blob: 9daaeef731900da386b7627044e515ecc5baf6f7 [file] [log] [blame]
// SPDX-License-Identifier: GPL-2.0+
/*
* EMIF programming
*
* (C) Copyright 2010
* Texas Instruments, <www.ti.com>
*
* Aneesh V <aneesh@ti.com>
*/
#include <common.h>
#include <hang.h>
#include <init.h>
#include <log.h>
#include <net.h>
#include <asm/emif.h>
#include <asm/arch/clock.h>
#include <asm/arch/sys_proto.h>
#include <asm/omap_common.h>
#include <asm/omap_sec_common.h>
#include <asm/utils.h>
#include <linux/compiler.h>
#include <asm/ti-common/ti-edma3.h>
#include <linux/delay.h>
static int emif1_enabled = -1, emif2_enabled = -1;
void set_lpmode_selfrefresh(u32 base)
{
struct emif_reg_struct *emif = (struct emif_reg_struct *)base;
u32 reg;
reg = readl(&emif->emif_pwr_mgmt_ctrl);
reg &= ~EMIF_REG_LP_MODE_MASK;
reg |= LP_MODE_SELF_REFRESH << EMIF_REG_LP_MODE_SHIFT;
reg &= ~EMIF_REG_SR_TIM_MASK;
writel(reg, &emif->emif_pwr_mgmt_ctrl);
/* dummy read for the new SR_TIM to be loaded */
readl(&emif->emif_pwr_mgmt_ctrl);
}
void force_emif_self_refresh(void)
{
set_lpmode_selfrefresh(EMIF1_BASE);
if (!is_dra72x())
set_lpmode_selfrefresh(EMIF2_BASE);
}
inline u32 emif_num(u32 base)
{
if (base == EMIF1_BASE)
return 1;
else if (base == EMIF2_BASE)
return 2;
else
return 0;
}
static inline u32 get_mr(u32 base, u32 cs, u32 mr_addr)
{
u32 mr;
struct emif_reg_struct *emif = (struct emif_reg_struct *)base;
mr_addr |= cs << EMIF_REG_CS_SHIFT;
writel(mr_addr, &emif->emif_lpddr2_mode_reg_cfg);
if (omap_revision() == OMAP4430_ES2_0)
mr = readl(&emif->emif_lpddr2_mode_reg_data_es2);
else
mr = readl(&emif->emif_lpddr2_mode_reg_data);
debug("get_mr: EMIF%d cs %d mr %08x val 0x%x\n", emif_num(base),
cs, mr_addr, mr);
if (((mr & 0x0000ff00) >> 8) == (mr & 0xff) &&
((mr & 0x00ff0000) >> 16) == (mr & 0xff) &&
((mr & 0xff000000) >> 24) == (mr & 0xff))
return mr & 0xff;
else
return mr;
}
static inline void set_mr(u32 base, u32 cs, u32 mr_addr, u32 mr_val)
{
struct emif_reg_struct *emif = (struct emif_reg_struct *)base;
mr_addr |= cs << EMIF_REG_CS_SHIFT;
writel(mr_addr, &emif->emif_lpddr2_mode_reg_cfg);
writel(mr_val, &emif->emif_lpddr2_mode_reg_data);
}
void emif_reset_phy(u32 base)
{
struct emif_reg_struct *emif = (struct emif_reg_struct *)base;
u32 iodft;
iodft = readl(&emif->emif_iodft_tlgc);
iodft |= EMIF_REG_RESET_PHY_MASK;
writel(iodft, &emif->emif_iodft_tlgc);
}
static void do_lpddr2_init(u32 base, u32 cs)
{
u32 mr_addr;
const struct lpddr2_mr_regs *mr_regs;
get_lpddr2_mr_regs(&mr_regs);
/* Wait till device auto initialization is complete */
while (get_mr(base, cs, LPDDR2_MR0) & LPDDR2_MR0_DAI_MASK)
;
set_mr(base, cs, LPDDR2_MR10, mr_regs->mr10);
/*
* tZQINIT = 1 us
* Enough loops assuming a maximum of 2GHz
*/
sdelay(2000);
set_mr(base, cs, LPDDR2_MR1, mr_regs->mr1);
set_mr(base, cs, LPDDR2_MR16, mr_regs->mr16);
/*
* Enable refresh along with writing MR2
* Encoding of RL in MR2 is (RL - 2)
*/
mr_addr = LPDDR2_MR2 | EMIF_REG_REFRESH_EN_MASK;
set_mr(base, cs, mr_addr, mr_regs->mr2);
if (mr_regs->mr3 > 0)
set_mr(base, cs, LPDDR2_MR3, mr_regs->mr3);
}
static void lpddr2_init(u32 base, const struct emif_regs *regs)
{
struct emif_reg_struct *emif = (struct emif_reg_struct *)base;
/* Not NVM */
clrbits_le32(&emif->emif_lpddr2_nvm_config, EMIF_REG_CS1NVMEN_MASK);
/*
* Keep REG_INITREF_DIS = 1 to prevent re-initialization of SDRAM
* when EMIF_SDRAM_CONFIG register is written
*/
setbits_le32(&emif->emif_sdram_ref_ctrl, EMIF_REG_INITREF_DIS_MASK);
/*
* Set the SDRAM_CONFIG and PHY_CTRL for the
* un-locked frequency & default RL
*/
writel(regs->sdram_config_init, &emif->emif_sdram_config);
writel(regs->emif_ddr_phy_ctlr_1_init, &emif->emif_ddr_phy_ctrl_1);
do_ext_phy_settings(base, regs);
do_lpddr2_init(base, CS0);
if (regs->sdram_config & EMIF_REG_EBANK_MASK)
do_lpddr2_init(base, CS1);
writel(regs->sdram_config, &emif->emif_sdram_config);
writel(regs->emif_ddr_phy_ctlr_1, &emif->emif_ddr_phy_ctrl_1);
/* Enable refresh now */
clrbits_le32(&emif->emif_sdram_ref_ctrl, EMIF_REG_INITREF_DIS_MASK);
}
__weak void do_ext_phy_settings(u32 base, const struct emif_regs *regs)
{
}
void emif_update_timings(u32 base, const struct emif_regs *regs)
{
struct emif_reg_struct *emif = (struct emif_reg_struct *)base;
if (!is_dra7xx())
writel(regs->ref_ctrl, &emif->emif_sdram_ref_ctrl_shdw);
else
writel(regs->ref_ctrl_final, &emif->emif_sdram_ref_ctrl_shdw);
writel(regs->sdram_tim1, &emif->emif_sdram_tim_1_shdw);
writel(regs->sdram_tim2, &emif->emif_sdram_tim_2_shdw);
writel(regs->sdram_tim3, &emif->emif_sdram_tim_3_shdw);
if (omap_revision() == OMAP4430_ES1_0) {
/* ES1 bug EMIF should be in force idle during freq_update */
writel(0, &emif->emif_pwr_mgmt_ctrl);
} else {
writel(EMIF_PWR_MGMT_CTRL, &emif->emif_pwr_mgmt_ctrl);
writel(EMIF_PWR_MGMT_CTRL_SHDW, &emif->emif_pwr_mgmt_ctrl_shdw);
}
writel(regs->read_idle_ctrl, &emif->emif_read_idlectrl_shdw);
writel(regs->zq_config, &emif->emif_zq_config);
writel(regs->temp_alert_config, &emif->emif_temp_alert_config);
writel(regs->emif_ddr_phy_ctlr_1, &emif->emif_ddr_phy_ctrl_1_shdw);
if ((omap_revision() >= OMAP5430_ES1_0) || is_dra7xx()) {
writel(EMIF_L3_CONFIG_VAL_SYS_10_MPU_5_LL_0,
&emif->emif_l3_config);
} else if (omap_revision() >= OMAP4460_ES1_0) {
writel(EMIF_L3_CONFIG_VAL_SYS_10_MPU_3_LL_0,
&emif->emif_l3_config);
} else {
writel(EMIF_L3_CONFIG_VAL_SYS_10_LL_0,
&emif->emif_l3_config);
}
}
#ifndef CONFIG_OMAP44XX
static void omap5_ddr3_leveling(u32 base, const struct emif_regs *regs)
{
struct emif_reg_struct *emif = (struct emif_reg_struct *)base;
/* keep sdram in self-refresh */
writel(((LP_MODE_SELF_REFRESH << EMIF_REG_LP_MODE_SHIFT)
& EMIF_REG_LP_MODE_MASK), &emif->emif_pwr_mgmt_ctrl);
__udelay(130);
/*
* Set invert_clkout (if activated)--DDR_PHYCTRL_1
* Invert clock adds an additional half cycle delay on the
* command interface. The additional half cycle, is usually
* meant to enable leveling in the situation that DQS is later
* than CK on the board.It also helps provide some additional
* margin for leveling.
