xref: /openbmc/u-boot/board/xes/xpedite537x/ddr.c (revision 0cb77bfa)
1 /*
2  * Copyright 2008 Extreme Engineering Solutions, Inc.
3  * Copyright 2008 Freescale Semiconductor, Inc.
4  *
5  * See file CREDITS for list of people who contributed to this
6  * project.
7  *
8  * This program is free software; you can redistribute it and/or
9  * modify it under the terms of the GNU General Public License as
10  * published by the Free Software Foundation; either version 2 of
11  * the License, or (at your option) any later version.
12  *
13  * This program is distributed in the hope that it will be useful,
14  * but WITHOUT ANY WARRANTY; without even the implied warranty of
15  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.	 See the
16  * GNU General Public License for more details.
17  *
18  * You should have received a copy of the GNU General Public License
19  * along with this program; if not, write to the Free Software
20  * Foundation, Inc., 59 Temple Place, Suite 330, Boston,
21  * MA 02111-1307 USA
22  */
23 
24 #include <common.h>
25 #include <i2c.h>
26 
27 #include <asm/fsl_ddr_sdram.h>
28 #include <asm/fsl_ddr_dimm_params.h>
29 
30 void get_spd(ddr2_spd_eeprom_t *spd, u8 i2c_address)
31 {
32 	i2c_read(i2c_address, SPD_EEPROM_OFFSET, 2, (uchar *)spd,
33 		 sizeof(ddr2_spd_eeprom_t));
34 }
35 
36 /*
37  * There are four board-specific SDRAM timing parameters which must be
38  * calculated based on the particular PCB artwork.  These are:
39  *   1.) CPO (Read Capture Delay)
40  *           - TIMING_CFG_2 register
41  *           Source: Calculation based on board trace lengths and
42  *                   chip-specific internal delays.
43  *   2.) WR_DATA_DELAY (Write Command to Data Strobe Delay)
44  *           - TIMING_CFG_2 register
45  *           Source: Calculation based on board trace lengths.
46  *                   Unless clock and DQ lanes are very different
47  *                   lengths (>2"), this should be set to the nominal value
48  *                   of 1/2 clock delay.
49  *   3.) CLK_ADJUST (Clock and Addr/Cmd alignment control)
50  *           - DDR_SDRAM_CLK_CNTL register
51  *           Source: Signal Integrity Simulations
52  *   4.) 2T Timing on Addr/Ctl
53  *           - TIMING_CFG_2 register
54  *           Source: Signal Integrity Simulations
55  *           Usually only needed with heavy load/very high speed (>DDR2-800)
56  *
57  *     ====== XPedite5370 DDR2-600 read delay calculations ======
58  *
59  *     See Freescale's App Note AN2583 as refrence.  This document also
60  *     contains the chip-specific delays for 8548E, 8572, etc.
61  *
62  *     For MPC8572E
63  *     Minimum chip delay (Ch 0): 1.372ns
64  *     Maximum chip delay (Ch 0): 2.914ns
65  *     Minimum chip delay (Ch 1): 1.220ns
66  *     Maximum chip delay (Ch 1): 2.595ns
67  *
68  *     CLK adjust = 5 (from simulations) = 5/8* 3.33ns = 2080ps
69  *
70  *     Minimum delay calc (Ch 0):
71  *     clock prop - dram skew + min dqs prop delay + clk_adjust + min chip dly
72  *     2.3" * 180 - 400ps     + 1.9" * 180         + 2080ps     + 1372ps
73  *                                                 = 3808ps
74  *                                                 = 3.808ns
75  *
76  *     Maximum delay calc (Ch 0):
77  *     clock prop + dram skew + max dqs prop delay + clk_adjust + max chip dly
78  *     2.3" * 180 + 400ps     + 2.4" * 180         + 2080ps     + 2914ps
79  *                                                 = 6240ps
80  *                                                 = 6.240ns
81  *
82  *     Minimum delay calc (Ch 1):
83  *     clock prop - dram skew + min dqs prop delay + clk_adjust + min chip dly
84  *     1.46" * 180- 400ps     + 0.7" * 180         + 2080ps     + 1220ps
85  *                                                 = 3288ps
86  *                                                 = 3.288ns
87  *
88  *     Maximum delay calc (Ch 1):
89  *     clock prop + dram skew + max dqs prop delay + clk_adjust + min chip dly
90  *     1.46" * 180+ 400ps     + 1.1" * 180         + 2080ps     + 2595ps
91  *                                                 = 5536ps
92  *                                                 = 5.536ns
93  *
94  *     Ch.0: 3.808ns to 6.240ns additional delay needed  (pick 5ns as target)
95  *              This is 1.5 clock cycles, pick CPO = READ_LAT + 3/2 (0x8)
96  *     Ch.1: 3.288ns to 5.536ns additional delay needed  (pick 4.4ns as target)
97  *              This is 1.32 clock cycles, pick CPO = READ_LAT + 5/4 (0x7)
98  *
99  *
100  *     ====== XPedite5370 DDR2-800 read delay calculations ======
101  *
102  *     See Freescale's App Note AN2583 as refrence.  This document also
103  *     contains the chip-specific delays for 8548E, 8572, etc.
