]>
Commit | Line | Data |
---|---|---|
abe1ee3a | 1 | |
a7f290da | 2 | /* |
a7f290da BH |
3 | * Copyright (C) 2004 Benjamin Herrenschmidt, IBM Corp. |
4 | * <benh@kernel.crashing.org> | |
5 | * | |
6 | * This program is free software; you can redistribute it and/or | |
7 | * modify it under the terms of the GNU General Public License | |
8 | * as published by the Free Software Foundation; either version | |
9 | * 2 of the License, or (at your option) any later version. | |
10 | */ | |
11 | ||
a7f290da BH |
12 | #include <linux/errno.h> |
13 | #include <linux/sched.h> | |
14 | #include <linux/kernel.h> | |
15 | #include <linux/mm.h> | |
16 | #include <linux/smp.h> | |
a7f290da BH |
17 | #include <linux/stddef.h> |
18 | #include <linux/unistd.h> | |
19 | #include <linux/slab.h> | |
20 | #include <linux/user.h> | |
21 | #include <linux/elf.h> | |
22 | #include <linux/security.h> | |
23 | #include <linux/bootmem.h> | |
95f72d1e | 24 | #include <linux/memblock.h> |
a7f290da BH |
25 | |
26 | #include <asm/pgtable.h> | |
a7f290da BH |
27 | #include <asm/processor.h> |
28 | #include <asm/mmu.h> | |
29 | #include <asm/mmu_context.h> | |
d9b2b2a2 | 30 | #include <asm/prom.h> |
a7f290da BH |
31 | #include <asm/machdep.h> |
32 | #include <asm/cputable.h> | |
33 | #include <asm/sections.h> | |
e8222502 | 34 | #include <asm/firmware.h> |
a7f290da BH |
35 | #include <asm/vdso.h> |
36 | #include <asm/vdso_datapage.h> | |
37 | ||
0909c8c2 BH |
38 | #include "setup.h" |
39 | ||
a7f290da BH |
40 | #undef DEBUG |
41 | ||
42 | #ifdef DEBUG | |
43 | #define DBG(fmt...) printk(fmt) | |
44 | #else | |
45 | #define DBG(fmt...) | |
46 | #endif | |
47 | ||
48 | /* Max supported size for symbol names */ | |
49 | #define MAX_SYMNAME 64 | |
50 | ||
348aa303 AS |
51 | /* The alignment of the vDSO */ |
52 | #define VDSO_ALIGNMENT (1 << 16) | |
53 | ||
a7f290da BH |
54 | extern char vdso32_start, vdso32_end; |
55 | static void *vdso32_kbase = &vdso32_start; | |
7ac9a137 BH |
56 | static unsigned int vdso32_pages; |
57 | static struct page **vdso32_pagelist; | |
a7f290da BH |
58 | unsigned long vdso32_sigtramp; |
59 | unsigned long vdso32_rt_sigtramp; | |
60 | ||
61 | #ifdef CONFIG_PPC64 | |
62 | extern char vdso64_start, vdso64_end; | |
63 | static void *vdso64_kbase = &vdso64_start; | |
7ac9a137 BH |
64 | static unsigned int vdso64_pages; |
65 | static struct page **vdso64_pagelist; | |
a7f290da BH |
66 | unsigned long vdso64_rt_sigtramp; |
67 | #endif /* CONFIG_PPC64 */ | |
68 | ||
7ac9a137 BH |
69 | static int vdso_ready; |
70 | ||
a7f290da BH |
71 | /* |
72 | * The vdso data page (aka. systemcfg for old ppc64 fans) is here. | |
73 | * Once the early boot kernel code no longer needs to muck around | |
74 | * with it, it will become dynamically allocated | |
75 | */ | |
76 | static union { | |
77 | struct vdso_data data; | |
78 | u8 page[PAGE_SIZE]; | |
abe1ee3a | 79 | } vdso_data_store __page_aligned_data; |
a7f290da BH |
80 | struct vdso_data *vdso_data = &vdso_data_store.data; |
81 | ||
82 | /* Format of the patch table */ | |
83 | struct vdso_patch_def | |
84 | { | |
85 | unsigned long ftr_mask, ftr_value; | |
86 | const char *gen_name; | |
87 | const char *fix_name; | |
88 | }; | |
89 | ||
90 | /* Table of functions to patch based on the CPU type/revision | |
91 | * | |
92 | * Currently, we only change sync_dicache to do nothing on processors | |
93 | * with a coherent icache | |
94 | */ | |
95 | static struct vdso_patch_def vdso_patches[] = { | |
96 | { | |
97 | CPU_FTR_COHERENT_ICACHE, CPU_FTR_COHERENT_ICACHE, | |
98 | "__kernel_sync_dicache", "__kernel_sync_dicache_p5" | |
99 | }, | |
100 | { | |
101 | CPU_FTR_USE_TB, 0, | |
102 | "__kernel_gettimeofday", NULL | |
103 | }, | |
7b5acbaa BH |
104 | { |
105 | CPU_FTR_USE_TB, 0, | |
106 | "__kernel_clock_gettime", NULL | |