*/
writel(regs->emif_ddr_phy_ctlr_1,
&emif->emif_ddr_phy_ctrl_1);
writel(regs->emif_ddr_phy_ctlr_1,
&emif->emif_ddr_phy_ctrl_1_shdw);
__udelay(130);
writel(((LP_MODE_DISABLE << EMIF_REG_LP_MODE_SHIFT)
& EMIF_REG_LP_MODE_MASK), &emif->emif_pwr_mgmt_ctrl);
/* Launch Full leveling */
writel(DDR3_FULL_LVL, &emif->emif_rd_wr_lvl_ctl);
/* Wait till full leveling is complete */
readl(&emif->emif_rd_wr_lvl_ctl);
__udelay(130);
/* Read data eye leveling no of samples */
config_data_eye_leveling_samples(base);
/*
* Launch 8 incremental WR_LVL- to compensate for
* PHY limitation.
*/
writel(0x2 << EMIF_REG_WRLVLINC_INT_SHIFT,
&emif->emif_rd_wr_lvl_ctl);
__udelay(130);
/* Launch Incremental leveling */
writel(DDR3_INC_LVL, &emif->emif_rd_wr_lvl_ctl);
__udelay(130);
}
static void update_hwleveling_output(u32 base, const struct emif_regs *regs)
{
struct emif_reg_struct *emif = (struct emif_reg_struct *)base;
u32 *emif_ext_phy_ctrl_reg, *emif_phy_status;
u32 reg, i, phy;
emif_phy_status = (u32 *)&emif->emif_ddr_phy_status[6];
phy = readl(&emif->emif_ddr_phy_ctrl_1);
/* Update PHY_REG_RDDQS_RATIO */
emif_ext_phy_ctrl_reg = (u32 *)&emif->emif_ddr_ext_phy_ctrl_7;
if (!(phy & EMIF_DDR_PHY_CTRL_1_RDLVL_MASK_MASK))
for (i = 0; i < PHY_RDDQS_RATIO_REGS; i++) {
reg = readl(emif_phy_status++);
writel(reg, emif_ext_phy_ctrl_reg++);
writel(reg, emif_ext_phy_ctrl_reg++);
}
/* Update PHY_REG_FIFO_WE_SLAVE_RATIO */
emif_ext_phy_ctrl_reg = (u32 *)&emif->emif_ddr_ext_phy_ctrl_2;
emif_phy_status = (u32 *)&emif->emif_ddr_phy_status[11];
if (!(phy & EMIF_DDR_PHY_CTRL_1_RDLVLGATE_MASK_MASK))
for (i = 0; i < PHY_FIFO_WE_SLAVE_RATIO_REGS; i++) {
reg = readl(emif_phy_status++);
writel(reg, emif_ext_phy_ctrl_reg++);
writel(reg, emif_ext_phy_ctrl_reg++);
}
/* Update PHY_REG_WR_DQ/DQS_SLAVE_RATIO */
emif_ext_phy_ctrl_reg = (u32 *)&emif->emif_ddr_ext_phy_ctrl_12;
emif_phy_status = (u32 *)&emif->emif_ddr_phy_status[16];
if (!(phy & EMIF_DDR_PHY_CTRL_1_WRLVL_MASK_MASK))
for (i = 0; i < PHY_REG_WR_DQ_SLAVE_RATIO_REGS; i++) {
reg = readl(emif_phy_status++);
writel(reg, emif_ext_phy_ctrl_reg++);
writel(reg, emif_ext_phy_ctrl_reg++);
}
/* Disable Leveling */
writel(regs->emif_ddr_phy_ctlr_1, &emif->emif_ddr_phy_ctrl_1);
writel(regs->emif_ddr_phy_ctlr_1, &emif->emif_ddr_phy_ctrl_1_shdw);
writel(0x0, &emif->emif_rd_wr_lvl_rmp_ctl);
}
static void dra7_ddr3_leveling(u32 base, const struct emif_regs *regs)
{
struct emif_reg_struct *emif = (struct emif_reg_struct *)base;
/* Clear Error Status */
clrsetbits_le32(&emif->emif_ddr_ext_phy_ctrl_36,
EMIF_REG_PHY_FIFO_WE_IN_MISALINED_CLR,
EMIF_REG_PHY_FIFO_WE_IN_MISALINED_CLR);
clrsetbits_le32(&emif->emif_ddr_ext_phy_ctrl_36_shdw,
EMIF_REG_PHY_FIFO_WE_IN_MISALINED_CLR,
EMIF_REG_PHY_FIFO_WE_IN_MISALINED_CLR);
/* Disable refreshed before leveling */
clrsetbits_le32(&emif->emif_sdram_ref_ctrl, EMIF_REG_INITREF_DIS_MASK,
EMIF_REG_INITREF_DIS_MASK);
/* Start Full leveling */
writel(DDR3_FULL_LVL, &emif->emif_rd_wr_lvl_ctl);
__udelay(300);
/* Check for leveling timeout */
if (readl(&emif->emif_status) & EMIF_REG_LEVELING_TO_MASK) {
printf("Leveling timeout on EMIF%d\n", emif_num(base));
return;
}
/* Enable refreshes after leveling */
clrbits_le32(&emif->emif_sdram_ref_ctrl, EMIF_REG_INITREF_DIS_MASK);
debug("HW leveling success\n");
/*
* Update slave ratios in EXT_PHY_CTRLx registers
* as per HW leveling output
*/
update_hwleveling_output(base, regs);
}
static void dra7_reset_ddr_data(u32 base, u32 size)
{
#if defined(CONFIG_TI_EDMA3) && !defined(CONFIG_DMA)
enable_edma3_clocks();
edma3_fill(EDMA3_BASE, 1, (void *)base, 0, size);
disable_edma3_clocks();
#else
memset((void *)base, 0, size);
#endif
}
static void dra7_enable_ecc(u32 base, const struct emif_regs *regs)
{
struct emif_reg_struct *emif = (struct emif_reg_struct *)base;
u32 rgn, rgn_start, size, ctrl_reg;
/* ECC available only on dra76x EMIF1 */
if ((base != EMIF1_BASE) || !is_dra76x())
return;
if (regs->emif_ecc_ctrl_reg & EMIF_ECC_CTRL_REG_ECC_EN_MASK) {
