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/module.h> |
13 | #include <linux/errno.h> | |
14 | #include <linux/sched.h> | |
15 | #include <linux/kernel.h> | |
16 | #include <linux/mm.h> | |
17 | #include <linux/smp.h> | |
a7f290da BH |
18 | #include <linux/stddef.h> |
19 | #include <linux/unistd.h> | |
20 | #include <linux/slab.h> | |
21 | #include <linux/user.h> | |
22 | #include <linux/elf.h> | |
23 | #include <linux/security.h> | |
24 | #include <linux/bootmem.h> | |
95f72d1e | 25 | #include <linux/memblock.h> |
a7f290da BH |
26 | |
27 | #include <asm/pgtable.h> | |
28 | #include <asm/system.h> | |
29 | #include <asm/processor.h> | |
30 | #include <asm/mmu.h> | |
31 | #include <asm/mmu_context.h> | |
d9b2b2a2 | 32 | #include <asm/prom.h> |
a7f290da BH |
33 | #include <asm/machdep.h> |
34 | #include <asm/cputable.h> | |
35 | #include <asm/sections.h> | |
e8222502 | 36 | #include <asm/firmware.h> |
a7f290da BH |
37 | #include <asm/vdso.h> |
38 | #include <asm/vdso_datapage.h> | |
39 | ||
0909c8c2 BH |
40 | #include "setup.h" |
41 | ||
a7f290da BH |
42 | #undef DEBUG |
43 | ||
44 | #ifdef DEBUG | |
45 | #define DBG(fmt...) printk(fmt) | |
46 | #else | |
47 | #define DBG(fmt...) | |
48 | #endif | |
49 | ||
50 | /* Max supported size for symbol names */ | |
51 | #define MAX_SYMNAME 64 | |
52 | ||
348aa303 AS |
53 | /* The alignment of the vDSO */ |
54 | #define VDSO_ALIGNMENT (1 << 16) | |
55 | ||
a7f290da BH |
56 | extern char vdso32_start, vdso32_end; |
57 | static void *vdso32_kbase = &vdso32_start; | |
7ac9a137 BH |
58 | static unsigned int vdso32_pages; |
59 | static struct page **vdso32_pagelist; | |
a7f290da BH |
60 | unsigned long vdso32_sigtramp; |
61 | unsigned long vdso32_rt_sigtramp; | |
62 | ||
63 | #ifdef CONFIG_PPC64 | |
64 | extern char vdso64_start, vdso64_end; | |
65 | static void *vdso64_kbase = &vdso64_start; | |
7ac9a137 BH |
66 | static unsigned int vdso64_pages; |
67 | static struct page **vdso64_pagelist; | |
a7f290da BH |
68 | unsigned long vdso64_rt_sigtramp; |
69 | #endif /* CONFIG_PPC64 */ | |
70 | ||
7ac9a137 BH |
71 | static int vdso_ready; |
72 | ||
a7f290da BH |
73 | /* |
74 | * The vdso data page (aka. systemcfg for old ppc64 fans) is here. | |
75 | * Once the early boot kernel code no longer needs to muck around | |
76 | * with it, it will become dynamically allocated | |
77 | */ | |
78 | static union { | |
79 | struct vdso_data data; | |
80 | u8 page[PAGE_SIZE]; | |
abe1ee3a | 81 | } vdso_data_store __page_aligned_data; |
a7f290da BH |
82 | struct vdso_data *vdso_data = &vdso_data_store.data; |
83 | ||
84 | /* Format of the patch table */ | |
85 | struct vdso_patch_def | |
86 | { | |
87 | unsigned long ftr_mask, ftr_value; | |
88 | const char *gen_name; | |
89 | const char *fix_name; | |
90 | }; | |
91 | ||
92 | /* Table of functions to patch based on the CPU type/revision | |
93 | * | |
94 | * Currently, we only change sync_dicache to do nothing on processors | |
95 | * with a coherent icache | |
96 | */ | |
97 | static struct vdso_patch_def vdso_patches[] = { | |
98 | { | |
99 | CPU_FTR_COHERENT_ICACHE, CPU_FTR_COHERENT_ICACHE, | |
100 | "__kernel_sync_dicache", "__kernel_sync_dicache_p5" | |
101 | }, | |
102 | { | |
103 | CPU_FTR_USE_TB, 0, | |
104 | "__kernel_gettimeofday", NULL | |
105 | }, | |
7b5acbaa BH |
106 | { |
107 | CPU_FTR_USE_TB, 0, | |
108 | "__kernel_clock_gettime", NULL | |
109 | }, | |
110 | { | |
111 | CPU_FTR_USE_TB, 0, | |
112 | "__kernel_clock_getres", NULL | |
113 | }, | |
114 | { | |
115 | CPU_FTR_USE_TB, 0, | |
116 | "__kernel_get_tbfreq", NULL | |
117 | }, | |
a7f290da BH |
118 | }; |
119 | ||
120 | /* | |
