate common voltage.c code from SoC-specific code; reuse common code.

- Reorganize structures to avoid unnecessary memory loss due to unpacked
  fields.

There is much left to be done.  VC code and VP code should be separated out
into vc*.c and vp*.c files.  Many fields in the existing structures are
superfluous, and should be removed.  Some code in voltage.c seems to be
duplicated; that code should be moved into functions of its own.  Proper
voltage domain code should be created, as was done with the powerdomain
and clockdomains, and powerdomains should reference voltagedomains.

Thanks to Shweta Gulati <shweta.gulati@ti.com> for comments.  Thanks
to Rajendra Nayak <rnayak@ti.com> for finding and fixing some bugs
that prevented OMAP4 from booting:

   https://patchwork.kernel.org/patch/587311/

His patch has been folded into this one to avoid breaking OMAP4
between patches.  Thanks also to Kevin Hilman <khilman@ti.com> for
finding and fixing a compile problem when !CONFIG_PM:

   http://www.spinics.net/lists/arm-kernel/msg118067.html

His patch has also been folded into this one to avoid breaking
!CONFIG_PM builds.

Signed-off-by: Paul Walmsley <paul@pwsan.com>
Cc: Shweta Gulati <shweta.gulati@ti.com>
Cc: Rajendra Nayak <rnayak@ti.com>
Cc: Kevin Hilman <khilman@ti.com>