/* Disable high-order interleaving */
clrbits_le32(MA_PRIORITY, MA_HIMEM_INTERLEAVE_UN_MASK);
#ifdef CONFIG_DRA7XX
/* Clear the status flags and other history */
writel(readl(&emif->emif_1b_ecc_err_cnt),
&emif->emif_1b_ecc_err_cnt);
writel(0xffffffff, &emif->emif_1b_ecc_err_dist_1);
writel(0x2, &emif->emif_1b_ecc_err_addr_log);
writel(0x1, &emif->emif_2b_ecc_err_addr_log);
writel(EMIF_INT_WR_ECC_ERR_SYS_MASK |
EMIF_INT_TWOBIT_ECC_ERR_SYS_MASK |
EMIF_INT_ONEBIT_ECC_ERR_SYS_MASK,
&emif->emif_irqstatus_sys);
#endif
writel(regs->emif_ecc_address_range_1,
&emif->emif_ecc_address_range_1);
writel(regs->emif_ecc_address_range_2,
&emif->emif_ecc_address_range_2);
/* Disable RMW and ECC verification for read accesses */
ctrl_reg = (regs->emif_ecc_ctrl_reg &
~EMIF_ECC_REG_RMW_EN_MASK) |
EMIF_ECC_CTRL_REG_ECC_VERIFY_DIS_MASK;
writel(ctrl_reg, &emif->emif_ecc_ctrl_reg);
/* Set region1 memory with 0 */
rgn_start = (regs->emif_ecc_address_range_1 &
EMIF_ECC_REG_ECC_START_ADDR_MASK) << 16;
rgn = rgn_start + CFG_SYS_SDRAM_BASE;
size = (regs->emif_ecc_address_range_1 &
EMIF_ECC_REG_ECC_END_ADDR_MASK) + 0x10000 - rgn_start;
if (regs->emif_ecc_ctrl_reg &
EMIF_ECC_REG_ECC_ADDR_RGN_1_EN_MASK)
dra7_reset_ddr_data(rgn, size);
/* Set region2 memory with 0 */
rgn_start = (regs->emif_ecc_address_range_2 &
EMIF_ECC_REG_ECC_START_ADDR_MASK) << 16;
rgn = rgn_start + CFG_SYS_SDRAM_BASE;
size = (regs->emif_ecc_address_range_2 &
EMIF_ECC_REG_ECC_END_ADDR_MASK) + 0x10000 - rgn_start;
if (regs->emif_ecc_ctrl_reg &
EMIF_ECC_REG_ECC_ADDR_RGN_2_EN_MASK)
dra7_reset_ddr_data(rgn, size);
/* Default value enables RMW and ECC verification */
writel(regs->emif_ecc_ctrl_reg, &emif->emif_ecc_ctrl_reg);
}
}
static void dra7_ddr3_init(u32 base, const struct emif_regs *regs)
{
struct emif_reg_struct *emif = (struct emif_reg_struct *)base;
if (warm_reset()) {
emif_reset_phy(base);
writel(0x0, &emif->emif_pwr_mgmt_ctrl);
}
do_ext_phy_settings(base, regs);
writel(regs->ref_ctrl | EMIF_REG_INITREF_DIS_MASK,
&emif->emif_sdram_ref_ctrl);
/* Update timing registers */
writel(regs->sdram_tim1, &emif->emif_sdram_tim_1);
writel(regs->sdram_tim2, &emif->emif_sdram_tim_2);
writel(regs->sdram_tim3, &emif->emif_sdram_tim_3);
writel(EMIF_L3_CONFIG_VAL_SYS_10_MPU_5_LL_0, &emif->emif_l3_config);
writel(regs->read_idle_ctrl, &emif->emif_read_idlectrl);
writel(regs->zq_config, &emif->emif_zq_config);
writel(regs->temp_alert_config, &emif->emif_temp_alert_config);
writel(regs->emif_rd_wr_lvl_rmp_ctl, &emif->emif_rd_wr_lvl_rmp_ctl);
writel(regs->emif_rd_wr_lvl_ctl, &emif->emif_rd_wr_lvl_ctl);
writel(regs->emif_ddr_phy_ctlr_1_init, &emif->emif_ddr_phy_ctrl_1);
writel(regs->emif_rd_wr_exec_thresh, &emif->emif_rd_wr_exec_thresh);
writel(regs->ref_ctrl, &emif->emif_sdram_ref_ctrl);
writel(regs->sdram_config2, &emif->emif_lpddr2_nvm_config);
writel(regs->sdram_config_init, &emif->emif_sdram_config);
__udelay(1000);
writel(regs->ref_ctrl_final, &emif->emif_sdram_ref_ctrl);
if (regs->emif_rd_wr_lvl_rmp_ctl & EMIF_REG_RDWRLVL_EN_MASK) {
/*
* Perform Dummy ECC setup just to allow hardware
* leveling of ECC memories
*/
if (is_dra76x() && (base == EMIF1_BASE) &&
(regs->emif_ecc_ctrl_reg & EMIF_ECC_CTRL_REG_ECC_EN_MASK)) {
writel(0, &emif->emif_ecc_address_range_1);
writel(0, &emif->emif_ecc_address_range_2);
writel(EMIF_ECC_CTRL_REG_ECC_EN_MASK |
EMIF_ECC_CTRL_REG_ECC_ADDR_RGN_PROT_MASK,
&emif->emif_ecc_ctrl_reg);
}
dra7_ddr3_leveling(base, regs);
/* Disable ECC */
if (is_dra76x())
writel(0, &emif->emif_ecc_ctrl_reg);
}
/* Enable ECC as necessary */
dra7_enable_ecc(base, regs);
}
static void omap5_ddr3_init(u32 base, const struct emif_regs *regs)
{
struct emif_reg_struct *emif = (struct emif_reg_struct *)base;
writel(regs->ref_ctrl, &emif->emif_sdram_ref_ctrl);
writel(regs->sdram_config_init, &emif->emif_sdram_config);
/*
* Set SDRAM_CONFIG and PHY control registers to locked frequency
* and RL =7. As the default values of the Mode Registers are not
* defined, contents of mode Registers must be fully initialized.