107 | }, | |
108 | { | |
109 | CPU_FTR_USE_TB, 0, | |
110 | "__kernel_clock_getres", NULL | |
111 | }, | |
112 | { | |
113 | CPU_FTR_USE_TB, 0, | |
114 | "__kernel_get_tbfreq", NULL | |
115 | }, | |
fcb41a20 AZ |
116 | { |
117 | CPU_FTR_USE_TB, 0, | |
118 | "__kernel_time", NULL | |
119 | }, | |
a7f290da BH |
120 | }; |
121 | ||
122 | /* | |
123 | * Some infos carried around for each of them during parsing at | |
124 | * boot time. | |
125 | */ | |
126 | struct lib32_elfinfo | |
127 | { | |
128 | Elf32_Ehdr *hdr; /* ptr to ELF */ | |
129 | Elf32_Sym *dynsym; /* ptr to .dynsym section */ | |
130 | unsigned long dynsymsize; /* size of .dynsym section */ | |
131 | char *dynstr; /* ptr to .dynstr section */ | |
132 | unsigned long text; /* offset of .text section in .so */ | |
133 | }; | |
134 | ||
135 | struct lib64_elfinfo | |
136 | { | |
137 | Elf64_Ehdr *hdr; | |
138 | Elf64_Sym *dynsym; | |
139 | unsigned long dynsymsize; | |
140 | char *dynstr; | |
141 | unsigned long text; | |
142 | }; | |
143 | ||
144 | ||
145 | #ifdef __DEBUG | |
146 | static void dump_one_vdso_page(struct page *pg, struct page *upg) | |
147 | { | |
148 | printk("kpg: %p (c:%d,f:%08lx)", __va(page_to_pfn(pg) << PAGE_SHIFT), | |
149 | page_count(pg), | |
150 | pg->flags); | |
89f5b7da | 151 | if (upg && !IS_ERR(upg) /* && pg != upg*/) { |
a7f290da BH |
152 | printk(" upg: %p (c:%d,f:%08lx)", __va(page_to_pfn(upg) |
153 | << PAGE_SHIFT), | |
154 | page_count(upg), | |
155 | upg->flags); | |
156 | } | |
157 | printk("\n"); | |
158 | } | |
159 | ||
160 | static void dump_vdso_pages(struct vm_area_struct * vma) | |
161 | { | |
162 | int i; | |
163 | ||
cab175f9 | 164 | if (!vma || is_32bit_task()) { |
a7f290da BH |
165 | printk("vDSO32 @ %016lx:\n", (unsigned long)vdso32_kbase); |
166 | for (i=0; i<vdso32_pages; i++) { | |
167 | struct page *pg = virt_to_page(vdso32_kbase + | |
168 | i*PAGE_SIZE); | |
169 | struct page *upg = (vma && vma->vm_mm) ? | |
6aab341e | 170 | follow_page(vma, vma->vm_start + i*PAGE_SIZE, 0) |
a7f290da BH |
171 | : NULL; |
172 | dump_one_vdso_page(pg, upg); | |
173 | } | |
174 | } | |
cab175f9 | 175 | if (!vma || !is_32bit_task()) { |
a7f290da BH |
176 | printk("vDSO64 @ %016lx:\n", (unsigned long)vdso64_kbase); |
177 | for (i=0; i<vdso64_pages; i++) { | |
178 | struct page *pg = virt_to_page(vdso64_kbase + | |
179 | i*PAGE_SIZE); | |
180 | struct page *upg = (vma && vma->vm_mm) ? | |
6aab341e | 181 | follow_page(vma, vma->vm_start + i*PAGE_SIZE, 0) |
a7f290da BH |
182 | : NULL; |
183 | dump_one_vdso_page(pg, upg); | |
184 | } | |
185 | } | |
186 | } | |
187 | #endif /* DEBUG */ | |
188 | ||
a7f290da BH |
189 | /* |
190 | * This is called from binfmt_elf, we create the special vma for the | |
191 | * vDSO and insert it into the mm struct tree | |
192 | */ | |
fc5243d9 | 193 | int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp) |
a7f290da BH |
194 | { |
195 | struct mm_struct *mm = current->mm; | |
c13e4ca2 | 196 | struct page **vdso_pagelist; |
a7f290da BH |
197 | unsigned long vdso_pages; |
198 | unsigned long vdso_base; | |
a5bba930 | 199 | int rc; |
a7f290da | 200 | |
7ac9a137 BH |
201 | if (!vdso_ready) |
202 | return 0; | |
203 | ||
a7f290da | 204 | #ifdef CONFIG_PPC64 |
cab175f9 | 205 | if (is_32bit_task()) { |
c13e4ca2 | 206 | vdso_pagelist = vdso32_pagelist; |
a7f290da BH |
207 | vdso_pages = vdso32_pages; |
208 | vdso_base = VDSO32_MBASE; | |
209 | } else { | |
c13e4ca2 | 210 | vdso_pagelist = vdso64_pagelist; |
a7f290da | 211 | vdso_pages = vdso64_pages; |
30d0b368 AB |
212 | /* |
213 | * On 64bit we don't have a preferred map address. This | |
214 | * allows get_unmapped_area to find an area near other mmaps | |
215 | * and most likely share a SLB entry. | |
216 | */ | |
217 | vdso_base = 0; | |
