xref: /kvm-unit-tests/lib/x86/setup.c (revision b397e5a52c02e6a15bd38e0199fe7b8c55a4634e)
1 /*
2  * Initialize machine setup information
3  *
4  * Copyright (C) 2017, Red Hat Inc, Andrew Jones <drjones@redhat.com>
5  * Copyright (C) 2021, Google Inc, Zixuan Wang <zixuanwang@google.com>
6  *
7  * This work is licensed under the terms of the GNU LGPL, version 2.
8  */
9 #include "libcflat.h"
10 #include "fwcfg.h"
11 #include "alloc_phys.h"
12 #include "argv.h"
13 #include "desc.h"
14 #include "apic.h"
15 #include "apic-defs.h"
16 #include "asm/setup.h"
17 
18 extern char edata;
19 
20 struct mbi_bootinfo {
21 	u32 flags;
22 	u32 mem_lower;
23 	u32 mem_upper;
24 	u32 boot_device;
25 	u32 cmdline;
26 	u32 mods_count;
27 	u32 mods_addr;
28 	u32 reserved[4];   /* 28-43 */
29 	u32 mmap_length;
30 	u32 mmap_addr;
31 	u32 reserved0[3];  /* 52-63 */
32 	u32 bootloader;
33 	u32 reserved1[5];  /* 68-87 */
34 	u32 size;
35 };
36 
37 struct mbi_module {
38 	u32 start, end;
39 	u32 cmdline;
40 	u32 unused;
41 };
42 
43 struct mbi_mem {
44 	u32 size;
45 	u64 base_addr;
46 	u64 length;
47 	u32 type;
48 } __attribute__((packed));
49 
50 #define ENV_SIZE 16384
51 
52 void setup_env(char *env, int size);
53 void setup_multiboot(struct mbi_bootinfo *bootinfo);
54 void setup_libcflat(void);
55 
56 char *initrd;
57 u32 initrd_size;
58 
59 static char env[ENV_SIZE];
60 static struct mbi_bootinfo *bootinfo;
61 
62 #define HUGEPAGE_SIZE (1 << 21)
63 
64 #ifdef __x86_64__
65 void find_highmem(void)
66 {
67 	/* Memory above 4 GB is only supported on 64-bit systems.  */
68 	if (!(bootinfo->flags & 64))
69 	    	return;
70 
71 	u64 upper_end = bootinfo->mem_upper * 1024ull;
72 	u64 best_start = (uintptr_t) &edata;
73 	u64 best_end = upper_end;
74 	u64 max_end = fwcfg_get_u64(FW_CFG_MAX_RAM);
75 	if (max_end == 0)
76 		max_end = -1ull;
77 	bool found = false;
78 
79 	uintptr_t mmap = bootinfo->mmap_addr;
80 	while (mmap < bootinfo->mmap_addr + bootinfo->mmap_length) {
81 		struct mbi_mem *mem = (void *)mmap;
82 		mmap += mem->size + 4;
83 		if (mem->type != 1)
84 			continue;
85 		if (mem->base_addr <= (uintptr_t) &edata ||
86 		    (mem->base_addr <= upper_end && mem->base_addr + mem->length <= upper_end))
87 			continue;
88 		if (mem->length < best_end - best_start)
89 			continue;
90 		if (mem->base_addr >= max_end)
91 			continue;
92 		best_start = mem->base_addr;
93 		best_end = mem->base_addr + mem->length;
94 		if (best_end > max_end)
95 			best_end = max_end;
96 		found = true;
97 	}
98 
99 	if (found) {
100 		best_start = (best_start + HUGEPAGE_SIZE - 1) & -HUGEPAGE_SIZE;
101 		best_end = best_end & -HUGEPAGE_SIZE;
102 		phys_alloc_init(best_start, best_end - best_start);
103 	}
104 }
105 
106 /* Setup TSS for the current processor, and return TSS offset within GDT */
107 unsigned long setup_tss(u8 *stacktop)
108 {
109 	u32 id;
110 	tss64_t *tss_entry;
111 
112 	id = pre_boot_apic_id();
113 
114 	/* Runtime address of current TSS */
115 	tss_entry = &tss[id];
116 
117 	/* Update TSS */
118 	memset((void *)tss_entry, 0, sizeof(tss64_t));
119 