1 files changed, 143 insertions, 0 deletions

diff --git a/arch/arm/mach-omap2/vp.h b/arch/arm/mach-omap2/vp.h
new file mode 100644
index 000000000000..7ce134f7de79
--- /dev/null
+++ b/arch/arm/mach-omap2/vp.h
@@ -0,0 +1,143 @@
+/*
+ * OMAP3/4 Voltage Processor (VP) structure and macro definitions
+ *
+ * Copyright (C) 2007, 2010 Texas Instruments, Inc.
+ * Rajendra Nayak <rnayak@ti.com>
+ * Lesly A M <x0080970@ti.com>
+ * Thara Gopinath <thara@ti.com>
+ *
+ * Copyright (C) 2008, 2011 Nokia Corporation
+ * Kalle Jokiniemi
+ * Paul Walmsley
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License version
+ * 2 as published by the Free Software Foundation.
+ */
+#ifndef __ARCH_ARM_MACH_OMAP2_VP_H
+#define __ARCH_ARM_MACH_OMAP2_VP_H
+
+#include <linux/kernel.h>
+
+/* XXX document */
+#define VP_IDLE_TIMEOUT         200
+#define VP_TRANXDONE_TIMEOUT    300
+
+
+/**
+ * struct omap_vp_common_data - register data common to all VDDs
+ * @vpconfig_errorgain_mask: ERRORGAIN bitmask in the PRM_VP*_CONFIG reg
+ * @vpconfig_initvoltage_mask: INITVOLTAGE bitmask in the PRM_VP*_CONFIG reg
+ * @vpconfig_timeouten_mask: TIMEOUT bitmask in the PRM_VP*_CONFIG reg
+ * @vpconfig_initvdd: INITVDD bitmask in the PRM_VP*_CONFIG reg
+ * @vpconfig_forceupdate: FORCEUPDATE bitmask in the PRM_VP*_CONFIG reg
+ * @vpconfig_vpenable: VPENABLE bitmask in the PRM_VP*_CONFIG reg
+ * @vpconfig_erroroffset_shift: ERROROFFSET field shift in PRM_VP*_CONFIG reg
+ * @vpconfig_errorgain_shift: ERRORGAIN field shift in PRM_VP*_CONFIG reg
+ * @vpconfig_initvoltage_shift: INITVOLTAGE field shift in PRM_VP*_CONFIG reg
+ * @vpconfig_stepmin_shift: VSTEPMIN field shift in the PRM_VP*_VSTEPMIN reg
+ * @vpconfig_smpswaittimemin_shift: SMPSWAITTIMEMIN field shift in PRM_VP*_VSTEPMIN reg
+ * @vpconfig_stepmax_shift: VSTEPMAX field shift in the PRM_VP*_VSTEPMAX reg
+ * @vpconfig_smpswaittimemax_shift: SMPSWAITTIMEMAX field shift in PRM_VP*_VSTEPMAX reg
+ * @vpconfig_vlimitto_vddmin_shift: VDDMIN field shift in PRM_VP*_VLIMITTO reg
+ * @vpconfig_vlimitto_vddmax_shift: VDDMAX field shift in PRM_VP*_VLIMITTO reg
+ * @vpconfig_vlimitto_timeout_shift: TIMEOUT field shift in PRM_VP*_VLIMITTO reg
+ *
+ * XXX It it not necessary to have both a mask and a shift for the same
+ *     bitfield - remove one
+ * XXX Many of these fields are wrongly named -- e.g., vpconfig_smps* -- fix!
+ */
+struct omap_vp_common_data {
+        u32 vpconfig_errorgain_mask;
+        u32 vpconfig_initvoltage_mask;
+        u32 vpconfig_timeouten;
+        u32 vpconfig_initvdd;
+        u32 vpconfig_forceupdate;
+        u32 vpconfig_vpenable;
+        u8 vpconfig_erroroffset_shift;
+        u8 vpconfig_errorgain_shift;
+        u8 vpconfig_initvoltage_shift;
+        u8 vstepmin_stepmin_shift;
+        u8 vstepmin_smpswaittimemin_shift;
+        u8 vstepmax_stepmax_shift;
+        u8 vstepmax_smpswaittimemax_shift;
+        u8 vlimitto_vddmin_shift;
+        u8 vlimitto_vddmax_shift;
+        u8 vlimitto_timeout_shift;
+};
+
+/**
+ * struct omap_vp_prm_irqst_data - PRM_IRQSTATUS_MPU.VP_TRANXDONE_ST data
+ * @prm_irqst_reg: reg offset for PRM_IRQSTATUS_MPU from top of PRM
+ * @tranxdone_status: VP_TRANXDONE_ST bitmask in PRM_IRQSTATUS_MPU reg
+ *
+ * XXX prm_irqst_reg does not belong here
+ * XXX Note that on OMAP3, VP_TRANXDONE interrupt may not work due to a
+ *     hardware bug
+ * XXX This structure is probably not needed
+ */
+struct omap_vp_prm_irqst_data {
+        u8 prm_irqst_reg;
+        u32 tranxdone_status;
+};
+
+/**
+ * struct omap_vp_instance_data - VP register offsets (per-VDD)
+ * @vp_common: pointer to struct omap_vp_common_data * for this SoC
+ * @prm_irqst_data: pointer to struct omap_vp_prm_irqst_data for this VDD
+ * @vpconfig: PRM_VP*_CONFIG reg offset from PRM start
+ * @vstepmin: PRM_VP*_VSTEPMIN reg offset from PRM start
+ * @vlimitto: PRM_VP*_VLIMITTO reg offset from PRM start
+ * @vstatus: PRM_VP*_VSTATUS reg offset from PRM start
+ * @voltage: PRM_VP*_VOLTAGE reg offset from PRM start
+ *
+ * XXX vp_common is probably not needed since it is per-SoC
+ */
+struct omap_vp_instance_data {
+        const struct omap_vp_common_data *vp_common;
+        const struct omap_vp_prm_irqst_data *prm_irqst_data;
+        u8 vpconfig;
+        u8 vstepmin;
+        u8 vstepmax;
+        u8 vlimitto;
+        u8 vstatus;
+        u8 voltage;
+};
+
+/**
+ * struct omap_vp_runtime_data - VP data populated at runtime by code
+ * @vpconfig_erroroffset: value of ERROROFFSET bitfield in PRM_VP*_CONFIG
+ * @vpconfig_errorgain: value of ERRORGAIN bitfield in PRM_VP*_CONFIG
+ * @vstepmin_smpswaittimemin: value of SMPSWAITTIMEMIN bitfield in PRM_VP*_VSTEPMIN
+ * @vstepmax_smpswaittimemax: value of SMPSWAITTIMEMAX bitfield in PRM_VP*_VSTEPMAX
+ * @vlimitto_timeout: value of TIMEOUT bitfield in PRM_VP*_VLIMITTO
+ * @vstepmin_stepmin: value of VSTEPMIN bitfield in PRM_VP*_VSTEPMIN
+ * @vstepmax_stepmax: value of VSTEPMAX bitfield in PRM_VP*_VSTEPMAX
+ * @vlimitto_vddmin: value of VDDMIN bitfield in PRM_VP*_VLIMITTO
+ * @vlimitto_vddmax: value of VDDMAX bitfield in PRM_VP*_VLIMITTO
+ *
+ * XXX Is this structure really needed?  Why not just program the
+ * device directly?  They are in PRM space, therefore in the WKUP
+ * powerdomain, so register contents should not be lost in off-mode.
+ * XXX Some of these fields are incorrectly named, e.g., vstep*
+ */
+struct omap_vp_runtime_data {
+        u32 vpconfig_erroroffset;
+        u16 vpconfig_errorgain;
+        u16 vstepmin_smpswaittimemin;
+        u16 vstepmax_smpswaittimemax;
+        u16 vlimitto_timeout;
+        u8 vstepmin_stepmin;
+        u8 vstepmax_stepmax;
+        u8 vlimitto_vddmin;
+        u8 vlimitto_vddmax;
+};
+
+extern struct omap_vp_instance_data omap3_vp1_data;
+extern struct omap_vp_instance_data omap3_vp2_data;
+
+extern struct omap_vp_instance_data omap4_vp_mpu_data;
+extern struct omap_vp_instance_data omap4_vp_iva_data;
+extern struct omap_vp_instance_data omap4_vp_core_data;
+
+#endif