a7f290da BH |
218 | } |
219 | #else | |
c13e4ca2 | 220 | vdso_pagelist = vdso32_pagelist; |
a7f290da BH |
221 | vdso_pages = vdso32_pages; |
222 | vdso_base = VDSO32_MBASE; | |
223 | #endif | |
224 | ||
a5bba930 | 225 | current->mm->context.vdso_base = 0; |
a7f290da BH |
226 | |
227 | /* vDSO has a problem and was disabled, just don't "enable" it for the | |
228 | * process | |
229 | */ | |
230 | if (vdso_pages == 0) | |
231 | return 0; | |
a7f290da BH |
232 | /* Add a page to the vdso size for the data page */ |
233 | vdso_pages ++; | |
234 | ||
235 | /* | |
236 | * pick a base address for the vDSO in process space. We try to put it | |
237 | * at vdso_base which is the "natural" base for it, but we might fail | |
238 | * and end up putting it elsewhere. | |
348aa303 | 239 | * Add enough to the size so that the result can be aligned. |
a7f290da | 240 | */ |
a5bba930 | 241 | down_write(&mm->mmap_sem); |
a7f290da | 242 | vdso_base = get_unmapped_area(NULL, vdso_base, |
348aa303 AS |
243 | (vdso_pages << PAGE_SHIFT) + |
244 | ((VDSO_ALIGNMENT - 1) & PAGE_MASK), | |
245 | 0, 0); | |
a5bba930 BH |
246 | if (IS_ERR_VALUE(vdso_base)) { |
247 | rc = vdso_base; | |
248 | goto fail_mmapsem; | |
a7f290da BH |
249 | } |
250 | ||
348aa303 AS |
251 | /* Add required alignment. */ |
252 | vdso_base = ALIGN(vdso_base, VDSO_ALIGNMENT); | |
253 | ||
f2053f1a AB |
254 | /* |
255 | * Put vDSO base into mm struct. We need to do this before calling | |
256 | * install_special_mapping or the perf counter mmap tracking code | |
257 | * will fail to recognise it as a vDSO (since arch_vma_name fails). | |
258 | */ | |
259 | current->mm->context.vdso_base = vdso_base; | |
260 | ||
a7f290da BH |
261 | /* |
262 | * our vma flags don't have VM_WRITE so by default, the process isn't | |
263 | * allowed to write those pages. | |
264 | * gdb can break that with ptrace interface, and thus trigger COW on | |
265 | * those pages but it's then your responsibility to never do that on | |
266 | * the "data" page of the vDSO or you'll stop getting kernel updates | |
267 | * and your nice userland gettimeofday will be totally dead. | |
268 | * It's fine to use that for setting breakpoints in the vDSO code | |
909af768 | 269 | * pages though. |
3a0cfadb | 270 | */ |
c13e4ca2 RM |
271 | rc = install_special_mapping(mm, vdso_base, vdso_pages << PAGE_SHIFT, |
272 | VM_READ|VM_EXEC| | |
909af768 | 273 | VM_MAYREAD|VM_MAYWRITE|VM_MAYEXEC, |
c13e4ca2 | 274 | vdso_pagelist); |
f2053f1a AB |
275 | if (rc) { |
276 | current->mm->context.vdso_base = 0; | |
c13e4ca2 | 277 | goto fail_mmapsem; |
f2053f1a | 278 | } |
c13e4ca2 | 279 | |
a7f290da | 280 | up_write(&mm->mmap_sem); |
a7f290da | 281 | return 0; |
a5bba930 | 282 | |
a5bba930 BH |
283 | fail_mmapsem: |
284 | up_write(&mm->mmap_sem); | |
285 | return rc; | |
286 | } | |
287 | ||
288 | const char *arch_vma_name(struct vm_area_struct *vma) | |
289 | { | |
290 | if (vma->vm_mm && vma->vm_start == vma->vm_mm->context.vdso_base) | |
291 | return "[vdso]"; | |
292 | return NULL; | |
a7f290da BH |
293 | } |
294 | ||
a5bba930 BH |
295 | |
296 | ||
a7f290da BH |
297 | static void * __init find_section32(Elf32_Ehdr *ehdr, const char *secname, |
298 | unsigned long *size) | |
299 | { | |
300 | Elf32_Shdr *sechdrs; | |
301 | unsigned int i; | |
302 | char *secnames; | |
303 | ||
304 | /* Grab section headers and strings so we can tell who is who */ | |
305 | sechdrs = (void *)ehdr + ehdr->e_shoff; | |
306 | secnames = (void *)ehdr + sechdrs[ehdr->e_shstrndx].sh_offset; | |
307 | ||
308 | /* Find the section they want */ | |
309 | for (i = 1; i < ehdr->e_shnum; i++) { | |
310 | if (strcmp(secnames+sechdrs[i].sh_name, secname) == 0) { | |
311 | if (size) | |
312 | *size = sechdrs[i].sh_size; | |
313 | return (void *)ehdr + sechdrs[i].sh_offset; | |