104  *
105  *     For MPC8572E
106  *     Minimum chip delay (Ch 0): 1.372ns
107  *     Maximum chip delay (Ch 0): 2.914ns
108  *     Minimum chip delay (Ch 1): 1.220ns
109  *     Maximum chip delay (Ch 1): 2.595ns
110  *
111  *     CLK adjust = 5 (from simulations) = 5/8* 2.5ns = 1563ps
112  *
113  *     Minimum delay calc (Ch 0):
114  *     clock prop - dram skew + min dqs prop delay + clk_adjust + min chip dly
115  *     2.3" * 180 - 350ps     + 1.9" * 180         + 1563ps     + 1372ps
116  *                                                 = 3341ps
117  *                                                 = 3.341ns
118  *
119  *     Maximum delay calc (Ch 0):
120  *     clock prop + dram skew + max dqs prop delay + clk_adjust + max chip dly
121  *     2.3" * 180 + 350ps     + 2.4" * 180         + 1563ps     + 2914ps
122  *                                                 = 5673ps
123  *                                                 = 5.673ns
124  *
125  *     Minimum delay calc (Ch 1):
126  *     clock prop - dram skew + min dqs prop delay + clk_adjust + min chip dly
127  *     1.46" * 180- 350ps     + 0.7" * 180         + 1563ps     + 1220ps
128  *                                                 = 2822ps
129  *                                                 = 2.822ns
130  *
131  *     Maximum delay calc (Ch 1):
132  *     clock prop + dram skew + max dqs prop delay + clk_adjust + min chip dly
133  *     1.46" * 180+ 350ps     + 1.1" * 180         + 1563ps     + 2595ps
134  *                                                 = 4968ps
135  *                                                 = 4.968ns
136  *
137  *     Ch.0: 3.341ns to 5.673ns additional delay needed  (pick 4.5ns as target)
138  *              This is 1.8 clock cycles, pick CPO = READ_LAT + 7/4 (0x9)
139  *     Ch.1: 2.822ns to 4.968ns additional delay needed  (pick 3.9ns as target)
140  *              This is 1.56 clock cycles, pick CPO = READ_LAT + 3/2 (0x8)
141  *
142  * Write latency (WR_DATA_DELAY) is calculated by doing the following:
143  *
144  *      The DDR SDRAM specification requires DQS be received no sooner than
145  *      75% of an SDRAM clock period—and no later than 125% of a clock
146  *      period—from the capturing clock edge of the command/address at the
147  *      SDRAM.