121 | * Some infos carried around for each of them during parsing at | |
122 | * boot time. | |
123 | */ | |
124 | struct lib32_elfinfo | |
125 | { | |
126 | Elf32_Ehdr *hdr; /* ptr to ELF */ | |
127 | Elf32_Sym *dynsym; /* ptr to .dynsym section */ | |
128 | unsigned long dynsymsize; /* size of .dynsym section */ | |
129 | char *dynstr; /* ptr to .dynstr section */ | |
130 | unsigned long text; /* offset of .text section in .so */ | |
131 | }; | |
132 | ||
133 | struct lib64_elfinfo | |
134 | { | |
135 | Elf64_Ehdr *hdr; | |
136 | Elf64_Sym *dynsym; | |
137 | unsigned long dynsymsize; | |
138 | char *dynstr; | |
139 | unsigned long text; | |
140 | }; | |
141 | ||
142 | ||
143 | #ifdef __DEBUG | |
144 | static void dump_one_vdso_page(struct page *pg, struct page *upg) | |
145 | { | |
146 | printk("kpg: %p (c:%d,f:%08lx)", __va(page_to_pfn(pg) << PAGE_SHIFT), | |
147 | page_count(pg), | |
148 | pg->flags); | |
89f5b7da | 149 | if (upg && !IS_ERR(upg) /* && pg != upg*/) { |
a7f290da BH |
150 | printk(" upg: %p (c:%d,f:%08lx)", __va(page_to_pfn(upg) |
151 | << PAGE_SHIFT), | |
152 | page_count(upg), | |
153 | upg->flags); | |
154 | } | |
155 | printk("\n"); | |
156 | } | |
157 | ||
158 | static void dump_vdso_pages(struct vm_area_struct * vma) | |
159 | { | |
160 | int i; | |
161 | ||
cab175f9 | 162 | if (!vma || is_32bit_task()) { |
a7f290da BH |
163 | printk("vDSO32 @ %016lx:\n", (unsigned long)vdso32_kbase); |
164 | for (i=0; i<vdso32_pages; i++) { | |
165 | struct page *pg = virt_to_page(vdso32_kbase + | |
166 | i*PAGE_SIZE); | |
167 | struct page *upg = (vma && vma->vm_mm) ? | |
6aab341e | 168 | follow_page(vma, vma->vm_start + i*PAGE_SIZE, 0) |
a7f290da BH |
169 | : NULL; |
170 | dump_one_vdso_page(pg, upg); | |
171 | } | |
172 | } | |
cab175f9 | 173 | if (!vma || !is_32bit_task()) { |
a7f290da BH |
174 | printk("vDSO64 @ %016lx:\n", (unsigned long)vdso64_kbase); |
175 | for (i=0; i<vdso64_pages; i++) { | |
176 | struct page *pg = virt_to_page(vdso64_kbase + | |
177 | i*PAGE_SIZE); | |
178 | struct page *upg = (vma && vma->vm_mm) ? | |
6aab341e | 179 | follow_page(vma, vma->vm_start + i*PAGE_SIZE, 0) |
a7f290da BH |
180 | : NULL; |
181 | dump_one_vdso_page(pg, upg); | |
182 | } | |
183 | } | |
184 | } | |
185 | #endif /* DEBUG */ | |
186 | ||
a7f290da BH |
187 | /* |
188 | * This is called from binfmt_elf, we create the special vma for the | |
189 | * vDSO and insert it into the mm struct tree | |
190 | */ | |
fc5243d9 | 191 | int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp) |
a7f290da BH |
192 | { |
193 | struct mm_struct *mm = current->mm; | |
c13e4ca2 | 194 | struct page **vdso_pagelist; |
a7f290da BH |
195 | unsigned long vdso_pages; |
196 | unsigned long vdso_base; | |
a5bba930 | 197 | int rc; |
a7f290da | 198 | |
7ac9a137 BH |
199 | if (!vdso_ready) |
200 | return 0; | |
201 | ||
a7f290da | 202 | #ifdef CONFIG_PPC64 |
cab175f9 | 203 | if (is_32bit_task()) { |
c13e4ca2 | 204 | vdso_pagelist = vdso32_pagelist; |
a7f290da BH |
205 | vdso_pages = vdso32_pages; |
206 | vdso_base = VDSO32_MBASE; | |
207 | } else { | |
c13e4ca2 | 208 | vdso_pagelist = vdso64_pagelist; |
a7f290da | 209 | vdso_pages = vdso64_pages; |
30d0b368 AB |
210 | /* |
211 | * On 64bit we don't have a preferred map address. This | |
212 | * allows get_unmapped_area to find an area near other mmaps | |
213 | * and most likely share a SLB entry. | |
214 | */ | |
215 | vdso_base = 0; | |
a7f290da BH |
216 | } |
217 | #else | |
c13e4ca2 | 218 | vdso_pagelist = vdso32_pagelist; |
a7f290da BH |
219 | vdso_pages = vdso32_pages; |
220 | vdso_base = VDSO32_MBASE; | |
221 | #endif | |
222 | ||
a5bba930 | 223 | current->mm->context.vdso_base = 0; |