314 | } | |
315 | } | |
316 | *size = 0; | |
317 | return NULL; | |
318 | } | |
319 | ||
320 | static Elf32_Sym * __init find_symbol32(struct lib32_elfinfo *lib, | |
321 | const char *symname) | |
322 | { | |
323 | unsigned int i; | |
324 | char name[MAX_SYMNAME], *c; | |
325 | ||
326 | for (i = 0; i < (lib->dynsymsize / sizeof(Elf32_Sym)); i++) { | |
327 | if (lib->dynsym[i].st_name == 0) | |
328 | continue; | |
329 | strlcpy(name, lib->dynstr + lib->dynsym[i].st_name, | |
330 | MAX_SYMNAME); | |
331 | c = strchr(name, '@'); | |
332 | if (c) | |
333 | *c = 0; | |
334 | if (strcmp(symname, name) == 0) | |
335 | return &lib->dynsym[i]; | |
336 | } | |
337 | return NULL; | |
338 | } | |
339 | ||
340 | /* Note that we assume the section is .text and the symbol is relative to | |
341 | * the library base | |
342 | */ | |
343 | static unsigned long __init find_function32(struct lib32_elfinfo *lib, | |
344 | const char *symname) | |
345 | { | |
346 | Elf32_Sym *sym = find_symbol32(lib, symname); | |
347 | ||
348 | if (sym == NULL) { | |
349 | printk(KERN_WARNING "vDSO32: function %s not found !\n", | |
350 | symname); | |
351 | return 0; | |
352 | } | |
353 | return sym->st_value - VDSO32_LBASE; | |
354 | } | |
355 | ||
cf8918fe AB |
356 | static int __init vdso_do_func_patch32(struct lib32_elfinfo *v32, |
357 | struct lib64_elfinfo *v64, | |
358 | const char *orig, const char *fix) | |
a7f290da BH |
359 | { |
360 | Elf32_Sym *sym32_gen, *sym32_fix; | |
361 | ||
362 | sym32_gen = find_symbol32(v32, orig); | |
363 | if (sym32_gen == NULL) { | |
364 | printk(KERN_ERR "vDSO32: Can't find symbol %s !\n", orig); | |
365 | return -1; | |
366 | } | |
367 | if (fix == NULL) { | |
368 | sym32_gen->st_name = 0; | |
369 | return 0; | |
370 | } | |
371 | sym32_fix = find_symbol32(v32, fix); | |
372 | if (sym32_fix == NULL) { | |
373 | printk(KERN_ERR "vDSO32: Can't find symbol %s !\n", fix); | |
374 | return -1; | |
375 | } | |
376 | sym32_gen->st_value = sym32_fix->st_value; | |
377 | sym32_gen->st_size = sym32_fix->st_size; | |
378 | sym32_gen->st_info = sym32_fix->st_info; | |
379 | sym32_gen->st_other = sym32_fix->st_other; | |
380 | sym32_gen->st_shndx = sym32_fix->st_shndx; | |
381 | ||
382 | return 0; | |
383 | } | |
384 | ||
385 | ||
386 | #ifdef CONFIG_PPC64 | |
387 | ||
388 | static void * __init find_section64(Elf64_Ehdr *ehdr, const char *secname, | |
389 | unsigned long *size) | |
390 | { | |
391 | Elf64_Shdr *sechdrs; | |
392 | unsigned int i; | |
393 | char *secnames; | |
394 | ||
395 | /* Grab section headers and strings so we can tell who is who */ | |
396 | sechdrs = (void *)ehdr + ehdr->e_shoff; | |
397 | secnames = (void *)ehdr + sechdrs[ehdr->e_shstrndx].sh_offset; | |
398 | ||
399 | /* Find the section they want */ | |
400 | for (i = 1; i < ehdr->e_shnum; i++) { | |
401 | if (strcmp(secnames+sechdrs[i].sh_name, secname) == 0) { | |
402 | if (size) | |
403 | *size = sechdrs[i].sh_size; | |
404 | return (void *)ehdr + sechdrs[i].sh_offset; | |
405 | } | |
406 | } | |
407 | if (size) | |
408 | *size = 0; | |
409 | return NULL; | |
410 | } | |
411 | ||
412 | static Elf64_Sym * __init find_symbol64(struct lib64_elfinfo *lib, | |
413 | const char *symname) | |
414 | { | |
415 | unsigned int i; | |
416 | char name[MAX_SYMNAME], *c; | |
417 | ||
418 | for (i = 0; i < (lib->dynsymsize / sizeof(Elf64_Sym)); i++) { | |
419 | if (lib->dynsym[i].st_name == 0) | |
420 | continue; | |
421 | strlcpy(name, lib->dynstr + lib->dynsym[i].st_name, | |
422 | MAX_SYMNAME); | |
423 | c = strchr(name, '@'); | |
424 | if (c) | |
425 | *c = 0; | |
426 | if (strcmp(symname, name) == 0) | |
427 | return &lib->dynsym[i]; | |
428 | } | |
429 | return NULL; | |
430 | } | |
431 | ||
432 | /* Note that we assume the section is .text and the symbol is relative to | |
