summaryrefslogtreecommitdiff
path: root/src/cpu/intel/haswell/smmrelocate.c
blob: 1ccc9bb5b3b768aaafb14cc2661d96e9db392141 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
/*
 * This file is part of the coreboot project.
 *
 * Copyright (C) 2013 ChromeOS Authors
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; version 2 of the License.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write to the Free Software
 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301 USA
 */

#include <types.h>
#include <string.h>
#include <device/device.h>
#include <device/pci.h>
#include <cpu/cpu.h>
#include <cpu/x86/cache.h>
#include <cpu/x86/msr.h>
#include <cpu/x86/mtrr.h>
#include <cpu/x86/smm.h>
#include <console/console.h>
#include <northbridge/intel/haswell/haswell.h>
#include <southbridge/intel/lynxpoint/pch.h>
#include "haswell.h"

#define EMRRphysBase_MSR 0x1f4
#define EMRRphysMask_MSR 0x1f5
#define UNCORE_EMRRphysBase_MSR 0x2f4
#define UNCORE_EMRRphysMask_MSR 0x2f5

#define SMRR_SUPPORTED (1<<11)
#define EMRR_SUPPORTED (1<<12)

struct smm_relocation_params {
	u32 smram_base;
	u32 smram_size;
	u32 ied_base;
	u32 ied_size;
	msr_t smrr_base;
	msr_t smrr_mask;
	msr_t emrr_base;
	msr_t emrr_mask;
	msr_t uncore_emrr_base;
	msr_t uncore_emrr_mask;
};

/* This gets filled in and used during relocation. */
static struct smm_relocation_params smm_reloc_params;

static inline void write_smrr(struct smm_relocation_params *relo_params)
{
	printk(BIOS_DEBUG, "Writing SMRR. base = 0x%08x, mask=0x%08x\n",
	       relo_params->smrr_base.lo, relo_params->smrr_mask.lo);
	wrmsr(SMRRphysBase_MSR, relo_params->smrr_base);
	wrmsr(SMRRphysMask_MSR, relo_params->smrr_mask);
}

static inline void write_emrr(struct smm_relocation_params *relo_params)
{
	printk(BIOS_DEBUG, "Writing EMRR. base = 0x%08x, mask=0x%08x\n",
	       relo_params->emrr_base.lo, relo_params->emrr_mask.lo);
	wrmsr(EMRRphysBase_MSR, relo_params->emrr_base);
	wrmsr(EMRRphysMask_MSR, relo_params->emrr_mask);
}

static inline void write_uncore_emrr(struct smm_relocation_params *relo_params)
{
	printk(BIOS_DEBUG,
	       "Writing UNCORE_EMRR. base = 0x%08x, mask=0x%08x\n",
	       relo_params->uncore_emrr_base.lo,
	       relo_params->uncore_emrr_mask.lo);
	wrmsr(UNCORE_EMRRphysBase_MSR, relo_params->uncore_emrr_base);
	wrmsr(UNCORE_EMRRphysMask_MSR, relo_params->uncore_emrr_mask);
}

/* The relocation work is actually performed in SMM context, but the code
 * resides in the ramstage module. This occurs by trampolining from the default
 * SMRAM entry point to here. */
static void cpu_smm_do_relocation(void *arg, int cpu,
                                  const struct smm_runtime *runtime)
{
	em64t101_smm_state_save_area_t *save_state;
	msr_t mtrr_cap;
	struct smm_relocation_params *relo_params = arg;

	if (cpu >= CONFIG_MAX_CPUS) {
		printk(BIOS_CRIT,
		       "Invalid CPU number assigned in SMM stub: %d\n", cpu);
		return;
	}

	printk(BIOS_DEBUG, "In relocation handler: cpu %d\n", cpu);

	/* All threads need to set IEDBASE and SMBASE in the save state area.
	 * Since one thread runs at a time during the relocation the save state
	 * is the same for all cpus. */
	save_state = (void *)(runtime->smbase + SMM_DEFAULT_SIZE -
                              runtime->save_state_size);