a7f290da BH |
224 | |
225 | /* vDSO has a problem and was disabled, just don't "enable" it for the | |
226 | * process | |
227 | */ | |
228 | if (vdso_pages == 0) | |
229 | return 0; | |
a7f290da BH |
230 | /* Add a page to the vdso size for the data page */ |
231 | vdso_pages ++; | |
232 | ||
233 | /* | |
234 | * pick a base address for the vDSO in process space. We try to put it | |
235 | * at vdso_base which is the "natural" base for it, but we might fail | |
236 | * and end up putting it elsewhere. | |
348aa303 | 237 | * Add enough to the size so that the result can be aligned. |
a7f290da | 238 | */ |
a5bba930 | 239 | down_write(&mm->mmap_sem); |
a7f290da | 240 | vdso_base = get_unmapped_area(NULL, vdso_base, |
348aa303 AS |
241 | (vdso_pages << PAGE_SHIFT) + |
242 | ((VDSO_ALIGNMENT - 1) & PAGE_MASK), | |
243 | 0, 0); | |
a5bba930 BH |
244 | if (IS_ERR_VALUE(vdso_base)) { |
245 | rc = vdso_base; | |
246 | goto fail_mmapsem; | |
a7f290da BH |
247 | } |
248 | ||
348aa303 AS |
249 | /* Add required alignment. */ |
250 | vdso_base = ALIGN(vdso_base, VDSO_ALIGNMENT); | |
251 | ||
f2053f1a AB |
252 | /* |
253 | * Put vDSO base into mm struct. We need to do this before calling | |
254 | * install_special_mapping or the perf counter mmap tracking code | |
255 | * will fail to recognise it as a vDSO (since arch_vma_name fails). | |
256 | */ | |
257 | current->mm->context.vdso_base = vdso_base; | |
258 | ||
a7f290da BH |
259 | /* |
260 | * our vma flags don't have VM_WRITE so by default, the process isn't | |
261 | * allowed to write those pages. | |
262 | * gdb can break that with ptrace interface, and thus trigger COW on | |
263 | * those pages but it's then your responsibility to never do that on | |
264 | * the "data" page of the vDSO or you'll stop getting kernel updates | |
265 | * and your nice userland gettimeofday will be totally dead. | |
266 | * It's fine to use that for setting breakpoints in the vDSO code | |
267 | * pages though | |
c13e4ca2 | 268 | * |
3a0cfadb RM |
269 | * Make sure the vDSO gets into every core dump. |
270 | * Dumping its contents makes post-mortem fully interpretable later | |
271 | * without matching up the same kernel and hardware config to see | |
272 | * what PC values meant. | |
273 | */ | |
c13e4ca2 RM |
274 | rc = install_special_mapping(mm, vdso_base, vdso_pages << PAGE_SHIFT, |
275 | VM_READ|VM_EXEC| | |
276 | VM_MAYREAD|VM_MAYWRITE|VM_MAYEXEC| | |
277 | VM_ALWAYSDUMP, | |
278 | vdso_pagelist); | |
f2053f1a AB |
279 | if (rc) { |
280 | current->mm->context.vdso_base = 0; | |
c13e4ca2 | 281 | goto fail_mmapsem; |
f2053f1a | 282 | } |
c13e4ca2 | 283 | |
a7f290da | 284 | up_write(&mm->mmap_sem); |
a7f290da | 285 | return 0; |
a5bba930 | 286 | |
a5bba930 BH |
287 | fail_mmapsem: |
288 | up_write(&mm->mmap_sem); | |
289 | return rc; | |
290 | } | |
291 | ||
292 | const char *arch_vma_name(struct vm_area_struct *vma) | |
293 | { | |
294 | if (vma->vm_mm && vma->vm_start == vma->vm_mm->context.vdso_base) | |
295 | return "[vdso]"; | |
296 | return NULL; | |
a7f290da BH |
297 | } |
298 | ||
a5bba930 BH |
299 | |
300 | ||
a7f290da BH |
301 | static void * __init find_section32(Elf32_Ehdr *ehdr, const char *secname, |
302 | unsigned long *size) | |
303 | { | |
304 | Elf32_Shdr *sechdrs; | |
305 | unsigned int i; | |
306 | char *secnames; | |
307 | ||
308 | /* Grab section headers and strings so we can tell who is who */ | |
309 | sechdrs = (void *)ehdr + ehdr->e_shoff; | |
310 | secnames = (void *)ehdr + sechdrs[ehdr->e_shstrndx].sh_offset; | |
311 | ||
312 | /* Find the section they want */ | |
313 | for (i = 1; i < ehdr->e_shnum; i++) { | |
