1 /*- 2 * Copyright (c) 2013 The FreeBSD Foundation 3 * All rights reserved. 4 * 5 * This software was developed by Benno Rice under sponsorship from 6 * the FreeBSD Foundation. 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 18 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 19 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 20 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 21 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 22 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 24 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 25 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 26 * SUCH DAMAGE. 27 */ 28 29 #include <sys/cdefs.h> 30 __FBSDID("$FreeBSD$"); 31 32 #include <sys/param.h> 33 34 #include <stand.h> 35 #include <bootstrap.h> 36 37 #include <efi.h> 38 #include <efilib.h> 39 40 #include "loader_efi.h" 41 42 #if defined(__i386__) || defined(__amd64__) 43 #include <machine/cpufunc.h> 44 #include <machine/specialreg.h> 45 46 /* 47 * The code is excerpted from sys/x86/x86/identcpu.c: identify_cpu(), 48 * identify_hypervisor(), and dev/hyperv/vmbus/hyperv.c: hyperv_identify(). 49 */ 50 #define CPUID_LEAF_HV_MAXLEAF 0x40000000 51 #define CPUID_LEAF_HV_INTERFACE 0x40000001 52 #define CPUID_LEAF_HV_FEATURES 0x40000003 53 #define CPUID_LEAF_HV_LIMITS 0x40000005 54 #define CPUID_HV_IFACE_HYPERV 0x31237648 /* HV#1 */ 55 #define CPUID_HV_MSR_TIME_REFCNT 0x0002 /* MSR_HV_TIME_REF_COUNT */ 56 #define CPUID_HV_MSR_HYPERCALL 0x0020 57 58 static int 59 running_on_hyperv(void) 60 { 61 char hv_vendor[16]; 62 uint32_t regs[4]; 63 64 do_cpuid(1, regs); 65 if ((regs[2] & CPUID2_HV) == 0) 66 return (0); 67 68 do_cpuid(CPUID_LEAF_HV_MAXLEAF, regs); 69 if (regs[0] < CPUID_LEAF_HV_LIMITS) 70 return (0); 71 72 ((uint32_t *)&hv_vendor)[0] = regs[1]; 73 ((uint32_t *)&hv_vendor)[1] = regs[2]; 74 ((uint32_t *)&hv_vendor)[2] = regs[3]; 75 hv_vendor[12] = '\0'; 76 if (strcmp(hv_vendor, "Microsoft Hv") != 0) 77 return (0); 78 79 do_cpuid(CPUID_LEAF_HV_INTERFACE, regs); 80 if (regs[0] != CPUID_HV_IFACE_HYPERV) 81 return (0); 82 83 do_cpuid(CPUID_LEAF_HV_FEATURES, regs); 84 if ((regs[0] & CPUID_HV_MSR_HYPERCALL) == 0) 85 return (0); 86 if ((regs[0] & CPUID_HV_MSR_TIME_REFCNT) == 0) 87 return (0); 88 89 return (1); 90 } 91 92 #define KERNEL_PHYSICAL_BASE (2*1024*1024) 93 94 static void 95 efi_verify_staging_size(unsigned long *nr_pages) 96 { 97 UINTN sz; 98 EFI_MEMORY_DESCRIPTOR *map = NULL, *p; 99 EFI_PHYSICAL_ADDRESS start, end; 100 UINTN key, dsz; 101 UINT32 dver; 102 EFI_STATUS status; 103 int i, ndesc; 104 unsigned long available_pages = 0; 105 106 sz = 0; 107 108 for (;;) { 109 status = BS->GetMemoryMap(&sz, map, &key, &dsz, &dver); 110 if (!EFI_ERROR(status)) 111 break; 112 113 if (status != EFI_BUFFER_TOO_SMALL) { 114 printf("Can't read memory map: %lu\n", 115 EFI_ERROR_CODE(status)); 116 goto out; 117 } 118 119 free(map); 120 121 /* Allocate 10 descriptors more than the size reported, 122 * to allow for any fragmentation caused by calling 123 * malloc */ 124 map = malloc(sz + (10 * dsz)); 125 if (map == NULL) { 126 printf("Unable to allocate memory\n"); 127 goto out; 128 } 129 } 130 131 ndesc = sz / dsz; 132 for (i = 0, p = map; i < ndesc; 133 i++, p = NextMemoryDescriptor(p, dsz)) { 134 start = p->PhysicalStart; 135 end = start + p->NumberOfPages * EFI_PAGE_SIZE; 136 137 if (KERNEL_PHYSICAL_BASE < start || 138 KERNEL_PHYSICAL_BASE >= end) 139 continue; 140 141 available_pages = p->NumberOfPages - 142 ((KERNEL_PHYSICAL_BASE - start) >> EFI_PAGE_SHIFT); 143 break; 144 } 145 146 if (available_pages == 0) { 147 printf("Can't find valid memory map for staging area!