	/* The relocated handler runs with all CPUs concurrently. Therefore
	 * stagger the entry points adjusting SMBASE downwards by save state
	 * size * CPU num. */
	save_state->smbase = relo_params->smram_base -
	                     cpu * runtime->save_state_size;
	save_state->iedbase = relo_params->ied_base;

	printk(BIOS_DEBUG, "New SMBASE=0x%08x IEDBASE=0x%08x @ %p\n",
	       save_state->smbase, save_state->iedbase, save_state);

	/* Write EMRR and SMRR MSRs based on indicated support. */
	mtrr_cap = rdmsr(MTRRcap_MSR);
	if (mtrr_cap.lo & SMRR_SUPPORTED)
		write_smrr(relo_params);

	if (mtrr_cap.lo & EMRR_SUPPORTED) {
		write_emrr(relo_params);
		/* UNCORE_EMRR msrs are package level. Therefore, only
		 * configure these MSRs on the BSP. */
		if (cpu == 0)
			write_uncore_emrr(relo_params);
	}

	southbridge_clear_smi_status();
}

static u32 northbridge_get_base_reg(device_t dev, int reg)
{
	u32 value;

	value = pci_read_config32(dev, reg);
	/* Base registers are at 1MiB granularity. */
	value &= ~((1 << 20) - 1);
	return value;
}

static void fill_in_relocation_params(device_t dev,
                                      struct smm_relocation_params *params)
{
	u32 tseg_size;
	u32 tsegmb;
	u32 bgsm;
	u32 emrr_base;
	u32 emrr_size;
	int phys_bits;
	/* All range registers are aligned to 4KiB */
	const u32 rmask = ~((1 << 12) - 1);

	/* Some of the range registers are dependent on the number of physical
	 * address bits supported. */
	phys_bits = cpuid_eax(0x80000008) & 0xff;

	/* The range bounded by the TSEGMB and BGSM registers encompasses the
	 * SMRAM range as well as the IED range. However, the SMRAM available
	 * to the handler is 4MiB since the IEDRAM lives TSEGMB + 4MiB.
	 */
	tsegmb = northbridge_get_base_reg(dev, TSEG);
	bgsm = northbridge_get_base_reg(dev, BGSM);
	tseg_size = bgsm - tsegmb;

	params->smram_base = tsegmb;
	params->smram_size = 4 << 20;
	params->ied_base = tsegmb + params->smram_size;
	params->ied_size = tseg_size - params->smram_size;

	/* SMRR has 32-bits of valid address aligned to 4KiB. */
	params->smrr_base.lo = (params->smram_base & rmask) | MTRR_TYPE_WRBACK;
	params->smrr_base.hi = 0;
	params->smrr_mask.lo = (~(tseg_size - 1) & rmask) | MTRRphysMaskValid;
	params->smrr_mask.hi = 0;

	/* The EMRR and UNCORE_EMRR are at IEDBASE + 2MiB */
	emrr_base = (params->ied_base + (2 << 20)) & rmask;
	emrr_size = params->ied_size - (2 << 20);

	/* EMRR has 46 bits of valid address aligned to 4KiB. It's dependent
	 * on the number of physical address bits supported. */
	params->emrr_base.lo = emrr_base | MTRR_TYPE_WRBACK;
	params->emrr_base.hi = 0;
	params->emrr_mask.lo = (~(emrr_size - 1) & rmask) | MTRRphysMaskValid;
	params->emrr_mask.hi = (1 << (phys_bits - 32)) - 1;

	/* UNCORE_EMRR has 39 bits of valid address aligned to 4KiB. */
	params->uncore_emrr_base.lo = emrr_base;
	params->uncore_emrr_base.hi = 0;
	params->uncore_emrr_mask.lo = (~(emrr_size - 1) & rmask) |
	                              MTRRphysMaskValid;
	params->uncore_emrr_mask.hi = (1 << (39 - 32)) - 1;
}

static int install_relocation_handler(int num_cpus,
                                      struct smm_relocation_params *relo_params)
{
	/* The default SMM entry happens serially at the default location.
	 * Therefore, there is only 1 concurrent save state area. Set the
	 * stack size to the save state size, and call into the
	 * do_relocation handler. */
	int save_state_size = sizeof(em64t101_smm_state_save_area_t);
	struct smm_loader_params smm_params = {
		.per_cpu_stack_size = save_state_size,
		.num_concurrent_stacks = num_cpus,
		.per_cpu_save_state_size = save_state_size,
		.num_concurrent_save_states = 1,
		.handler = &cpu_smm_do_relocation,
		.handler_arg = (void *)relo_params,
	};

	return smm_setup_relocation_handler(&smm_params);
}

static void setup_ied_area(struct smm_relocation_params *params)
{
	char *ied_base;

	struct ied_header ied = {
		.signature = "INTEL RSVD",
		.size = params->ied_size,
		.reserved = {0},
	};

	ied_base = (void *)params->ied_base;