* H/W takes care of this initialization
*/
writel(regs->emif_ddr_phy_ctlr_1_init, &emif->emif_ddr_phy_ctrl_1);
/* Update timing registers */
writel(regs->sdram_tim1, &emif->emif_sdram_tim_1);
writel(regs->sdram_tim2, &emif->emif_sdram_tim_2);
writel(regs->sdram_tim3, &emif->emif_sdram_tim_3);
writel(regs->read_idle_ctrl, &emif->emif_read_idlectrl);
writel(regs->sdram_config2, &emif->emif_lpddr2_nvm_config);
writel(regs->sdram_config_init, &emif->emif_sdram_config);
do_ext_phy_settings(base, regs);
writel(regs->emif_rd_wr_lvl_rmp_ctl, &emif->emif_rd_wr_lvl_rmp_ctl);
omap5_ddr3_leveling(base, regs);
}
static void ddr3_init(u32 base, const struct emif_regs *regs)
{
if (is_omap54xx())
omap5_ddr3_init(base, regs);
else
dra7_ddr3_init(base, regs);
}
#endif
#ifndef CONFIG_SYS_EMIF_PRECALCULATED_TIMING_REGS
#define print_timing_reg(reg) debug(#reg" - 0x%08x\n", (reg))
/*
* Organization and refresh requirements for LPDDR2 devices of different
* types and densities. Derived from JESD209-2 section 2.4
*/
const struct lpddr2_addressing addressing_table[] = {
/* Banks tREFIx10 rowx32,rowx16 colx32,colx16 density */
{BANKS4, T_REFI_15_6, {ROW_12, ROW_12}, {COL_7, COL_8} },/*64M */
{BANKS4, T_REFI_15_6, {ROW_12, ROW_12}, {COL_8, COL_9} },/*128M */
{BANKS4, T_REFI_7_8, {ROW_13, ROW_13}, {COL_8, COL_9} },/*256M */
{BANKS4, T_REFI_7_8, {ROW_13, ROW_13}, {COL_9, COL_10} },/*512M */
{BANKS8, T_REFI_7_8, {ROW_13, ROW_13}, {COL_9, COL_10} },/*1GS4 */
{BANKS8, T_REFI_3_9, {ROW_14, ROW_14}, {COL_9, COL_10} },/*2GS4 */
{BANKS8, T_REFI_3_9, {ROW_14, ROW_14}, {COL_10, COL_11} },/*4G */
{BANKS8, T_REFI_3_9, {ROW_15, ROW_15}, {COL_10, COL_11} },/*8G */
{BANKS4, T_REFI_7_8, {ROW_14, ROW_14}, {COL_9, COL_10} },/*1GS2 */
{BANKS4, T_REFI_3_9, {ROW_15, ROW_15}, {COL_9, COL_10} },/*2GS2 */
};
static const u32 lpddr2_density_2_size_in_mbytes[] = {
8, /* 64Mb */
16, /* 128Mb */
32, /* 256Mb */
64, /* 512Mb */
128, /* 1Gb */
256, /* 2Gb */
512, /* 4Gb */
1024, /* 8Gb */
2048, /* 16Gb */
4096 /* 32Gb */
};
/*
* Calculate the period of DDR clock from frequency value and set the
* denominator and numerator in global variables for easy access later
*/
static void set_ddr_clk_period(u32 freq)
{
/*
* period = 1/freq
* period_in_ns = 10^9/freq
*/
*T_num = 1000000000;
*T_den = freq;
cancel_out(T_num, T_den, 200);
}
/*
* Convert time in nano seconds to number of cycles of DDR clock
*/
static inline u32 ns_2_cycles(u32 ns)
{
return ((ns * (*T_den)) + (*T_num) - 1) / (*T_num);
}
/*
* ns_2_cycles with the difference that the time passed is 2 times the actual
* value(to avoid fractions). The cycles returned is for the original value of
* the timing parameter
*/
static inline u32 ns_x2_2_cycles(u32 ns)
{
return ((ns * (*T_den)) + (*T_num) * 2 - 1) / ((*T_num) * 2);
}
/*
* Find addressing table index based on the device's type(S2 or S4) and
* density
*/
s8 addressing_table_index(u8 type, u8 density, u8 width)
{
u8 index;
if ((density > LPDDR2_DENSITY_8Gb) || (width == LPDDR2_IO_WIDTH_8))
return -1;
/*
* Look at the way ADDR_TABLE_INDEX* values have been defined
* in emif.h compared to LPDDR2_DENSITY_* values
* The table is layed out in the increasing order of density
* (ignoring type). The exceptions 1GS2 and 2GS2 have been placed
* at the end
*/
if ((type == LPDDR2_TYPE_S2) && (density == LPDDR2_DENSITY_1Gb))
index = ADDR_TABLE_INDEX1GS2;
else if ((type == LPDDR2_TYPE_S2) && (density == LPDDR2_DENSITY_2Gb))
index = ADDR_TABLE_INDEX2GS2;
else
index = density;
debug("emif: addressing table index %d\n", index);
return index;
}
/*
* Find the the right timing table from the array of timing
* tables of the device using DDR clock frequency
*/
static const struct lpddr2_ac_timings *get_timings_table(const struct
lpddr2_ac_timings *const *device_timings,
u32 freq)
{
u32 i, temp, freq_nearest;
const struct lpddr2_ac_timings *timings = 0;
emif_assert(freq <= MAX_LPDDR2_FREQ);
emif_assert(device_timings);
/*
* Start with the maximum allowed frequency - that is always safe
*/
freq_nearest = MAX_LPDDR2_FREQ;
/*
* Find the timings table that has the max frequency value:
* i. Above or equal to the DDR frequency - safe
* ii. The lowest that satisfies condition (i) - optimal
*/
for (i = 0; (i < MAX_NUM_SPEEDBINS) && device_timings[i]; i++) {
temp = device_timings[i]->max_freq;
if ((temp >= freq) && (temp <= freq_nearest)) {
freq_nearest = temp;
timings = device_timings[i];
}
}
debug("emif: timings table: %d\n", freq_nearest);
return timings;
}
/*
* Finds the value of emif_sdram_config_reg
* All parameters are programmed based on the device on CS0.
* If there is a device on CS1, it will be same as that on CS0 or
* it will be NVM. We don't support NVM yet.
* If cs1_device pointer is NULL it is assumed that there is no device
* on CS1
*/
static u32 get_sdram_config_reg(const struct lpddr2_device_details *cs0_device,
const struct lpddr2_device_details *cs1_device,
const struct lpddr2_addressing *addressing,
u8 RL)
{
u32 config_reg = 0;
config_reg |= (cs0_device->type + 4) << EMIF_REG_SDRAM_TYPE_SHIFT;
config_reg |= EMIF_INTERLEAVING_POLICY_MAX_INTERLEAVING <<
EMIF_REG_IBANK_POS_SHIFT;
config_reg |= cs0_device->io_width << EMIF_REG_NARROW_MODE_SHIFT;
config_reg |= RL << EMIF_REG_CL_SHIFT;
config_reg |= addressing->row_sz[cs0_device->io_width] <<
EMIF_REG_ROWSIZE_SHIFT;
config_reg |= addressing->num_banks << EMIF_REG_IBANK_SHIFT;
config_reg |= (cs1_device ? EBANK_CS1_EN : EBANK_CS1_DIS) <<
EMIF_REG_EBANK_SHIFT;
config_reg |= addressing->col_sz[cs0_device->io_width] <<
EMIF_REG_PAGESIZE_SHIFT;
return config_reg;
}
static u32 get_sdram_ref_ctrl(u32 freq,
const struct lpddr2_addressing *addressing)
{
u32 ref_ctrl = 0, val = 0, freq_khz;
freq_khz = freq / 1000;
/*
* refresh rate to be set is 'tREFI * freq in MHz
* division by 10000 to account for khz and x10 in t_REFI_us_x10
*/
val = addressing->t_REFI_us_x10 * freq_khz / 10000;