314 | if (strcmp(secnames+sechdrs[i].sh_name, secname) == 0) { | |
315 | if (size) | |
316 | *size = sechdrs[i].sh_size; | |
317 | return (void *)ehdr + sechdrs[i].sh_offset; | |
318 | } | |
319 | } | |
320 | *size = 0; | |
321 | return NULL; | |
322 | } | |
323 | ||
324 | static Elf32_Sym * __init find_symbol32(struct lib32_elfinfo *lib, | |
325 | const char *symname) | |
326 | { | |
327 | unsigned int i; | |
328 | char name[MAX_SYMNAME], *c; | |
329 | ||
330 | for (i = 0; i < (lib->dynsymsize / sizeof(Elf32_Sym)); i++) { | |
331 | if (lib->dynsym[i].st_name == 0) | |
332 | continue; | |
333 | strlcpy(name, lib->dynstr + lib->dynsym[i].st_name, | |
334 | MAX_SYMNAME); | |
335 | c = strchr(name, '@'); | |
336 | if (c) | |
337 | *c = 0; | |
338 | if (strcmp(symname, name) == 0) | |
339 | return &lib->dynsym[i]; | |
340 | } | |
341 | return NULL; | |
342 | } | |
343 | ||
344 | /* Note that we assume the section is .text and the symbol is relative to | |
345 | * the library base | |
346 | */ | |
347 | static unsigned long __init find_function32(struct lib32_elfinfo *lib, | |
348 | const char *symname) | |
349 | { | |
350 | Elf32_Sym *sym = find_symbol32(lib, symname); | |
351 | ||
352 | if (sym == NULL) { | |
353 | printk(KERN_WARNING "vDSO32: function %s not found !\n", | |
354 | symname); | |
355 | return 0; | |
356 | } | |
357 | return sym->st_value - VDSO32_LBASE; | |
358 | } | |
359 | ||
cf8918fe AB |
360 | static int __init vdso_do_func_patch32(struct lib32_elfinfo *v32, |
361 | struct lib64_elfinfo *v64, | |
362 | const char *orig, const char *fix) | |
a7f290da BH |
363 | { |
364 | Elf32_Sym *sym32_gen, *sym32_fix; | |
365 | ||
366 | sym32_gen = find_symbol32(v32, orig); | |
367 | if (sym32_gen == NULL) { | |
368 | printk(KERN_ERR "vDSO32: Can't find symbol %s !\n", orig); | |
369 | return -1; | |
370 | } | |
371 | if (fix == NULL) { | |
372 | sym32_gen->st_name = 0; | |
373 | return 0; | |
374 | } | |
375 | sym32_fix = find_symbol32(v32, fix); | |
376 | if (sym32_fix == NULL) { | |
377 | printk(KERN_ERR "vDSO32: Can't find symbol %s !\n", fix); | |
378 | return -1; | |
379 | } | |
380 | sym32_gen->st_value = sym32_fix->st_value; | |
381 | sym32_gen->st_size = sym32_fix->st_size; | |
382 | sym32_gen->st_info = sym32_fix->st_info; | |
383 | sym32_gen->st_other = sym32_fix->st_other; | |
384 | sym32_gen->st_shndx = sym32_fix->st_shndx; | |
385 | ||
386 | return 0; | |
387 | } | |
388 | ||
389 | ||
390 | #ifdef CONFIG_PPC64 | |
391 | ||
392 | static void * __init find_section64(Elf64_Ehdr *ehdr, const char *secname, | |
393 | unsigned long *size) | |
394 | { | |
395 | Elf64_Shdr *sechdrs; | |
396 | unsigned int i; | |
397 | char *secnames; | |
398 | ||
399 | /* Grab section headers and strings so we can tell who is who */ | |
400 | sechdrs = (void *)ehdr + ehdr->e_shoff; | |
401 | secnames = (void *)ehdr + sechdrs[ehdr->e_shstrndx].sh_offset; | |
402 | ||
403 | /* Find the section they want */ | |
404 | for (i = 1; i < ehdr->e_shnum; i++) { | |
405 | if (strcmp(secnames+sechdrs[i].sh_name, secname) == 0) { | |
406 | if (size) | |
407 | *size = sechdrs[i].sh_size; | |
408 | return (void *)ehdr + sechdrs[i].sh_offset; | |
409 | } | |
410 | } | |
411 | if (size) | |
412 | *size = 0; | |
413 | return NULL; | |
414 | } | |
415 | ||
416 | static Elf64_Sym * __init find_symbol64(struct lib64_elfinfo *lib, | |
417 | const char *symname) | |
418 | { | |
419 | unsigned int i; | |
420 | char name[MAX_SYMNAME], *c; | |
421 | ||
422 | for (i = 0; i < (lib->dynsymsize / sizeof(Elf64_Sym)); i++) { | |
423 | if (lib->dynsym[i].st_name == 0) | |
424 | continue; | |
425 | strlcpy(name, lib->dynstr + lib->dynsym[i].st_name, | |