\n"); 148 goto out; 149 } 150 151 i++; 152 p = NextMemoryDescriptor(p, dsz); 153 154 for ( ; i < ndesc; 155 i++, p = NextMemoryDescriptor(p, dsz)) { 156 if (p->Type != EfiConventionalMemory && 157 p->Type != EfiLoaderData) 158 break; 159 160 if (p->PhysicalStart != end) 161 break; 162 163 end = p->PhysicalStart + p->NumberOfPages * EFI_PAGE_SIZE; 164 165 available_pages += p->NumberOfPages; 166 } 167 168 if (*nr_pages > available_pages) { 169 printf("Staging area's size is reduced: %ld -> %ld!\n", 170 *nr_pages, available_pages); 171 *nr_pages = available_pages; 172 } 173 out: 174 free(map); 175 } 176 #endif /* __i386__ || __amd64__ */ 177 178 #ifndef EFI_STAGING_SIZE 179 #if defined(__amd64__) 180 #define EFI_STAGING_SIZE 100 181 #elif defined(__arm__) 182 #define EFI_STAGING_SIZE 32 183 #else 184 #define EFI_STAGING_SIZE 64 185 #endif 186 #endif 187 188 EFI_PHYSICAL_ADDRESS staging, staging_end, staging_base; 189 int stage_offset_set = 0; 190 ssize_t stage_offset; 191 192 int 193 efi_copy_init(void) 194 { 195 EFI_STATUS status; 196 197 unsigned long nr_pages; 198 199 nr_pages = EFI_SIZE_TO_PAGES((EFI_STAGING_SIZE) * 1024 * 1024); 200 201 #if defined(__i386__) || defined(__amd64__) 202 /* 203 * We'll decrease nr_pages, if it's too big. Currently we only 204 * apply this to FreeBSD VM running on Hyper-V. Why? Please see 205 * https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=211746#c28 206 */ 207 if (running_on_hyperv()) 208 efi_verify_staging_size(&nr_pages); 209 210 /* 211 * The staging area must reside in the the first 1GB physical 212 * memory: see elf64_exec() in 213 * boot/efi/loader/arch/amd64/elf64_freebsd.c. 214 */ 215 staging = 1024*1024*1024; 216 status = BS->AllocatePages(AllocateMaxAddress, EfiLoaderData, 217 nr_pages, &staging); 218 #else 219 status = BS->AllocatePages(AllocateAnyPages, EfiLoaderData, 220 nr_pages, &staging); 221 #endif 222 if (EFI_ERROR(status)) { 223 printf("failed to allocate staging area: %lu\n", 224 EFI_ERROR_CODE(status)); 225 return (status); 226 } 227 staging_base = staging; 228 staging_end = staging + nr_pages * EFI_PAGE_SIZE; 229 230 #if defined(__aarch64__) || defined(__arm__) || defined(__riscv) 231 /* 232 * Round the kernel load address to a 2MiB value. This is needed 233 * because the kernel builds a page table based on where it has 234 * been loaded in physical address space. As the kernel will use 235 * either a 1MiB or 2MiB page for this we need to make sure it 236 * is correctly aligned for both cases. 