ref_ctrl |= val << EMIF_REG_REFRESH_RATE_SHIFT;
return ref_ctrl;
}
static u32 get_sdram_tim_1_reg(const struct lpddr2_ac_timings *timings,
const struct lpddr2_min_tck *min_tck,
const struct lpddr2_addressing *addressing)
{
u32 tim1 = 0, val = 0;
val = max(min_tck->tWTR, ns_x2_2_cycles(timings->tWTRx2)) - 1;
tim1 |= val << EMIF_REG_T_WTR_SHIFT;
if (addressing->num_banks == BANKS8)
val = (timings->tFAW * (*T_den) + 4 * (*T_num) - 1) /
(4 * (*T_num)) - 1;
else
val = max(min_tck->tRRD, ns_2_cycles(timings->tRRD)) - 1;
tim1 |= val << EMIF_REG_T_RRD_SHIFT;
val = ns_2_cycles(timings->tRASmin + timings->tRPab) - 1;
tim1 |= val << EMIF_REG_T_RC_SHIFT;
val = max(min_tck->tRAS_MIN, ns_2_cycles(timings->tRASmin)) - 1;
tim1 |= val << EMIF_REG_T_RAS_SHIFT;
val = max(min_tck->tWR, ns_2_cycles(timings->tWR)) - 1;
tim1 |= val << EMIF_REG_T_WR_SHIFT;
val = max(min_tck->tRCD, ns_2_cycles(timings->tRCD)) - 1;
tim1 |= val << EMIF_REG_T_RCD_SHIFT;
val = max(min_tck->tRP_AB, ns_2_cycles(timings->tRPab)) - 1;
tim1 |= val << EMIF_REG_T_RP_SHIFT;
return tim1;
}
static u32 get_sdram_tim_2_reg(const struct lpddr2_ac_timings *timings,
const struct lpddr2_min_tck *min_tck)
{
u32 tim2 = 0, val = 0;
val = max(min_tck->tCKE, timings->tCKE) - 1;
tim2 |= val << EMIF_REG_T_CKE_SHIFT;
val = max(min_tck->tRTP, ns_x2_2_cycles(timings->tRTPx2)) - 1;
tim2 |= val << EMIF_REG_T_RTP_SHIFT;
/*
* tXSRD = tRFCab + 10 ns. XSRD and XSNR should have the
* same value
*/
val = ns_2_cycles(timings->tXSR) - 1;
tim2 |= val << EMIF_REG_T_XSRD_SHIFT;
tim2 |= val << EMIF_REG_T_XSNR_SHIFT;
val = max(min_tck->tXP, ns_x2_2_cycles(timings->tXPx2)) - 1;
tim2 |= val << EMIF_REG_T_XP_SHIFT;
return tim2;
}
static u32 get_sdram_tim_3_reg(const struct lpddr2_ac_timings *timings,
const struct lpddr2_min_tck *min_tck,
const struct lpddr2_addressing *addressing)
{
u32 tim3 = 0, val = 0;
val = min(timings->tRASmax * 10 / addressing->t_REFI_us_x10 - 1, 0xF);
tim3 |= val << EMIF_REG_T_RAS_MAX_SHIFT;
val = ns_2_cycles(timings->tRFCab) - 1;
tim3 |= val << EMIF_REG_T_RFC_SHIFT;
val = ns_x2_2_cycles(timings->tDQSCKMAXx2) - 1;
tim3 |= val << EMIF_REG_T_TDQSCKMAX_SHIFT;
val = ns_2_cycles(timings->tZQCS) - 1;
tim3 |= val << EMIF_REG_ZQ_ZQCS_SHIFT;
val = max(min_tck->tCKESR, ns_2_cycles(timings->tCKESR)) - 1;
tim3 |= val << EMIF_REG_T_CKESR_SHIFT;
return tim3;
}
static u32 get_zq_config_reg(const struct lpddr2_device_details *cs1_device,
const struct lpddr2_addressing *addressing,
u8 volt_ramp)
{
u32 zq = 0, val = 0;
if (volt_ramp)
val =
EMIF_ZQCS_INTERVAL_DVFS_IN_US * 10 /
addressing->t_REFI_us_x10;
else
val =
EMIF_ZQCS_INTERVAL_NORMAL_IN_US * 10 /
addressing->t_REFI_us_x10;
zq |= val << EMIF_REG_ZQ_REFINTERVAL_SHIFT;
zq |= (REG_ZQ_ZQCL_MULT - 1) << EMIF_REG_ZQ_ZQCL_MULT_SHIFT;
zq |= (REG_ZQ_ZQINIT_MULT - 1) << EMIF_REG_ZQ_ZQINIT_MULT_SHIFT;
zq |= REG_ZQ_SFEXITEN_ENABLE << EMIF_REG_ZQ_SFEXITEN_SHIFT;
/*
* Assuming that two chipselects have a single calibration resistor
* If there are indeed two calibration resistors, then this flag should
* be enabled to take advantage of dual calibration feature.
* This data should ideally come from board files. But considering
* that none of the boards today have calibration resistors per CS,
* it would be an unnecessary overhead.
*/
zq |= REG_ZQ_DUALCALEN_DISABLE << EMIF_REG_ZQ_DUALCALEN_SHIFT;
zq |= REG_ZQ_CS0EN_ENABLE << EMIF_REG_ZQ_CS0EN_SHIFT;
zq |= (cs1_device ? 1 : 0) << EMIF_REG_ZQ_CS1EN_SHIFT;
return zq;
}
static u32 get_temp_alert_config(const struct lpddr2_device_details *cs1_device,
const struct lpddr2_addressing *addressing,
u8 is_derated)
{
u32 alert = 0, interval;
interval =
TEMP_ALERT_POLL_INTERVAL_MS * 10000 / addressing->t_REFI_us_x10;
if (is_derated)
interval *= 4;
alert |= interval << EMIF_REG_TA_REFINTERVAL_SHIFT;
alert |= TEMP_ALERT_CONFIG_DEVCT_1 << EMIF_REG_TA_DEVCNT_SHIFT;
alert |= TEMP_ALERT_CONFIG_DEVWDT_32 << EMIF_REG_TA_DEVWDT_SHIFT;
alert |= 1 << EMIF_REG_TA_SFEXITEN_SHIFT;
alert |= 1 << EMIF_REG_TA_CS0EN_SHIFT;
alert |= (cs1_device ? 1 : 0) << EMIF_REG_TA_CS1EN_SHIFT;
return alert;
}
static u32 get_read_idle_ctrl_reg(u8 volt_ramp)
{
u32 idle = 0, val = 0;
if (volt_ramp)
val = ns_2_cycles(READ_IDLE_INTERVAL_DVFS) / 64 - 1;
else
/*Maximum value in normal conditions - suggested by hw team */
val = 0x1FF;
idle |= val << EMIF_REG_READ_IDLE_INTERVAL_SHIFT;
idle |= EMIF_REG_READ_IDLE_LEN_VAL << EMIF_REG_READ_IDLE_LEN_SHIFT;
return idle;
}
static u32 get_ddr_phy_ctrl_1(u32 freq, u8 RL)
{
u32 phy = 0, val = 0;
phy |= (RL + 2) << EMIF_REG_READ_LATENCY_SHIFT;
if (freq <= 100000000)
val = EMIF_DLL_SLAVE_DLY_CTRL_100_MHZ_AND_LESS;
else if (freq <= 200000000)
val = EMIF_DLL_SLAVE_DLY_CTRL_200_MHZ;
else
val = EMIF_DLL_SLAVE_DLY_CTRL_400_MHZ;
phy |= val << EMIF_REG_DLL_SLAVE_DLY_CTRL_SHIFT;
/* Other fields are constant magic values. Hardcode them together */
phy |= EMIF_DDR_PHY_CTRL_1_BASE_VAL <<
EMIF_EMIF_DDR_PHY_CTRL_1_BASE_VAL_SHIFT;
return phy;
}
static u32 get_emif_mem_size(u32 base)
{
u32 size_mbytes = 0, temp;
struct emif_device_details dev_details;
struct lpddr2_device_details cs0_dev_details, cs1_dev_details;
u32 emif_nr = emif_num(base);
emif_reset_phy(base);
dev_details.cs0_device_details = emif_get_device_details(emif_nr, CS0,
&cs0_dev_details);
dev_details.cs1_device_details = emif_get_device_details(emif_nr, CS1,
&cs1_dev_details);
emif_reset_phy(base);
if (dev_details.cs0_device_details) {
temp = dev_details.cs0_device_details->density;
size_mbytes += lpddr2_density_2_size_in_mbytes[temp];
}
if (dev_details.cs1_device_details) {
temp = dev_details.cs1_device_details->density;
size_mbytes += lpddr2_density_2_size_in_mbytes[temp];
}
/* convert to bytes */
return size_mbytes << 20;
}
/* Gets the encoding corresponding to a given DMM section size */
u32 get_dmm_section_size_map(u32 section_size)
{
/*
* Section size mapping:
* 0x0: 16-MiB section
* 0x1: 32-MiB section
* 0x2: 64-MiB section
* 0x3: 128-MiB section