426 | MAX_SYMNAME); | |
427 | c = strchr(name, '@'); | |
428 | if (c) | |
429 | *c = 0; | |
430 | if (strcmp(symname, name) == 0) | |
431 | return &lib->dynsym[i]; | |
432 | } | |
433 | return NULL; | |
434 | } | |
435 | ||
436 | /* Note that we assume the section is .text and the symbol is relative to | |
437 | * the library base | |
438 | */ | |
439 | static unsigned long __init find_function64(struct lib64_elfinfo *lib, | |
440 | const char *symname) | |
441 | { | |
442 | Elf64_Sym *sym = find_symbol64(lib, symname); | |
443 | ||
444 | if (sym == NULL) { | |
445 | printk(KERN_WARNING "vDSO64: function %s not found !\n", | |
446 | symname); | |
447 | return 0; | |
448 | } | |
449 | #ifdef VDS64_HAS_DESCRIPTORS | |
450 | return *((u64 *)(vdso64_kbase + sym->st_value - VDSO64_LBASE)) - | |
451 | VDSO64_LBASE; | |
452 | #else | |
453 | return sym->st_value - VDSO64_LBASE; | |
454 | #endif | |
455 | } | |
456 | ||
cf8918fe AB |
457 | static int __init vdso_do_func_patch64(struct lib32_elfinfo *v32, |
458 | struct lib64_elfinfo *v64, | |
459 | const char *orig, const char *fix) | |
a7f290da BH |
460 | { |
461 | Elf64_Sym *sym64_gen, *sym64_fix; | |
462 | ||
463 | sym64_gen = find_symbol64(v64, orig); | |
464 | if (sym64_gen == NULL) { | |
465 | printk(KERN_ERR "vDSO64: Can't find symbol %s !\n", orig); | |
466 | return -1; | |
467 | } | |
468 | if (fix == NULL) { | |
469 | sym64_gen->st_name = 0; | |
470 | return 0; | |
471 | } | |
472 | sym64_fix = find_symbol64(v64, fix); | |
473 | if (sym64_fix == NULL) { | |
474 | printk(KERN_ERR "vDSO64: Can't find symbol %s !\n", fix); | |
475 | return -1; | |
476 | } | |
477 | sym64_gen->st_value = sym64_fix->st_value; | |
478 | sym64_gen->st_size = sym64_fix->st_size; | |
479 | sym64_gen->st_info = sym64_fix->st_info; | |
480 | sym64_gen->st_other = sym64_fix->st_other; | |
481 | sym64_gen->st_shndx = sym64_fix->st_shndx; | |
482 | ||
483 | return 0; | |
484 | } | |
485 | ||
486 | #endif /* CONFIG_PPC64 */ | |
487 | ||
488 | ||
489 | static __init int vdso_do_find_sections(struct lib32_elfinfo *v32, | |
490 | struct lib64_elfinfo *v64) | |
491 | { | |
492 | void *sect; | |
493 | ||
494 | /* | |
495 | * Locate symbol tables & text section | |
496 | */ | |
497 | ||
498 | v32->dynsym = find_section32(v32->hdr, ".dynsym", &v32->dynsymsize); | |
499 | v32->dynstr = find_section32(v32->hdr, ".dynstr", NULL); | |
500 | if (v32->dynsym == NULL || v32->dynstr == NULL) { | |
501 | printk(KERN_ERR "vDSO32: required symbol section not found\n"); | |
502 | return -1; | |
503 | } | |
504 | sect = find_section32(v32->hdr, ".text", NULL); | |
505 | if (sect == NULL) { | |
506 | printk(KERN_ERR "vDSO32: the .text section was not found\n"); | |
507 | return -1; | |
508 | } | |
509 | v32->text = sect - vdso32_kbase; | |
510 | ||
511 | #ifdef CONFIG_PPC64 | |
512 | v64->dynsym = find_section64(v64->hdr, ".dynsym", &v64->dynsymsize); | |
513 | v64->dynstr = find_section64(v64->hdr, ".dynstr", NULL); | |
514 | if (v64->dynsym == NULL || v64->dynstr == NULL) { | |
515 | printk(KERN_ERR "vDSO64: required symbol section not found\n"); | |
516 | return -1; | |
517 | } | |
518 | sect = find_section64(v64->hdr, ".text", NULL); | |
519 | if (sect == NULL) { | |
520 | printk(KERN_ERR "vDSO64: the .text section was not found\n"); | |
521 | return -1; | |
522 | } | |
523 | v64->text = sect - vdso64_kbase; | |
524 | #endif /* CONFIG_PPC64 */ | |
525 | ||
526 | return 0; | |
527 | } | |
528 | ||
529 | static __init void vdso_setup_trampolines(struct lib32_elfinfo *v32, | |
530 | struct lib64_elfinfo *v64) | |
531 | { | |
532 | /* | |
533 | * Find signal trampolines | |