148  *
149  * Based on the above tracelengths, the following are calculated:
150  *      Ch. 0 8572 to DRAM propagation (DQ lanes) : 1.9" * 180 =  0.342ns
151  *      Ch. 0 8572 to DRAM propagation (CLKs) :     2.3" * 180 =  0.414ns
152  *      Ch. 1 8572 to DRAM propagation (DQ lanes) : 0.7" * 180 =  0.126ns
153  *      Ch. 1 8572 to DRAM propagation (CLKs   ) : 1.47" * 180 =  0.264ns
154  *
155  * Difference in arrival time CLK vs. DQS:
156  *      Ch. 0 0.072ns
157  *      Ch. 1 0.138ns
158  *
159  *      Both of these values are much less than 25% of the clock
160  *      period at DDR2-600 or DDR2-800, so no additional delay is needed over
161  *      the 1/2 cycle which normally aligns the first DQS transition
162  *      exactly WL (CAS latency minus one cycle) after the CAS strobe.
163  *      See Figure 9-53 in MPC8572E manual: "1/2 delay" in Freescale's
164  *      terminology corresponds to exactly one clock period delay after
165  *      the CAS strobe. (due to the fact that the "delay" is referenced
166  *      from the *falling* edge of the CLK, just after the rising edge
167  *      which the CAS strobe is latched on.
168  */
169 
170 typedef struct board_memctl_options {
171 	uint16_t datarate_mhz_low;
172 	uint16_t datarate_mhz_high;
173 	uint8_t clk_adjust;
174 	uint8_t cpo_override;
175 	uint8_t write_data_delay;
176 } board_memctl_options_t;
177 
178 static struct board_memctl_options bopts_ctrl[][2] = {
179 	{
180 		/* Controller 0 */
181 		{
182 			/* DDR2 600/667 */
183 			.datarate_mhz_low	= 500,
184 			.datarate_mhz_high	= 750,
185 			.clk_adjust		= 5,
186 			.cpo_override		= 8,
187 			.write_data_delay	= 2,
188 		},
189 		{
190 			/* DDR2 800 */
191 			.datarate_mhz_low	= 750,
192 			.datarate_mhz_high	= 850,
193 			.clk_adjust		= 5,
194 			.cpo_override		= 9,
195 			.write_data_delay	= 2,
196 		},
197 	},
198 	{
199 		/* Controller 1 */
200 		{
201 			/* DDR2 600/667 */
202 			.datarate_mhz_low	= 500,
203 			.datarate_mhz_high	= 750,
204 			.clk_adjust		= 5,
205 			.cpo_override		= 7,
206 			.write_data_delay	= 2,
207 		},
208 		{
209 			/* DDR2 800 */
210 			.datarate_mhz_low	= 750,
211 			.datarate_mhz_high	= 850,
212 			.clk_adjust		= 5,
213 			.cpo_override		= 8,
214 			.write_data_delay	= 2,
215 		},
216 	},
217 };
218 
219 void fsl_ddr_board_options(memctl_options_t *popts,
220 			   dimm_params_t *pdimm,
221 			   unsigned int ctrl_num)
222 {
223 	struct board_memctl_options *bopts = bopts_ctrl[ctrl_num];
224 	sys_info_t sysinfo;
225 	int i;
226 	unsigned int datarate;
227 
228 	get_sys_info(&sysinfo);
229 	datarate = sysinfo.freqDDRBus / 1000 / 1000;
230 
231 	for (i = 0; i < ARRAY_SIZE(bopts_ctrl[ctrl_num]); i++) {
232 		if ((bopts[i].datarate_mhz_low <= datarate) &&
233 		    (bopts[i].datarate_mhz_high >= datarate)) {
234 			debug("controller %d:\n", ctrl_num);
235 			debug(" clk_adjust = %d\n", bopts[i].clk_adjust);
236 			debug(" cpo = %d\n", bopts[i].cpo_override);
237 			debug(" write_data_delay = %d\n",
238 			      bopts[i].write_data_delay);
239 			popts->clk_adjust = bopts[i].clk_adjust;
240 			popts->cpo_override = bopts[i].cpo_override;
241 			popts->write_data_delay = bopts[i].write_data_delay;
242 		}
243 	}
244 
245 	/*
246 	 * Factors to consider for half-strength driver enable:
247 	 *	- number of DIMMs installed
248 	 */
249 	popts->half_strength_driver_enable = 0;
250 }
251