237 */ 238 staging = roundup2(staging, 2 * 1024 * 1024); 239 #endif 240 241 return (0); 242 } 243 244 static bool 245 efi_check_space(vm_offset_t end) 246 { 247 EFI_PHYSICAL_ADDRESS addr; 248 EFI_STATUS status; 249 unsigned long nr_pages; 250 251 /* There is already enough space */ 252 if (end <= staging_end) 253 return (true); 254 255 end = roundup2(end, EFI_PAGE_SIZE); 256 nr_pages = EFI_SIZE_TO_PAGES(end - staging_end); 257 258 #if defined(__i386__) || defined(__amd64__) 259 /* X86 needs all memory to be allocated under the 1G boundary */ 260 if (end > 1024*1024*1024) 261 goto before_staging; 262 #endif 263 264 /* Try to allocate more space after the previous allocation */ 265 addr = staging_end; 266 status = BS->AllocatePages(AllocateAddress, EfiLoaderData, nr_pages, 267 &addr); 268 if (!EFI_ERROR(status)) { 269 staging_end = staging_end + nr_pages * EFI_PAGE_SIZE; 270 return (true); 271 } 272 273 before_staging: 274 /* Try allocating space before the previous allocation */ 275 if (staging < nr_pages * EFI_PAGE_SIZE) { 276 printf("Not enough space before allocation\n"); 277 return (false); 278 } 279 addr = staging - nr_pages * EFI_PAGE_SIZE; 280 #if defined(__aarch64__) || defined(__arm__) || defined(__riscv) 281 /* See efi_copy_init for why this is needed */ 282 addr = rounddown2(addr, 2 * 1024 * 1024); 283 #endif 284 nr_pages = EFI_SIZE_TO_PAGES(staging_base - addr); 285 status = BS->AllocatePages(AllocateAddress, EfiLoaderData, nr_pages, 286 &addr); 287 if (!EFI_ERROR(status)) { 288 /* 289 * Move the old allocation and update the state so 290 * translation still works. 291 */ 292 staging_base = addr; 293 memmove((void *)staging_base, (void *)staging, 294 staging_end - staging); 295 stage_offset -= (staging - staging_base); 296 staging = staging_base; 297 return (true); 298 } 299 300 printf("efi_check_space: Unable to expand staging area\n"); 301 return (false); 302 } 303 304 void * 305 efi_translate(vm_offset_t ptr) 306 { 307 308 return ((void *)(ptr + stage_offset)); 309 } 310 311 ssize_t 312 efi_copyin(const void *src, vm_offset_t dest, const size_t len) 313 { 314 315 if (!stage_offset_set) { 316 stage_offset = (vm_offset_t)staging - dest; 317 stage_offset_set = 1; 318 } 319 320 /* XXX: Callers do not check for failure. */ 321 if (!efi_check_space(dest + stage_offset + len)) { 322 errno = ENOMEM; 323 return (-1); 324 } 325 bcopy(src, (void *)(dest + stage_offset), len); 326 return (len); 327 } 328 329 ssize_t 330 efi_copyout(const vm_offset_t src, void *dest, const size_t len) 331 { 332 333 /* XXX: Callers do not check for failure. */ 334 if (src + stage_offset + len > staging_end) { 335 errno = ENOMEM; 336 return (-1); 337 } 338 bcopy((void *)(src + stage_offset), dest, len); 339 return (len); 340 } 341 342 343 ssize_t 344 efi_readin(readin_handle_t fd, vm_offset_t dest, const size_t len) 345 { 346 347 if (!stage_offset_set) { 348 stage_offset = (vm_offset_t)staging - dest; 349 stage_offset_set = 1; 350 } 351 352 if (!efi_check_space(dest + stage_offset + len)) { 353 errno = ENOMEM; 354 return (-1); 355 } 356 return (VECTX_READ(fd, (void *)(dest + stage_offset), len)); 357 } 358 359 void 360 efi_copy_finish(void) 361 { 362 uint64_t *src, *dst, *last; 363 364 src = (uint64_t *)(uintptr_t)staging; 365 dst = (uint64_t *)(uintptr_t)(staging - stage_offset); 366 last = (uint64_t *)(uintptr_t)staging_end; 367 368 while (src < last) 369 *dst++ = *src++; 370 } 371