534 | */ | |
535 | ||
536 | #ifdef CONFIG_PPC64 | |
537 | vdso64_rt_sigtramp = find_function64(v64, "__kernel_sigtramp_rt64"); | |
538 | #endif | |
539 | vdso32_sigtramp = find_function32(v32, "__kernel_sigtramp32"); | |
540 | vdso32_rt_sigtramp = find_function32(v32, "__kernel_sigtramp_rt32"); | |
541 | } | |
542 | ||
543 | static __init int vdso_fixup_datapage(struct lib32_elfinfo *v32, | |
544 | struct lib64_elfinfo *v64) | |
545 | { | |
546 | Elf32_Sym *sym32; | |
547 | #ifdef CONFIG_PPC64 | |
548 | Elf64_Sym *sym64; | |
549 | ||
550 | sym64 = find_symbol64(v64, "__kernel_datapage_offset"); | |
551 | if (sym64 == NULL) { | |
552 | printk(KERN_ERR "vDSO64: Can't find symbol " | |
553 | "__kernel_datapage_offset !\n"); | |
554 | return -1; | |
555 | } | |
556 | *((int *)(vdso64_kbase + sym64->st_value - VDSO64_LBASE)) = | |
557 | (vdso64_pages << PAGE_SHIFT) - | |
558 | (sym64->st_value - VDSO64_LBASE); | |
559 | #endif /* CONFIG_PPC64 */ | |
560 | ||
561 | sym32 = find_symbol32(v32, "__kernel_datapage_offset"); | |
562 | if (sym32 == NULL) { | |
563 | printk(KERN_ERR "vDSO32: Can't find symbol " | |
564 | "__kernel_datapage_offset !\n"); | |
565 | return -1; | |
566 | } | |
567 | *((int *)(vdso32_kbase + (sym32->st_value - VDSO32_LBASE))) = | |
568 | (vdso32_pages << PAGE_SHIFT) - | |
569 | (sym32->st_value - VDSO32_LBASE); | |
570 | ||
571 | return 0; | |
572 | } | |
573 | ||
0909c8c2 BH |
574 | |
575 | static __init int vdso_fixup_features(struct lib32_elfinfo *v32, | |
576 | struct lib64_elfinfo *v64) | |
577 | { | |
578 | void *start32; | |
579 | unsigned long size32; | |
580 | ||
581 | #ifdef CONFIG_PPC64 | |
582 | void *start64; | |
583 | unsigned long size64; | |
584 | ||
585 | start64 = find_section64(v64->hdr, "__ftr_fixup", &size64); | |
586 | if (start64) | |
587 | do_feature_fixups(cur_cpu_spec->cpu_features, | |
588 | start64, start64 + size64); | |
589 | ||
7c03d653 BH |
590 | start64 = find_section64(v64->hdr, "__mmu_ftr_fixup", &size64); |
591 | if (start64) | |
592 | do_feature_fixups(cur_cpu_spec->mmu_features, | |
593 | start64, start64 + size64); | |
594 | ||
0909c8c2 BH |
595 | start64 = find_section64(v64->hdr, "__fw_ftr_fixup", &size64); |
596 | if (start64) | |
597 | do_feature_fixups(powerpc_firmware_features, | |
598 | start64, start64 + size64); | |
2d1b2027 KG |
599 | |
600 | start64 = find_section64(v64->hdr, "__lwsync_fixup", &size64); | |
601 | if (start64) | |
602 | do_lwsync_fixups(cur_cpu_spec->cpu_features, | |
603 | start64, start64 + size64); | |
0909c8c2 BH |
604 | #endif /* CONFIG_PPC64 */ |
605 | ||
606 | start32 = find_section32(v32->hdr, "__ftr_fixup", &size32); | |
607 | if (start32) | |
608 | do_feature_fixups(cur_cpu_spec->cpu_features, | |
609 | start32, start32 + size32); | |
610 | ||
7c03d653 BH |
611 | start32 = find_section32(v32->hdr, "__mmu_ftr_fixup", &size32); |
612 | if (start32) | |
613 | do_feature_fixups(cur_cpu_spec->mmu_features, | |
614 | start32, start32 + size32); | |
615 | ||
0909c8c2 BH |
616 | #ifdef CONFIG_PPC64 |
617 | start32 = find_section32(v32->hdr, "__fw_ftr_fixup", &size32); | |
618 | if (start32) | |
619 | do_feature_fixups(powerpc_firmware_features, | |
620 | start32, start32 + size32); | |
621 | #endif /* CONFIG_PPC64 */ | |
622 | ||
2d1b2027 KG |
623 | start32 = find_section32(v32->hdr, "__lwsync_fixup", &size32); |
624 | if (start32) | |
625 | do_lwsync_fixups(cur_cpu_spec->cpu_features, | |
626 | start32, start32 + size32); | |
627 | ||
0909c8c2 BH |
628 | return 0; |
629 | } | |
630 | ||
a7f290da BH |
631 | static __init int vdso_fixup_alt_funcs(struct lib32_elfinfo *v32, |
632 | struct lib64_elfinfo *v64) | |