* 0x4: 256-MiB section
* 0x5: 512-MiB section
* 0x6: 1-GiB section
* 0x7: 2-GiB section
*/
section_size >>= 24; /* divide by 16 MB */
return log_2_n_round_down(section_size);
}
static void emif_calculate_regs(
const struct emif_device_details *emif_dev_details,
u32 freq, struct emif_regs *regs)
{
u32 temp, sys_freq;
const struct lpddr2_addressing *addressing;
const struct lpddr2_ac_timings *timings;
const struct lpddr2_min_tck *min_tck;
const struct lpddr2_device_details *cs0_dev_details =
emif_dev_details->cs0_device_details;
const struct lpddr2_device_details *cs1_dev_details =
emif_dev_details->cs1_device_details;
const struct lpddr2_device_timings *cs0_dev_timings =
emif_dev_details->cs0_device_timings;
emif_assert(emif_dev_details);
emif_assert(regs);
/*
* You can not have a device on CS1 without one on CS0
* So configuring EMIF without a device on CS0 doesn't
* make sense
*/
emif_assert(cs0_dev_details);
emif_assert(cs0_dev_details->type != LPDDR2_TYPE_NVM);
/*
* If there is a device on CS1 it should be same type as CS0
* (or NVM. But NVM is not supported in this driver yet)
*/
emif_assert((cs1_dev_details == NULL) ||
(cs1_dev_details->type == LPDDR2_TYPE_NVM) ||
(cs0_dev_details->type == cs1_dev_details->type));
emif_assert(freq <= MAX_LPDDR2_FREQ);
set_ddr_clk_period(freq);
/*
* The device on CS0 is used for all timing calculations
* There is only one set of registers for timings per EMIF. So, if the
* second CS(CS1) has a device, it should have the same timings as the
* device on CS0
*/
timings = get_timings_table(cs0_dev_timings->ac_timings, freq);
emif_assert(timings);
min_tck = cs0_dev_timings->min_tck;
temp = addressing_table_index(cs0_dev_details->type,
cs0_dev_details->density,
cs0_dev_details->io_width);
emif_assert((temp >= 0));
addressing = &(addressing_table[temp]);
emif_assert(addressing);
sys_freq = get_sys_clk_freq();
regs->sdram_config_init = get_sdram_config_reg(cs0_dev_details,
cs1_dev_details,
addressing, RL_BOOT);
regs->sdram_config = get_sdram_config_reg(cs0_dev_details,
cs1_dev_details,
addressing, RL_FINAL);
regs->ref_ctrl = get_sdram_ref_ctrl(freq, addressing);
regs->sdram_tim1 = get_sdram_tim_1_reg(timings, min_tck, addressing);
regs->sdram_tim2 = get_sdram_tim_2_reg(timings, min_tck);
regs->sdram_tim3 = get_sdram_tim_3_reg(timings, min_tck, addressing);
regs->read_idle_ctrl = get_read_idle_ctrl_reg(LPDDR2_VOLTAGE_STABLE);
regs->temp_alert_config =
get_temp_alert_config(cs1_dev_details, addressing, 0);
regs->zq_config = get_zq_config_reg(cs1_dev_details, addressing,
LPDDR2_VOLTAGE_STABLE);
regs->emif_ddr_phy_ctlr_1_init =
get_ddr_phy_ctrl_1(sys_freq / 2, RL_BOOT);
regs->emif_ddr_phy_ctlr_1 =
get_ddr_phy_ctrl_1(freq, RL_FINAL);
regs->freq = freq;
print_timing_reg(regs->sdram_config_init);
print_timing_reg(regs->sdram_config);
print_timing_reg(regs->ref_ctrl);
print_timing_reg(regs->sdram_tim1);
print_timing_reg(regs->sdram_tim2);
print_timing_reg(regs->sdram_tim3);
print_timing_reg(regs->read_idle_ctrl);
print_timing_reg(regs->temp_alert_config);
print_timing_reg(regs->zq_config);
print_timing_reg(regs->emif_ddr_phy_ctlr_1);
print_timing_reg(regs->emif_ddr_phy_ctlr_1_init);
}
#endif /* CONFIG_SYS_EMIF_PRECALCULATED_TIMING_REGS */
#ifdef CONFIG_SYS_AUTOMATIC_SDRAM_DETECTION
const char *get_lpddr2_type(u8 type_id)
{
switch (type_id) {
case LPDDR2_TYPE_S4:
return "LPDDR2-S4";
case LPDDR2_TYPE_S2:
return "LPDDR2-S2";
default:
return NULL;
}
}
const char *get_lpddr2_io_width(u8 width_id)
{
switch (width_id) {
case LPDDR2_IO_WIDTH_8:
return "x8";
case LPDDR2_IO_WIDTH_16:
return "x16";
case LPDDR2_IO_WIDTH_32:
return "x32";
default:
return NULL;
}
}
const char *get_lpddr2_manufacturer(u32 manufacturer)
{
switch (manufacturer) {
case LPDDR2_MANUFACTURER_SAMSUNG:
return "Samsung";
case LPDDR2_MANUFACTURER_QIMONDA:
return "Qimonda";
case LPDDR2_MANUFACTURER_ELPIDA:
return "Elpida";
case LPDDR2_MANUFACTURER_ETRON:
return "Etron";
case LPDDR2_MANUFACTURER_NANYA:
return "Nanya";
case LPDDR2_MANUFACTURER_HYNIX:
return "Hynix";
case LPDDR2_MANUFACTURER_MOSEL:
return "Mosel";
case LPDDR2_MANUFACTURER_WINBOND:
return "Winbond";
case LPDDR2_MANUFACTURER_ESMT:
return "ESMT";
case LPDDR2_MANUFACTURER_SPANSION:
return "Spansion";
case LPDDR2_MANUFACTURER_SST:
return "SST";
case LPDDR2_MANUFACTURER_ZMOS:
return "ZMOS";
case LPDDR2_MANUFACTURER_INTEL:
return "Intel";
case LPDDR2_MANUFACTURER_NUMONYX:
return "Numonyx";
case LPDDR2_MANUFACTURER_MICRON:
return "Micron";
default:
return NULL;
}
}
static void display_sdram_details(u32 emif_nr, u32 cs,
struct lpddr2_device_details *device)
{
const char *mfg_str;
const char *type_str;
char density_str[10];
u32 density;
debug("EMIF%d CS%d\t", emif_nr, cs);
if (!device) {
debug("None\n");
return;
}
mfg_str = get_lpddr2_manufacturer(device->manufacturer);
type_str = get_lpddr2_type(device->type);
density = lpddr2_density_2_size_in_mbytes[device->density];
if ((density / 1024 * 1024) == density) {
density /= 1024;
sprintf(density_str, "%d GB", density);
} else
sprintf(density_str, "%d MB", density);
if (mfg_str && type_str)
debug("%s\t\t%s\t%s\n", mfg_str, type_str, density_str);
}
static u8 is_lpddr2_sdram_present(u32 base, u32 cs,
struct lpddr2_device_details *lpddr2_device)
{
u32 mr = 0, temp;
mr = get_mr(base, cs, LPDDR2_MR0);
if (mr > 0xFF) {
/* Mode register value bigger than 8 bit */
return 0;
}
temp = (mr & LPDDR2_MR0_DI_MASK) >> LPDDR2_MR0_DI_SHIFT;
if (temp) {
/* Not SDRAM */
return 0;
}
temp = (mr & LPDDR2_MR0_DNVI_MASK) >> LPDDR2_MR0_DNVI_SHIFT;
if (temp) {
/* DNV supported - But DNV is only supported for NVM */
return 0;
}
mr = get_mr(base, cs, LPDDR2_MR4);
if (mr > 0xFF) {
/* Mode register value bigger than 8 bit */
return 0;
}
mr = get_mr(base, cs, LPDDR2_MR5);
if (mr > 0xFF) {
/* Mode register value bigger than 8 bit */
return 0;
}
if (!get_lpddr2_manufacturer(mr)) {
/* Manufacturer not identified */