120 	/* Update TSS descriptors; each descriptor takes up 2 entries */
121 	set_gdt_entry(TSS_MAIN + id * 16, (unsigned long)tss_entry, 0xffff, 0x89, 0);
122 
123 	return TSS_MAIN + id * 16;
124 }
125 #else
126 /* Setup TSS for the current processor, and return TSS offset within GDT */
127 unsigned long setup_tss(u8 *stacktop)
128 {
129 	u32 id;
130 	tss32_t *tss_entry;
131 
132 	id = pre_boot_apic_id();
133 
134 	/* Runtime address of current TSS */
135 	tss_entry = &tss[id];
136 
137 	/* Update TSS */
138 	memset((void *)tss_entry, 0, sizeof(tss32_t));
139 	tss_entry->ss0 = KERNEL_DS;
140 
141 	/* Update descriptors for TSS and percpu data segment.  */
142 	set_gdt_entry(TSS_MAIN + id * 8,
143 		      (unsigned long)tss_entry, 0xffff, 0x89, 0);
144 	set_gdt_entry(TSS_MAIN + MAX_TEST_CPUS * 8 + id * 8,
145 		      (unsigned long)stacktop - 4096, 0xfffff, 0x93, 0xc0);
146 
147 	return TSS_MAIN + id * 8;
148 }
149 #endif
150 
151 void setup_multiboot(struct mbi_bootinfo *bi)
152 {
153 	struct mbi_module *mods;
154 
155 	bootinfo = bi;
156 
157 	u64 best_start = (uintptr_t) &edata;
158 	u64 best_end = bootinfo->mem_upper * 1024ull;
159 	phys_alloc_init(best_start, best_end - best_start);
160 
161 	if (bootinfo->mods_count != 1)
162 		return;
163 
164 	mods = (struct mbi_module *)(uintptr_t) bootinfo->mods_addr;
165 
166 	initrd = (char *)(uintptr_t) mods->start;
167 	initrd_size = mods->end - mods->start;
168 }
169 
170 #ifdef CONFIG_EFI
171 
172 /* From x86/efi/efistart64.S */
173 extern void load_gdt_tss(size_t tss_offset);
174 
175 static efi_status_t setup_memory_allocator(efi_bootinfo_t *efi_bootinfo)
176 {
177 	int i;
178 	unsigned long free_mem_pages = 0;
179 	unsigned long free_mem_start = 0;
180 	struct efi_boot_memmap *map = &(efi_bootinfo->mem_map);
181 	efi_memory_desc_t *buffer = *map->map;
182 	efi_memory_desc_t *d = NULL;
183 
184 	/*
185 	 * The 'buffer' contains multiple descriptors that describe memory
186 	 * regions maintained by UEFI. This code records the largest free
187 	 * EFI_CONVENTIONAL_MEMORY region which will be used to set up the
188 	 * memory allocator, so that the memory allocator can work in the
189 	 * largest free continuous memory region.
190 	 */
191 	for (i = 0; i < *(map->map_size); i += *(map->desc_size)) {
192 		d = (efi_memory_desc_t *)(&((u8 *)buffer)[i]);
193 		if (d->type == EFI_CONVENTIONAL_MEMORY) {
194 			if (free_mem_pages < d->num_pages) {
195 				free_mem_pages = d->num_pages;
196 				free_mem_start = d->phys_addr;
197 			}
198 		}
199 	}
200 
201 	if (free_mem_pages == 0) {
202 		return EFI_OUT_OF_RESOURCES;
203 	}
204 
205 	phys_alloc_init(free_mem_start, free_mem_pages << EFI_PAGE_SHIFT);
206 
207 	return EFI_SUCCESS;
208 }
209 
210 static efi_status_t setup_rsdp(efi_bootinfo_t *efi_bootinfo)
211 {
212 	efi_status_t status;
213 	struct rsdp_descriptor *rsdp;
214 
215 	/*
216 	 * RSDP resides in an EFI_ACPI_RECLAIM_MEMORY region, which is not used
217 	 * by kvm-unit-tests x86's memory allocator. So it is not necessary to
218 	 * copy the data structure to another memory region to prevent
219 	 * unintentional overwrite.