633 | { | |
634 | int i; | |
635 | ||
636 | for (i = 0; i < ARRAY_SIZE(vdso_patches); i++) { | |
637 | struct vdso_patch_def *patch = &vdso_patches[i]; | |
638 | int match = (cur_cpu_spec->cpu_features & patch->ftr_mask) | |
639 | == patch->ftr_value; | |
640 | if (!match) | |
641 | continue; | |
642 | ||
643 | DBG("replacing %s with %s...\n", patch->gen_name, | |
644 | patch->fix_name ? "NONE" : patch->fix_name); | |
645 | ||
646 | /* | |
647 | * Patch the 32 bits and 64 bits symbols. Note that we do not | |
648 | * patch the "." symbol on 64 bits. | |
649 | * It would be easy to do, but doesn't seem to be necessary, | |
650 | * patching the OPD symbol is enough. | |
651 | */ | |
652 | vdso_do_func_patch32(v32, v64, patch->gen_name, | |
653 | patch->fix_name); | |
654 | #ifdef CONFIG_PPC64 | |
655 | vdso_do_func_patch64(v32, v64, patch->gen_name, | |
656 | patch->fix_name); | |
657 | #endif /* CONFIG_PPC64 */ | |
658 | } | |
659 | ||
660 | return 0; | |
661 | } | |
662 | ||
663 | ||
664 | static __init int vdso_setup(void) | |
665 | { | |
666 | struct lib32_elfinfo v32; | |
667 | struct lib64_elfinfo v64; | |
668 | ||
669 | v32.hdr = vdso32_kbase; | |
670 | #ifdef CONFIG_PPC64 | |
671 | v64.hdr = vdso64_kbase; | |
672 | #endif | |
673 | if (vdso_do_find_sections(&v32, &v64)) | |
674 | return -1; | |
675 | ||
676 | if (vdso_fixup_datapage(&v32, &v64)) | |
677 | return -1; | |
678 | ||
0909c8c2 BH |
679 | if (vdso_fixup_features(&v32, &v64)) |
680 | return -1; | |
681 | ||
a7f290da BH |
682 | if (vdso_fixup_alt_funcs(&v32, &v64)) |
683 | return -1; | |
684 | ||
685 | vdso_setup_trampolines(&v32, &v64); | |
686 | ||
687 | return 0; | |
688 | } | |
689 | ||
690 | /* | |
691 | * Called from setup_arch to initialize the bitmap of available | |
692 | * syscalls in the systemcfg page | |
693 | */ | |
694 | static void __init vdso_setup_syscall_map(void) | |
695 | { | |
696 | unsigned int i; | |
697 | extern unsigned long *sys_call_table; | |
698 | extern unsigned long sys_ni_syscall; | |
699 | ||
700 | ||
701 | for (i = 0; i < __NR_syscalls; i++) { | |
702 | #ifdef CONFIG_PPC64 | |
703 | if (sys_call_table[i*2] != sys_ni_syscall) | |
704 | vdso_data->syscall_map_64[i >> 5] |= | |
705 | 0x80000000UL >> (i & 0x1f); | |
706 | if (sys_call_table[i*2+1] != sys_ni_syscall) | |
707 | vdso_data->syscall_map_32[i >> 5] |= | |
708 | 0x80000000UL >> (i & 0x1f); | |
709 | #else /* CONFIG_PPC64 */ | |
710 | if (sys_call_table[i] != sys_ni_syscall) | |
711 | vdso_data->syscall_map_32[i >> 5] |= | |
712 | 0x80000000UL >> (i & 0x1f); | |
713 | #endif /* CONFIG_PPC64 */ | |
714 | } | |
715 | } | |
716 | ||
717 | ||
7ac9a137 | 718 | static int __init vdso_init(void) |
a7f290da BH |
719 | { |
720 | int i; | |
721 | ||
722 | #ifdef CONFIG_PPC64 | |
723 | /* | |
06fe9fb4 | 724 | * Fill up the "systemcfg" stuff for backward compatibility |
a7f290da | 725 | */ |
31fe5bf6 | 726 | strcpy((char *)vdso_data->eye_catcher, "SYSTEMCFG:PPC64"); |
a7f290da BH |
727 | vdso_data->version.major = SYSTEMCFG_MAJOR; |
728 | vdso_data->version.minor = SYSTEMCFG_MINOR; | |
729 | vdso_data->processor = mfspr(SPRN_PVR); | |
e8222502 BH |
730 | /* |
731 | * Fake the old platform number for pSeries and iSeries and add | |
732 | * in LPAR bit if necessary | |
733 | */ | |
734 | vdso_data->platform = machine_is(iseries) ? 0x200 : 0x100; | |
735 | if (firmware_has_feature(FW_FEATURE_LPAR)) | |
736 | vdso_data->platform |= 1; | |
95f72d1e | 737 | vdso_data->physicalMemorySize = memblock_phys_mem_size(); |
a7f290da BH |
738 | vdso_data->dcache_size = ppc64_caches.dsize; |
739 | vdso_data->dcache_line_size = ppc64_caches.dline_size; | |