433 | * the library base | |
434 | */ | |
435 | static unsigned long __init find_function64(struct lib64_elfinfo *lib, | |
436 | const char *symname) | |
437 | { | |
438 | Elf64_Sym *sym = find_symbol64(lib, symname); | |
439 | ||
440 | if (sym == NULL) { | |
441 | printk(KERN_WARNING "vDSO64: function %s not found !\n", | |
442 | symname); | |
443 | return 0; | |
444 | } | |
445 | #ifdef VDS64_HAS_DESCRIPTORS | |
446 | return *((u64 *)(vdso64_kbase + sym->st_value - VDSO64_LBASE)) - | |
447 | VDSO64_LBASE; | |
448 | #else | |
449 | return sym->st_value - VDSO64_LBASE; | |
450 | #endif | |
451 | } | |
452 | ||
cf8918fe AB |
453 | static int __init vdso_do_func_patch64(struct lib32_elfinfo *v32, |
454 | struct lib64_elfinfo *v64, | |
455 | const char *orig, const char *fix) | |
a7f290da BH |
456 | { |
457 | Elf64_Sym *sym64_gen, *sym64_fix; | |
458 | ||
459 | sym64_gen = find_symbol64(v64, orig); | |
460 | if (sym64_gen == NULL) { | |
461 | printk(KERN_ERR "vDSO64: Can't find symbol %s !\n", orig); | |
462 | return -1; | |
463 | } | |
464 | if (fix == NULL) { | |
465 | sym64_gen->st_name = 0; | |
466 | return 0; | |
467 | } | |
468 | sym64_fix = find_symbol64(v64, fix); | |
469 | if (sym64_fix == NULL) { | |
470 | printk(KERN_ERR "vDSO64: Can't find symbol %s !\n", fix); | |
471 | return -1; | |
472 | } | |
473 | sym64_gen->st_value = sym64_fix->st_value; | |
474 | sym64_gen->st_size = sym64_fix->st_size; | |
475 | sym64_gen->st_info = sym64_fix->st_info; | |
476 | sym64_gen->st_other = sym64_fix->st_other; | |
477 | sym64_gen->st_shndx = sym64_fix->st_shndx; | |
478 | ||
479 | return 0; | |
480 | } | |
481 | ||
482 | #endif /* CONFIG_PPC64 */ | |
483 | ||
484 | ||
485 | static __init int vdso_do_find_sections(struct lib32_elfinfo *v32, | |
486 | struct lib64_elfinfo *v64) | |
487 | { | |
488 | void *sect; | |
489 | ||
490 | /* | |
491 | * Locate symbol tables & text section | |
492 | */ | |
493 | ||
494 | v32->dynsym = find_section32(v32->hdr, ".dynsym", &v32->dynsymsize); | |
495 | v32->dynstr = find_section32(v32->hdr, ".dynstr", NULL); | |
496 | if (v32->dynsym == NULL || v32->dynstr == NULL) { | |
497 | printk(KERN_ERR "vDSO32: required symbol section not found\n"); | |
498 | return -1; | |
499 | } | |
500 | sect = find_section32(v32->hdr, ".text", NULL); | |
501 | if (sect == NULL) { | |
502 | printk(KERN_ERR "vDSO32: the .text section was not found\n"); | |
503 | return -1; | |
504 | } | |
505 | v32->text = sect - vdso32_kbase; | |
506 | ||
507 | #ifdef CONFIG_PPC64 | |
508 | v64->dynsym = find_section64(v64->hdr, ".dynsym", &v64->dynsymsize); | |
509 | v64->dynstr = find_section64(v64->hdr, ".dynstr", NULL); | |
510 | if (v64->dynsym == NULL || v64->dynstr == NULL) { | |
511 | printk(KERN_ERR "vDSO64: required symbol section not found\n"); | |
512 | return -1; | |
513 | } | |
514 | sect = find_section64(v64->hdr, ".text", NULL); | |
515 | if (sect == NULL) { | |
516 | printk(KERN_ERR "vDSO64: the .text section was not found\n"); | |
517 | return -1; | |
518 | } | |
519 | v64->text = sect - vdso64_kbase; | |
520 | #endif /* CONFIG_PPC64 */ | |
521 | ||
522 | return 0; | |
523 | } | |
524 | ||
525 | static __init void vdso_setup_trampolines(struct lib32_elfinfo *v32, | |
526 | struct lib64_elfinfo *v64) | |
527 | { | |
528 | /* | |
529 | * Find signal trampolines | |
530 | */ | |
531 | ||
532 | #ifdef CONFIG_PPC64 | |
533 | vdso64_rt_sigtramp = find_function64(v64, "__kernel_sigtramp_rt64"); | |
534 | #endif | |
535 | vdso32_sigtramp = find_function32(v32, "__kernel_sigtramp32"); | |
536 | vdso32_rt_sigtramp = find_function32(v32, "__kernel_sigtramp_rt32"); | |
537 | } | |
538 | ||
539 | static __init int vdso_fixup_datapage(struct lib32_elfinfo *v32, | |
540 | struct lib64_elfinfo *v64) | |
541 | { | |