	/* Place IED header at IEDBASE. */
	memcpy(ied_base, &ied, sizeof(ied));

	/* Zero out 32KiB at IEDBASE + 1MiB */
	memset(ied_base + (1 << 20), 0, (32 << 10));

	/* According to the BWG MP init section 2MiB of memory at IEDBASE +
	 * 2MiB should be zeroed as well. However, I suspect what is inteneded
	 * is to clear the memory covered by EMRR. TODO(adurbin): figure out if 	 * this is really required. */
	//memset(ied_base + (2 << 20), 0, (2 << 20));
}

static int install_permanent_handler(int num_cpus,
                                     struct smm_relocation_params *relo_params)
{
	/* There are num_cpus concurrent stacks and num_cpus concurrent save
	 * state areas. Lastly, set the stack size to the save state size. */
	int save_state_size = sizeof(em64t101_smm_state_save_area_t);
	struct smm_loader_params smm_params = {
		.per_cpu_stack_size = save_state_size,
		.num_concurrent_stacks = num_cpus,
		.per_cpu_save_state_size = save_state_size,
		.num_concurrent_save_states = num_cpus,
	};

	printk(BIOS_DEBUG, "Installing SMM handler to 0x%08x\n",
	       relo_params->smram_base);
	return smm_load_module((void *)relo_params->smram_base,
	                       relo_params->smram_size, &smm_params);
}

static int cpu_smm_setup(void)
{
	device_t dev;
	int num_cpus;
	msr_t msr;

	printk(BIOS_DEBUG, "Setting up SMI for CPU\n");

	dev = dev_find_slot(0, PCI_DEVFN(0, 0));

	fill_in_relocation_params(dev, &smm_reloc_params);

	setup_ied_area(&smm_reloc_params);

	msr = rdmsr(CORE_THREAD_COUNT_MSR);
	num_cpus = msr.lo & 0xffff;
	if (num_cpus > CONFIG_MAX_CPUS) {
		printk(BIOS_CRIT,
		       "Error: Hardware CPUs (%d) > MAX_CPUS (%d)\n",
		       num_cpus, CONFIG_MAX_CPUS);
	}

	if (install_relocation_handler(num_cpus, &smm_reloc_params)) {
		printk(BIOS_CRIT, "SMM Relocation handler install failed.\n");
		return -1;
	}

	if (install_permanent_handler(num_cpus, &smm_reloc_params)) {
		printk(BIOS_CRIT, "SMM Permanent handler install failed.\n");
		return -1;
	}

	/* Ensure the SMM handlers hit DRAM before performing first SMI. */
	/* TODO(adurbin): Is this really needed? */
	wbinvd();

	return 0;
}

void smm_init(void)
{
	/* Return early if CPU SMM setup failed. */
	if (cpu_smm_setup())
		return;

	southbridge_smm_init();

	/* Initiate first SMI to kick off SMM-context relocation. Note: this
	 * SMI being triggered here queues up an SMI in the APs which are in
	 * wait-for-SIPI state. Once an AP gets an SIPI it will service the SMI
	 * at the SMM_DEFAULT_BASE before jumping to startup vector. */
	southbridge_trigger_smi();

	printk(BIOS_DEBUG, "Relocation complete.\n");

	/* Lock down the SMRAM space. */
	smm_lock();
}

void smm_lock(void)
{
	/* LOCK the SMM memory window and enable normal SMM.
	 * After running this function, only a full reset can
	 * make the SMM registers writable again.
	 */
	printk(BIOS_DEBUG, "Locking SMM.\n");
	pci_write_config8(dev_find_slot(0, PCI_DEVFN(0, 0)), SMRAM,
			D_LCK | G_SMRAME | C_BASE_SEG);
}