740 | vdso_data->icache_size = ppc64_caches.isize; | |
741 | vdso_data->icache_line_size = ppc64_caches.iline_size; | |
742 | ||
fbe48175 OJ |
743 | /* XXXOJN: Blocks should be added to ppc64_caches and used instead */ |
744 | vdso_data->dcache_block_size = ppc64_caches.dline_size; | |
745 | vdso_data->icache_block_size = ppc64_caches.iline_size; | |
746 | vdso_data->dcache_log_block_size = ppc64_caches.log_dline_size; | |
747 | vdso_data->icache_log_block_size = ppc64_caches.log_iline_size; | |
748 | ||
a7f290da BH |
749 | /* |
750 | * Calculate the size of the 64 bits vDSO | |
751 | */ | |
752 | vdso64_pages = (&vdso64_end - &vdso64_start) >> PAGE_SHIFT; | |
753 | DBG("vdso64_kbase: %p, 0x%x pages\n", vdso64_kbase, vdso64_pages); | |
fbe48175 OJ |
754 | #else |
755 | vdso_data->dcache_block_size = L1_CACHE_BYTES; | |
756 | vdso_data->dcache_log_block_size = L1_CACHE_SHIFT; | |
757 | vdso_data->icache_block_size = L1_CACHE_BYTES; | |
758 | vdso_data->icache_log_block_size = L1_CACHE_SHIFT; | |
a7f290da BH |
759 | #endif /* CONFIG_PPC64 */ |
760 | ||
761 | ||
762 | /* | |
763 | * Calculate the size of the 32 bits vDSO | |
764 | */ | |
765 | vdso32_pages = (&vdso32_end - &vdso32_start) >> PAGE_SHIFT; | |
766 | DBG("vdso32_kbase: %p, 0x%x pages\n", vdso32_kbase, vdso32_pages); | |
767 | ||
768 | ||
769 | /* | |
770 | * Setup the syscall map in the vDOS | |
771 | */ | |
772 | vdso_setup_syscall_map(); | |
0909c8c2 | 773 | |
a7f290da BH |
774 | /* |
775 | * Initialize the vDSO images in memory, that is do necessary | |
776 | * fixups of vDSO symbols, locate trampolines, etc... | |
777 | */ | |
778 | if (vdso_setup()) { | |
779 | printk(KERN_ERR "vDSO setup failure, not enabled !\n"); | |
780 | vdso32_pages = 0; | |
781 | #ifdef CONFIG_PPC64 | |
782 | vdso64_pages = 0; | |
783 | #endif | |
7ac9a137 | 784 | return 0; |
a7f290da BH |
785 | } |
786 | ||
787 | /* Make sure pages are in the correct state */ | |
7ac9a137 BH |
788 | vdso32_pagelist = kzalloc(sizeof(struct page *) * (vdso32_pages + 2), |
789 | GFP_KERNEL); | |
790 | BUG_ON(vdso32_pagelist == NULL); | |
a7f290da BH |
791 | for (i = 0; i < vdso32_pages; i++) { |
792 | struct page *pg = virt_to_page(vdso32_kbase + i*PAGE_SIZE); | |
793 | ClearPageReserved(pg); | |
794 | get_page(pg); | |
c13e4ca2 | 795 | vdso32_pagelist[i] = pg; |
a7f290da | 796 | } |
c13e4ca2 RM |
797 | vdso32_pagelist[i++] = virt_to_page(vdso_data); |
798 | vdso32_pagelist[i] = NULL; | |
799 | ||
a7f290da | 800 | #ifdef CONFIG_PPC64 |
7ac9a137 BH |
801 | vdso64_pagelist = kzalloc(sizeof(struct page *) * (vdso64_pages + 2), |
802 | GFP_KERNEL); | |
803 | BUG_ON(vdso64_pagelist == NULL); | |
a7f290da BH |
804 | for (i = 0; i < vdso64_pages; i++) { |
805 | struct page *pg = virt_to_page(vdso64_kbase + i*PAGE_SIZE); | |
806 | ClearPageReserved(pg); | |
807 | get_page(pg); | |
c13e4ca2 | 808 | vdso64_pagelist[i] = pg; |
a7f290da | 809 | } |
c13e4ca2 RM |
810 | vdso64_pagelist[i++] = virt_to_page(vdso_data); |
811 | vdso64_pagelist[i] = NULL; | |
a7f290da BH |
812 | #endif /* CONFIG_PPC64 */ |
813 | ||
814 | get_page(virt_to_page(vdso_data)); | |
7ac9a137 BH |
815 | |
816 | smp_wmb(); | |
817 | vdso_ready = 1; | |
818 | ||
819 | return 0; | |
a7f290da | 820 | } |
7ac9a137 | 821 | arch_initcall(vdso_init); |
a7f290da BH |
822 | |
823 | int in_gate_area_no_task(unsigned long addr) | |
824 | { | |
825 | return 0; | |
826 | } | |
827 | ||
828 | int in_gate_area(struct task_struct *task, unsigned long addr) | |
829 | { | |
830 | return 0; | |
831 | } | |
832 | ||
833 | struct vm_area_struct *get_gate_vma(struct task_struct *tsk) | |
834 | { | |
835 | return NULL; | |
836 | } | |
837 |