542 | Elf32_Sym *sym32; | |
543 | #ifdef CONFIG_PPC64 | |
544 | Elf64_Sym *sym64; | |
545 | ||
546 | sym64 = find_symbol64(v64, "__kernel_datapage_offset"); | |
547 | if (sym64 == NULL) { | |
548 | printk(KERN_ERR "vDSO64: Can't find symbol " | |
549 | "__kernel_datapage_offset !\n"); | |
550 | return -1; | |
551 | } | |
552 | *((int *)(vdso64_kbase + sym64->st_value - VDSO64_LBASE)) = | |
553 | (vdso64_pages << PAGE_SHIFT) - | |
554 | (sym64->st_value - VDSO64_LBASE); | |
555 | #endif /* CONFIG_PPC64 */ | |
556 | ||
557 | sym32 = find_symbol32(v32, "__kernel_datapage_offset"); | |
558 | if (sym32 == NULL) { | |
559 | printk(KERN_ERR "vDSO32: Can't find symbol " | |
560 | "__kernel_datapage_offset !\n"); | |
561 | return -1; | |
562 | } | |
563 | *((int *)(vdso32_kbase + (sym32->st_value - VDSO32_LBASE))) = | |
564 | (vdso32_pages << PAGE_SHIFT) - | |
565 | (sym32->st_value - VDSO32_LBASE); | |
566 | ||
567 | return 0; | |
568 | } | |
569 | ||
0909c8c2 BH |
570 | |
571 | static __init int vdso_fixup_features(struct lib32_elfinfo *v32, | |
572 | struct lib64_elfinfo *v64) | |
573 | { | |
574 | void *start32; | |
575 | unsigned long size32; | |
576 | ||
577 | #ifdef CONFIG_PPC64 | |
578 | void *start64; | |
579 | unsigned long size64; | |
580 | ||
581 | start64 = find_section64(v64->hdr, "__ftr_fixup", &size64); | |
582 | if (start64) | |
583 | do_feature_fixups(cur_cpu_spec->cpu_features, | |
584 | start64, start64 + size64); | |
585 | ||
7c03d653 BH |
586 | start64 = find_section64(v64->hdr, "__mmu_ftr_fixup", &size64); |
587 | if (start64) | |
588 | do_feature_fixups(cur_cpu_spec->mmu_features, | |
589 | start64, start64 + size64); | |
590 | ||
0909c8c2 BH |
591 | start64 = find_section64(v64->hdr, "__fw_ftr_fixup", &size64); |
592 | if (start64) | |
593 | do_feature_fixups(powerpc_firmware_features, | |
594 | start64, start64 + size64); | |
2d1b2027 KG |
595 | |
596 | start64 = find_section64(v64->hdr, "__lwsync_fixup", &size64); | |
597 | if (start64) | |
598 | do_lwsync_fixups(cur_cpu_spec->cpu_features, | |
599 | start64, start64 + size64); | |
0909c8c2 BH |
600 | #endif /* CONFIG_PPC64 */ |
601 | ||
602 | start32 = find_section32(v32->hdr, "__ftr_fixup", &size32); | |
603 | if (start32) | |
604 | do_feature_fixups(cur_cpu_spec->cpu_features, | |
605 | start32, start32 + size32); | |
606 | ||
7c03d653 BH |
607 | start32 = find_section32(v32->hdr, "__mmu_ftr_fixup", &size32); |
608 | if (start32) | |
609 | do_feature_fixups(cur_cpu_spec->mmu_features, | |
610 | start32, start32 + size32); | |
611 | ||
0909c8c2 BH |
612 | #ifdef CONFIG_PPC64 |
613 | start32 = find_section32(v32->hdr, "__fw_ftr_fixup", &size32); | |
614 | if (start32) | |
615 | do_feature_fixups(powerpc_firmware_features, | |
616 | start32, start32 + size32); | |
617 | #endif /* CONFIG_PPC64 */ | |
618 | ||
2d1b2027 KG |
619 | start32 = find_section32(v32->hdr, "__lwsync_fixup", &size32); |
620 | if (start32) | |
621 | do_lwsync_fixups(cur_cpu_spec->cpu_features, | |
622 | start32, start32 + size32); | |
623 | ||
0909c8c2 BH |
624 | return 0; |
625 | } | |
626 | ||
a7f290da BH |
627 | static __init int vdso_fixup_alt_funcs(struct lib32_elfinfo *v32, |
628 | struct lib64_elfinfo *v64) | |
629 | { | |
630 | int i; | |
631 | ||
632 | for (i = 0; i < ARRAY_SIZE(vdso_patches); i++) { | |
633 | struct vdso_patch_def *patch = &vdso_patches[i]; | |
634 | int match = (cur_cpu_spec->cpu_features & patch->ftr_mask) | |
635 | == patch->ftr_value; | |
636 | if (!match) | |
637 | continue; | |
638 | ||
639 | DBG("replacing %s with %s...\n", patch->gen_name, | |
640 | patch->fix_name ? "NONE" : patch->fix_name); | |
641 | ||
642 | /* | |
643 | * Patch the 32 bits and 64 bits symbols. Note that we do not | |
644 | * patch the "." symbol on 64 bits. | |