return 0;
}
lpddr2_device->manufacturer = mr;
mr = get_mr(base, cs, LPDDR2_MR6);
if (mr >= 0xFF) {
/* Mode register value bigger than 8 bit */
return 0;
}
mr = get_mr(base, cs, LPDDR2_MR7);
if (mr >= 0xFF) {
/* Mode register value bigger than 8 bit */
return 0;
}
mr = get_mr(base, cs, LPDDR2_MR8);
if (mr >= 0xFF) {
/* Mode register value bigger than 8 bit */
return 0;
}
temp = (mr & MR8_TYPE_MASK) >> MR8_TYPE_SHIFT;
if (!get_lpddr2_type(temp)) {
/* Not SDRAM */
return 0;
}
lpddr2_device->type = temp;
temp = (mr & MR8_DENSITY_MASK) >> MR8_DENSITY_SHIFT;
if (temp > LPDDR2_DENSITY_32Gb) {
/* Density not supported */
return 0;
}
lpddr2_device->density = temp;
temp = (mr & MR8_IO_WIDTH_MASK) >> MR8_IO_WIDTH_SHIFT;
if (!get_lpddr2_io_width(temp)) {
/* IO width unsupported value */
return 0;
}
lpddr2_device->io_width = temp;
/*
* If all the above tests pass we should
* have a device on this chip-select
*/
return 1;
}
struct lpddr2_device_details *emif_get_device_details(u32 emif_nr, u8 cs,
struct lpddr2_device_details *lpddr2_dev_details)
{
u32 phy;
u32 base = (emif_nr == 1) ? EMIF1_BASE : EMIF2_BASE;
struct emif_reg_struct *emif = (struct emif_reg_struct *)base;
if (!lpddr2_dev_details)
return NULL;
/* Do the minimum init for mode register accesses */
if (!(running_from_sdram() || warm_reset())) {
phy = get_ddr_phy_ctrl_1(get_sys_clk_freq() / 2, RL_BOOT);
writel(phy, &emif->emif_ddr_phy_ctrl_1);
}
if (!(is_lpddr2_sdram_present(base, cs, lpddr2_dev_details)))
return NULL;
display_sdram_details(emif_num(base), cs, lpddr2_dev_details);
return lpddr2_dev_details;
}
#endif /* CONFIG_SYS_AUTOMATIC_SDRAM_DETECTION */
static void do_sdram_init(u32 base)
{
const struct emif_regs *regs;
u32 in_sdram, emif_nr;
debug(">>do_sdram_init() %x\n", base);
in_sdram = running_from_sdram();
emif_nr = (base == EMIF1_BASE) ? 1 : 2;
#ifdef CONFIG_SYS_EMIF_PRECALCULATED_TIMING_REGS
emif_get_reg_dump(emif_nr, &regs);
if (!regs) {
debug("EMIF: reg dump not provided\n");
return;
}
#else
/*
* The user has not provided the register values. We need to
* calculate it based on the timings and the DDR frequency
*/
struct emif_device_details dev_details;
struct emif_regs calculated_regs;
/*
* Get device details:
* - Discovered if CONFIG_SYS_AUTOMATIC_SDRAM_DETECTION is set
* - Obtained from user otherwise
*/
struct lpddr2_device_details cs0_dev_details, cs1_dev_details;
emif_reset_phy(base);
dev_details.cs0_device_details = emif_get_device_details(emif_nr, CS0,
&cs0_dev_details);
dev_details.cs1_device_details = emif_get_device_details(emif_nr, CS1,
&cs1_dev_details);
emif_reset_phy(base);
/* Return if no devices on this EMIF */
if (!dev_details.cs0_device_details &&
!dev_details.cs1_device_details) {
return;
}
/*
* Get device timings:
* - Default timings specified by JESD209-2 if
* CONFIG_SYS_DEFAULT_LPDDR2_TIMINGS is set
* - Obtained from user otherwise
*/
emif_get_device_timings(emif_nr, &dev_details.cs0_device_timings,
&dev_details.cs1_device_timings);
/* Calculate the register values */
emif_calculate_regs(&dev_details, omap_ddr_clk(), &calculated_regs);
regs = &calculated_regs;
#endif /* CONFIG_SYS_EMIF_PRECALCULATED_TIMING_REGS */
/*
* Initializing the DDR device can not happen from SDRAM.
* Changing the timing registers in EMIF can happen(going from one
* OPP to another)
*/
if (!in_sdram && (!warm_reset() || is_dra7xx())) {
if (emif_sdram_type(regs->sdram_config) ==
EMIF_SDRAM_TYPE_LPDDR2)
lpddr2_init(base, regs);
#ifndef CONFIG_OMAP44XX
else
ddr3_init(base, regs);
#endif
}
#ifdef CONFIG_OMAP54XX
if (warm_reset() && (emif_sdram_type(regs->sdram_config) ==
EMIF_SDRAM_TYPE_DDR3) && !is_dra7xx()) {
set_lpmode_selfrefresh(base);
emif_reset_phy(base);
omap5_ddr3_leveling(base, regs);
}
#endif
/* Write to the shadow registers */
emif_update_timings(base, regs);
debug("<<do_sdram_init() %x\n", base);
}
void emif_post_init_config(u32 base)
{
struct emif_reg_struct *emif = (struct emif_reg_struct *)base;
u32 omap_rev = omap_revision();
/* reset phy on ES2.0 */
if (omap_rev == OMAP4430_ES2_0)
emif_reset_phy(base);
/* Put EMIF back in smart idle on ES1.0 */
if (omap_rev == OMAP4430_ES1_0)
writel(0x80000000, &emif->emif_pwr_mgmt_ctrl);
}
void dmm_init(u32 base)
{
const struct dmm_lisa_map_regs *lisa_map_regs;
u32 i, section, valid;
#ifdef CONFIG_SYS_EMIF_PRECALCULATED_TIMING_REGS
emif_get_dmm_regs(&lisa_map_regs);
#else
u32 emif1_size, emif2_size, mapped_size, section_map = 0;
u32 section_cnt, sys_addr;
struct dmm_lisa_map_regs lis_map_regs_calculated = {0};
mapped_size = 0;
section_cnt = 3;
sys_addr = CFG_SYS_SDRAM_BASE;
emif1_size = get_emif_mem_size(EMIF1_BASE);
emif2_size = get_emif_mem_size(EMIF2_BASE);
debug("emif1_size 0x%x emif2_size 0x%x\n", emif1_size, emif2_size);
if (!emif1_size && !emif2_size)
return;
/* symmetric interleaved section */
if (emif1_size && emif2_size) {
mapped_size = min(emif1_size, emif2_size);
section_map = DMM_LISA_MAP_INTERLEAVED_BASE_VAL;
section_map |= 0 << EMIF_SDRC_ADDR_SHIFT;
/* only MSB */
section_map |= (sys_addr >> 24) <<
EMIF_SYS_ADDR_SHIFT;
section_map |= get_dmm_section_size_map(mapped_size * 2)
<< EMIF_SYS_SIZE_SHIFT;
lis_map_regs_calculated.dmm_lisa_map_3 = section_map;
emif1_size -= mapped_size;
emif2_size -= mapped_size;
sys_addr += (mapped_size * 2);
section_cnt--;
}
/*
* Single EMIF section(we can have a maximum of 1 single EMIF
* section- either EMIF1 or EMIF2 or none, but not both)
*/
if (emif1_size) {
section_map = DMM_LISA_MAP_EMIF1_ONLY_BASE_VAL;
section_map |= get_dmm_section_size_map(emif1_size)
<< EMIF_SYS_SIZE_SHIFT;
/* only MSB */
section_map |= (mapped_size >> 24) <<
EMIF_SDRC_ADDR_SHIFT;
/* only MSB */
section_map |= (sys_addr >> 24) << EMIF_SYS_ADDR_SHIFT;
section_cnt--;
}
if (emif2_size) {
section_map = DMM_LISA_MAP_EMIF2_ONLY_BASE_VAL;
section_map |= get_dmm_section_size_map(emif2_size) <<
EMIF_SYS_SIZE_SHIFT;
/* only MSB */
section_map |= mapped_size >> 24 << EMIF_SDRC_ADDR_SHIFT;