220 	 */
221 	status = efi_get_system_config_table(ACPI_TABLE_GUID, (void **)&rsdp);
222 	if (status != EFI_SUCCESS) {
223 		return status;
224 	}
225 
226 	set_efi_rsdp(rsdp);
227 
228 	return EFI_SUCCESS;
229 }
230 
231 /* Defined in cstart64.S or efistart64.S */
232 extern u8 ptl4;
233 extern u8 ptl3;
234 extern u8 ptl2;
235 
236 static void setup_page_table(void)
237 {
238 	pgd_t *curr_pt;
239 	phys_addr_t flags;
240 	int i;
241 
242 	/* Set default flags */
243 	flags = PT_PRESENT_MASK | PT_WRITABLE_MASK | PT_USER_MASK;
244 
245 	/* Set AMD SEV C-Bit for page table entries */
246 	flags |= get_amd_sev_c_bit_mask();
247 
248 	/* Level 4 */
249 	curr_pt = (pgd_t *)&ptl4;
250 	curr_pt[0] = ((phys_addr_t)&ptl3) | flags;
251 	/* Level 3 */
252 	curr_pt = (pgd_t *)&ptl3;
253 	for (i = 0; i < 4; i++) {
254 		curr_pt[i] = (((phys_addr_t)&ptl2) + i * PAGE_SIZE) | flags;
255 	}
256 	/* Level 2 */
257 	curr_pt = (pgd_t *)&ptl2;
258 	flags |= PT_ACCESSED_MASK | PT_DIRTY_MASK | PT_PAGE_SIZE_MASK | PT_GLOBAL_MASK;
259 	for (i = 0; i < 4 * 512; i++)	{
260 		curr_pt[i] = ((phys_addr_t) i << 21) | flags;
261 	}
262 
263 	if (amd_sev_es_enabled()) {
264 		setup_ghcb_pte((pgd_t *)&ptl4);
265 	}
266 
267 	/* Load 4-level page table */
268 	write_cr3((ulong)&ptl4);
269 }
270 
271 static void setup_gdt_tss(void)
272 {
273 	size_t tss_offset;
274 
275 	/* 64-bit setup_tss does not use the stacktop argument.  */
276 	tss_offset = setup_tss(NULL);
277 	load_gdt_tss(tss_offset);
278 }
279 
280 efi_status_t setup_efi(efi_bootinfo_t *efi_bootinfo)
281 {
282 	efi_status_t status;
283 	const char *phase;
284 
285 	status = setup_memory_allocator(efi_bootinfo);
286 	if (status != EFI_SUCCESS) {
287 		printf("Failed to set up memory allocator: ");
288 		switch (status) {
289 		case EFI_OUT_OF_RESOURCES:
290 			printf("No free memory region\n");
291 			break;
292 		default:
293 			printf("Unknown error\n");
294 			break;
295 		}
296 		return status;
297 	}
298 
299 	status = setup_rsdp(efi_bootinfo);
300 	if (status != EFI_SUCCESS) {
301 		printf("Cannot find RSDP in EFI system table\n");
302 		return status;
303 	}
304 
305 	phase = "AMD SEV";
306 	status = setup_amd_sev();
307 
308 	/* Continue if AMD SEV is not supported, but skip SEV-ES setup */
309 	if (status == EFI_SUCCESS) {
310 		phase = "AMD SEV-ES";
311 		status = setup_amd_sev_es();
312 	}
313 
314 	if (status != EFI_SUCCESS && status != EFI_UNSUPPORTED) {
315 		printf("%s setup failed, error = 0x%lx\n", phase, status);
316 		return status;
317 	}
318 
319 	reset_apic();
320 	setup_gdt_tss();
321 	setup_idt();
322 	load_idt();
323 	mask_pic_interrupts();
324 	enable_apic();
325 	ap_init();
326 	enable_x2apic();
327 	smp_init();
328 	setup_page_table();
329 
330 	return EFI_SUCCESS;
331 }
332 
333 #endif /* CONFIG_EFI */
334 
335 void setup_libcflat(void)
336 {
337 	if (initrd) {
338 		/* environ is currently the only file in the initrd */
339 		u32 size = MIN(initrd_size, ENV_SIZE);
340 		const char *str;
341 
342 		memcpy(env, initrd, size);
343 		setup_env(env, size);
344 		if ((str = getenv("BOOTLOADER")) && atol(str) != 0)
345 			add_setup_arg("bootloader");
346 	}
347 }
348