645 | * It would be easy to do, but doesn't seem to be necessary, | |
646 | * patching the OPD symbol is enough. | |
647 | */ | |
648 | vdso_do_func_patch32(v32, v64, patch->gen_name, | |
649 | patch->fix_name); | |
650 | #ifdef CONFIG_PPC64 | |
651 | vdso_do_func_patch64(v32, v64, patch->gen_name, | |
652 | patch->fix_name); | |
653 | #endif /* CONFIG_PPC64 */ | |
654 | } | |
655 | ||
656 | return 0; | |
657 | } | |
658 | ||
659 | ||
660 | static __init int vdso_setup(void) | |
661 | { | |
662 | struct lib32_elfinfo v32; | |
663 | struct lib64_elfinfo v64; | |
664 | ||
665 | v32.hdr = vdso32_kbase; | |
666 | #ifdef CONFIG_PPC64 | |
667 | v64.hdr = vdso64_kbase; | |
668 | #endif | |
669 | if (vdso_do_find_sections(&v32, &v64)) | |
670 | return -1; | |
671 | ||
672 | if (vdso_fixup_datapage(&v32, &v64)) | |
673 | return -1; | |
674 | ||
0909c8c2 BH |
675 | if (vdso_fixup_features(&v32, &v64)) |
676 | return -1; | |
677 | ||
a7f290da BH |
678 | if (vdso_fixup_alt_funcs(&v32, &v64)) |
679 | return -1; | |
680 | ||
681 | vdso_setup_trampolines(&v32, &v64); | |
682 | ||
683 | return 0; | |
684 | } | |
685 | ||
686 | /* | |
687 | * Called from setup_arch to initialize the bitmap of available | |
688 | * syscalls in the systemcfg page | |
689 | */ | |
690 | static void __init vdso_setup_syscall_map(void) | |
691 | { | |
692 | unsigned int i; | |
693 | extern unsigned long *sys_call_table; | |
694 | extern unsigned long sys_ni_syscall; | |
695 | ||
696 | ||
697 | for (i = 0; i < __NR_syscalls; i++) { | |
698 | #ifdef CONFIG_PPC64 | |
699 | if (sys_call_table[i*2] != sys_ni_syscall) | |
700 | vdso_data->syscall_map_64[i >> 5] |= | |
701 | 0x80000000UL >> (i & 0x1f); | |
702 | if (sys_call_table[i*2+1] != sys_ni_syscall) | |
703 | vdso_data->syscall_map_32[i >> 5] |= | |
704 | 0x80000000UL >> (i & 0x1f); | |
705 | #else /* CONFIG_PPC64 */ | |
706 | if (sys_call_table[i] != sys_ni_syscall) | |
707 | vdso_data->syscall_map_32[i >> 5] |= | |
708 | 0x80000000UL >> (i & 0x1f); | |
709 | #endif /* CONFIG_PPC64 */ | |
710 | } | |
711 | } | |
712 | ||
18ad51dd | 713 | #ifdef CONFIG_PPC64 |
061d19f2 | 714 | int vdso_getcpu_init(void) |
18ad51dd AB |
715 | { |
716 | unsigned long cpu, node, val; | |
717 | ||
718 | /* | |
719 | * SPRG3 contains the CPU in the bottom 16 bits and the NUMA node in | |
720 | * the next 16 bits. The VDSO uses this to implement getcpu(). | |
721 | */ | |
722 | cpu = get_cpu(); | |
723 | WARN_ON_ONCE(cpu > 0xffff); | |
724 | ||
725 | node = cpu_to_node(cpu); | |
726 | WARN_ON_ONCE(node > 0xffff); | |
727 | ||
728 | val = (cpu & 0xfff) | ((node & 0xffff) << 16); | |
729 | mtspr(SPRN_SPRG3, val); | |
0127262c | 730 | get_paca()->sprg3 = val; |
18ad51dd AB |
731 | |
732 | put_cpu(); | |
733 | ||
734 | return 0; | |
735 | } | |
736 | /* We need to call this before SMP init */ | |
737 | early_initcall(vdso_getcpu_init); | |
738 | #endif | |
a7f290da | 739 | |
7ac9a137 | 740 | static int __init vdso_init(void) |
a7f290da BH |
741 | { |
742 | int i; | |
743 | ||
744 | #ifdef CONFIG_PPC64 | |
745 | /* | |
06fe9fb4 | 746 | * Fill up the "systemcfg" stuff for backward compatibility |
a7f290da | 747 | */ |
31fe5bf6 | 748 | strcpy((char *)vdso_data->eye_catcher, "SYSTEMCFG:PPC64"); |
a7f290da BH |
749 | vdso_data->version.major = SYSTEMCFG_MAJOR; |
750 | vdso_data->version.minor = SYSTEMCFG_MINOR; | |
751 | vdso_data->processor = mfspr(SPRN_PVR); | |
e8222502 | 752 | /* |
1d9a4731 | 753 | * Fake the old platform number for pSeries and add |
e8222502 BH |
754 | * in LPAR bit if necessary |
755 | */ | |
1d9a4731 | 756 | vdso_data->platform = 0x100; |
e8222502 BH |
757 | if (firmware_has_feature(FW_FEATURE_LPAR)) |
758 | vdso_data->platform |= 1; | |
95f72d1e | 759 | vdso_data->physicalMemorySize = memblock_phys_mem_size(); |
a7f290da BH |
760 | vdso_data->dcache_size = ppc64_caches.dsize; |