/* only MSB */
section_map |= sys_addr >> 24 << EMIF_SYS_ADDR_SHIFT;
section_cnt--;
}
if (section_cnt == 2) {
/* Only 1 section - either symmetric or single EMIF */
lis_map_regs_calculated.dmm_lisa_map_3 = section_map;
lis_map_regs_calculated.dmm_lisa_map_2 = 0;
lis_map_regs_calculated.dmm_lisa_map_1 = 0;
} else {
/* 2 sections - 1 symmetric, 1 single EMIF */
lis_map_regs_calculated.dmm_lisa_map_2 = section_map;
lis_map_regs_calculated.dmm_lisa_map_1 = 0;
}
/* TRAP for invalid TILER mappings in section 0 */
lis_map_regs_calculated.dmm_lisa_map_0 = DMM_LISA_MAP_0_INVAL_ADDR_TRAP;
if (omap_revision() >= OMAP4460_ES1_0)
lis_map_regs_calculated.is_ma_present = 1;
lisa_map_regs = &lis_map_regs_calculated;
#endif
struct dmm_lisa_map_regs *hw_lisa_map_regs =
(struct dmm_lisa_map_regs *)base;
writel(0, &hw_lisa_map_regs->dmm_lisa_map_3);
writel(0, &hw_lisa_map_regs->dmm_lisa_map_2);
writel(0, &hw_lisa_map_regs->dmm_lisa_map_1);
writel(0, &hw_lisa_map_regs->dmm_lisa_map_0);
writel(lisa_map_regs->dmm_lisa_map_3,
&hw_lisa_map_regs->dmm_lisa_map_3);
writel(lisa_map_regs->dmm_lisa_map_2,
&hw_lisa_map_regs->dmm_lisa_map_2);
writel(lisa_map_regs->dmm_lisa_map_1,
&hw_lisa_map_regs->dmm_lisa_map_1);
writel(lisa_map_regs->dmm_lisa_map_0,
&hw_lisa_map_regs->dmm_lisa_map_0);
if (lisa_map_regs->is_ma_present) {
hw_lisa_map_regs =
(struct dmm_lisa_map_regs *)MA_BASE;
writel(lisa_map_regs->dmm_lisa_map_3,
&hw_lisa_map_regs->dmm_lisa_map_3);
writel(lisa_map_regs->dmm_lisa_map_2,
&hw_lisa_map_regs->dmm_lisa_map_2);
writel(lisa_map_regs->dmm_lisa_map_1,
&hw_lisa_map_regs->dmm_lisa_map_1);
writel(lisa_map_regs->dmm_lisa_map_0,
&hw_lisa_map_regs->dmm_lisa_map_0);
setbits_le32(MA_PRIORITY, MA_HIMEM_INTERLEAVE_UN_MASK);
}
/*
* EMIF should be configured only when
* memory is mapped on it. Using emif1_enabled
* and emif2_enabled variables for this.
*/
emif1_enabled = 0;
emif2_enabled = 0;
for (i = 0; i < 4; i++) {
section = __raw_readl(DMM_BASE + i*4);
valid = (section & EMIF_SDRC_MAP_MASK) >>
(EMIF_SDRC_MAP_SHIFT);
if (valid == 3) {
emif1_enabled = 1;
emif2_enabled = 1;
break;
}
if (valid == 1)
emif1_enabled = 1;
if (valid == 2)
emif2_enabled = 1;
}
}
static void do_bug0039_workaround(u32 base)
{
u32 val, i, clkctrl;
struct emif_reg_struct *emif_base = (struct emif_reg_struct *)base;
const struct read_write_regs *bug_00339_regs;
u32 iterations;
u32 *phy_status_base = &emif_base->emif_ddr_phy_status[0];
u32 *phy_ctrl_base = &emif_base->emif_ddr_ext_phy_ctrl_1;
if (is_dra7xx())
phy_status_base++;
bug_00339_regs = get_bug_regs(&iterations);
/* Put EMIF in to idle */
clkctrl = __raw_readl((*prcm)->cm_memif_clkstctrl);
__raw_writel(0x0, (*prcm)->cm_memif_clkstctrl);
/* Copy the phy status registers in to phy ctrl shadow registers */
for (i = 0; i < iterations; i++) {
val = __raw_readl(phy_status_base +
bug_00339_regs[i].read_reg - 1);
__raw_writel(val, phy_ctrl_base +
((bug_00339_regs[i].write_reg - 1) << 1));
__raw_writel(val, phy_ctrl_base +
(bug_00339_regs[i].write_reg << 1) - 1);
}
/* Disable leveling */
writel(0x0, &emif_base->emif_rd_wr_lvl_rmp_ctl);
__raw_writel(clkctrl, (*prcm)->cm_memif_clkstctrl);
}
/*
* SDRAM initialization:
* SDRAM initialization has two parts:
* 1. Configuring the SDRAM device
* 2. Update the AC timings related parameters in the EMIF module
* (1) should be done only once and should not be done while we are
* running from SDRAM.
* (2) can and should be done more than once if OPP changes.
* Particularly, this may be needed when we boot without SPL and
* and using Configuration Header(CH). ROM code supports only at 50% OPP
* at boot (low power boot). So u-boot has to switch to OPP100 and update
* the frequency. So,
* Doing (1) and (2) makes sense - first time initialization
* Doing (2) and not (1) makes sense - OPP change (when using CH)
* Doing (1) and not (2) doen't make sense
* See do_sdram_init() for the details
*/
void sdram_init(void)
{
u32 in_sdram, size_prog, size_detect;
struct emif_reg_struct *emif = (struct emif_reg_struct *)EMIF1_BASE;
u32 sdram_type = emif_sdram_type(emif->emif_sdram_config);
debug(">>sdram_init()\n");
if (omap_hw_init_context() == OMAP_INIT_CONTEXT_UBOOT_AFTER_SPL)
return;
in_sdram = running_from_sdram();
debug("in_sdram = %d\n", in_sdram);
if (!in_sdram) {
if ((sdram_type == EMIF_SDRAM_TYPE_LPDDR2) && !warm_reset())
bypass_dpll((*prcm)->cm_clkmode_dpll_core);
else if (sdram_type == EMIF_SDRAM_TYPE_DDR3)
writel(CM_DLL_CTRL_NO_OVERRIDE, (*prcm)->cm_dll_ctrl);
}
if (!in_sdram)
dmm_init(DMM_BASE);
if (emif1_enabled)
do_sdram_init(EMIF1_BASE);
if (emif2_enabled)
do_sdram_init(EMIF2_BASE);
if (!(in_sdram || warm_reset())) {
if (emif1_enabled)
emif_post_init_config(EMIF1_BASE);
if (emif2_enabled)
emif_post_init_config(EMIF2_BASE);
}
/* for the shadow registers to take effect */
if (sdram_type == EMIF_SDRAM_TYPE_LPDDR2)
freq_update_core();
/* Do some testing after the init */
if (!in_sdram) {
size_prog = omap_sdram_size();
size_prog = log_2_n_round_down(size_prog);
size_prog = (1 << size_prog);
size_detect = get_ram_size((long *)CFG_SYS_SDRAM_BASE,
size_prog);
/* Compare with the size programmed */
if (size_detect != size_prog) {
printf("SDRAM: identified size not same as expected"
" size identified: %x expected: %x\n",
size_detect,
size_prog);
} else
debug("get_ram_size() successful");
}
#if defined(CONFIG_TI_SECURE_DEVICE)
/*
* On HS devices, do static EMIF firewall configuration
* but only do it if not already running in SDRAM
*/
if (!in_sdram)
if (0 != secure_emif_reserve())
hang();
/* On HS devices, ensure static EMIF firewall APIs are locked */
if (0 != secure_emif_firewall_lock())
hang();
#endif
if (sdram_type == EMIF_SDRAM_TYPE_DDR3 &&
(!in_sdram && !warm_reset()) && (!is_dra7xx())) {
if (emif1_enabled)
do_bug0039_workaround(EMIF1_BASE);
if (emif2_enabled)
do_bug0039_workaround(EMIF2_BASE);
}
debug("<<sdram_init()\n");
}