761 | vdso_data->dcache_line_size = ppc64_caches.dline_size; | |
762 | vdso_data->icache_size = ppc64_caches.isize; | |
763 | vdso_data->icache_line_size = ppc64_caches.iline_size; | |
764 | ||
fbe48175 OJ |
765 | /* XXXOJN: Blocks should be added to ppc64_caches and used instead */ |
766 | vdso_data->dcache_block_size = ppc64_caches.dline_size; | |
767 | vdso_data->icache_block_size = ppc64_caches.iline_size; | |
768 | vdso_data->dcache_log_block_size = ppc64_caches.log_dline_size; | |
769 | vdso_data->icache_log_block_size = ppc64_caches.log_iline_size; | |
770 | ||
a7f290da BH |
771 | /* |
772 | * Calculate the size of the 64 bits vDSO | |
773 | */ | |
774 | vdso64_pages = (&vdso64_end - &vdso64_start) >> PAGE_SHIFT; | |
775 | DBG("vdso64_kbase: %p, 0x%x pages\n", vdso64_kbase, vdso64_pages); | |
fbe48175 OJ |
776 | #else |
777 | vdso_data->dcache_block_size = L1_CACHE_BYTES; | |
778 | vdso_data->dcache_log_block_size = L1_CACHE_SHIFT; | |
779 | vdso_data->icache_block_size = L1_CACHE_BYTES; | |
780 | vdso_data->icache_log_block_size = L1_CACHE_SHIFT; | |
a7f290da BH |
781 | #endif /* CONFIG_PPC64 */ |
782 | ||
783 | ||
784 | /* | |
785 | * Calculate the size of the 32 bits vDSO | |
786 | */ | |
787 | vdso32_pages = (&vdso32_end - &vdso32_start) >> PAGE_SHIFT; | |
788 | DBG("vdso32_kbase: %p, 0x%x pages\n", vdso32_kbase, vdso32_pages); | |
789 | ||
790 | ||
791 | /* | |
792 | * Setup the syscall map in the vDOS | |
793 | */ | |
794 | vdso_setup_syscall_map(); | |
0909c8c2 | 795 | |
a7f290da BH |
796 | /* |
797 | * Initialize the vDSO images in memory, that is do necessary | |
798 | * fixups of vDSO symbols, locate trampolines, etc... | |
799 | */ | |
800 | if (vdso_setup()) { | |
801 | printk(KERN_ERR "vDSO setup failure, not enabled !\n"); | |
802 | vdso32_pages = 0; | |
803 | #ifdef CONFIG_PPC64 | |
804 | vdso64_pages = 0; | |
805 | #endif | |
7ac9a137 | 806 | return 0; |
a7f290da BH |
807 | } |
808 | ||
809 | /* Make sure pages are in the correct state */ | |
7ac9a137 BH |
810 | vdso32_pagelist = kzalloc(sizeof(struct page *) * (vdso32_pages + 2), |
811 | GFP_KERNEL); | |
812 | BUG_ON(vdso32_pagelist == NULL); | |
a7f290da BH |
813 | for (i = 0; i < vdso32_pages; i++) { |
814 | struct page *pg = virt_to_page(vdso32_kbase + i*PAGE_SIZE); | |
815 | ClearPageReserved(pg); | |
816 | get_page(pg); | |
c13e4ca2 | 817 | vdso32_pagelist[i] = pg; |
a7f290da | 818 | } |
c13e4ca2 RM |
819 | vdso32_pagelist[i++] = virt_to_page(vdso_data); |
820 | vdso32_pagelist[i] = NULL; | |
821 | ||
a7f290da | 822 | #ifdef CONFIG_PPC64 |
7ac9a137 BH |
823 | vdso64_pagelist = kzalloc(sizeof(struct page *) * (vdso64_pages + 2), |
824 | GFP_KERNEL); | |
825 | BUG_ON(vdso64_pagelist == NULL); | |
a7f290da BH |
826 | for (i = 0; i < vdso64_pages; i++) { |
827 | struct page *pg = virt_to_page(vdso64_kbase + i*PAGE_SIZE); | |
828 | ClearPageReserved(pg); | |
829 | get_page(pg); | |
c13e4ca2 | 830 | vdso64_pagelist[i] = pg; |
a7f290da | 831 | } |
c13e4ca2 RM |
832 | vdso64_pagelist[i++] = virt_to_page(vdso_data); |
833 | vdso64_pagelist[i] = NULL; | |
a7f290da BH |
834 | #endif /* CONFIG_PPC64 */ |
835 | ||
836 | get_page(virt_to_page(vdso_data)); | |
7ac9a137 BH |
837 | |
838 | smp_wmb(); | |
839 | vdso_ready = 1; | |
840 | ||
841 | return 0; | |
a7f290da | 842 | } |
7ac9a137 | 843 | arch_initcall(vdso_init); |
a7f290da | 844 | |
cae5d390 | 845 | int in_gate_area_no_mm(unsigned long addr) |
a7f290da BH |
846 | { |
847 | return 0; | |
848 | } | |
849 | ||
83b964bb | 850 | int in_gate_area(struct mm_struct *mm, unsigned long addr) |
a7f290da BH |
851 | { |
852 | return 0; | |
853 | } | |
854 | ||
31db58b3 | 855 | struct vm_area_struct *get_gate_vma(struct mm_struct *mm) |
a7f290da BH |
856 | { |
857 | return NULL; | |
858 | } | |
859 |