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> | |
d9b2b2a2 | 25 | #include <linux/lmb.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 | ||
53 | extern char vdso32_start, vdso32_end; | |
54 | static void *vdso32_kbase = &vdso32_start; | |
7ac9a137 BH |
55 | static unsigned int vdso32_pages; |
56 | static struct page **vdso32_pagelist; | |
a7f290da BH |
57 | unsigned long vdso32_sigtramp; |
58 | unsigned long vdso32_rt_sigtramp; | |
59 | ||
60 | #ifdef CONFIG_PPC64 | |
61 | extern char vdso64_start, vdso64_end; | |
62 | static void *vdso64_kbase = &vdso64_start; | |
7ac9a137 BH |
63 | static unsigned int vdso64_pages; |
64 | static struct page **vdso64_pagelist; | |
a7f290da BH |
65 | unsigned long vdso64_rt_sigtramp; |
66 | #endif /* CONFIG_PPC64 */ | |
67 | ||
7ac9a137 BH |
68 | static int vdso_ready; |
69 | ||
a7f290da BH |
70 | /* |
71 | * The vdso data page (aka. systemcfg for old ppc64 fans) is here. | |
72 | * Once the early boot kernel code no longer needs to muck around | |
73 | * with it, it will become dynamically allocated | |
74 | */ | |
75 | static union { | |
76 | struct vdso_data data; | |
77 | u8 page[PAGE_SIZE]; | |
abe1ee3a | 78 | } vdso_data_store __page_aligned_data; |
a7f290da BH |
79 | struct vdso_data *vdso_data = &vdso_data_store.data; |
80 | ||
81 | /* Format of the patch table */ | |
82 | struct vdso_patch_def | |
83 | { | |
84 | unsigned long ftr_mask, ftr_value; | |
85 | const char *gen_name; | |
86 | const char *fix_name; | |
87 | }; | |
88 | ||
89 | /* Table of functions to patch based on the CPU type/revision | |
90 | * | |
91 | * Currently, we only change sync_dicache to do nothing on processors | |
92 | * with a coherent icache | |
93 | */ | |
94 | static struct vdso_patch_def vdso_patches[] = { | |
95 | { | |
96 | CPU_FTR_COHERENT_ICACHE, CPU_FTR_COHERENT_ICACHE, | |
97 | "__kernel_sync_dicache", "__kernel_sync_dicache_p5" | |
98 | }, | |
99 | { | |
100 | CPU_FTR_USE_TB, 0, | |
101 | "__kernel_gettimeofday", NULL | |
102 | }, | |
7b5acbaa BH |
103 | { |
104 | CPU_FTR_USE_TB, 0, | |
105 | "__kernel_clock_gettime", NULL | |
106 | }, | |
107 | { | |
108 | CPU_FTR_USE_TB, 0, | |
109 | "__kernel_clock_getres", NULL | |
110 | }, | |
111 | { | |
112 | CPU_FTR_USE_TB, 0, | |
113 | "__kernel_get_tbfreq", NULL | |
114 | }, | |
a7f290da BH |
115 | }; |
116 | ||
117 | /* | |
118 | * Some infos carried around for each of them during parsing at | |
119 | * boot time. | |
120 | */ | |
121 | struct lib32_elfinfo | |
122 | { | |
123 | Elf32_Ehdr *hdr; /* ptr to ELF */ | |
124 | Elf32_Sym *dynsym; /* ptr to .dynsym section */ | |
125 | unsigned long dynsymsize; /* size of .dynsym section */ | |
126 | char *dynstr; /* ptr to .dynstr section */ | |
127 | unsigned long text; /* offset of .text section in .so */ | |
128 | }; | |
129 | ||
130 | struct lib64_elfinfo | |
131 | { | |
132 | Elf64_Ehdr *hdr; | |
133 | Elf64_Sym *dynsym; | |
134 | unsigned long dynsymsize; | |
135 | char *dynstr; | |
136 | unsigned long text; | |
137 | }; | |
138 | ||
139 | ||
140 | #ifdef __DEBUG | |
141 | static void dump_one_vdso_page(struct page *pg, struct page *upg) | |
142 | { | |
143 | printk("kpg: %p (c:%d,f:%08lx)", __va(page_to_pfn(pg) << PAGE_SHIFT), | |
144 | page_count(pg), | |
145 | pg->flags); | |
89f5b7da | 146 | if (upg && !IS_ERR(upg) /* && pg != upg*/) { |
a7f290da BH |
147 | printk(" upg: %p (c:%d,f:%08lx)", __va(page_to_pfn(upg) |
148 | << PAGE_SHIFT), | |
149 | page_count(upg), | |
150 | upg->flags); | |
151 | } | |
152 | printk("\n"); | |
153 | } | |
154 | ||
155 | static void dump_vdso_pages(struct vm_area_struct * vma) | |
156 | { | |
157 | int i; | |
158 | ||
159 | if (!vma || test_thread_flag(TIF_32BIT)) { | |
160 | printk("vDSO32 @ %016lx:\n", (unsigned long)vdso32_kbase); | |
161 | for (i=0; i<vdso32_pages; i++) { | |
162 | struct page *pg = virt_to_page(vdso32_kbase + | |
163 | i*PAGE_SIZE); | |
164 | struct page *upg = (vma && vma->vm_mm) ? | |
6aab341e | 165 | follow_page(vma, vma->vm_start + i*PAGE_SIZE, 0) |
a7f290da BH |
166 | : NULL; |
167 | dump_one_vdso_page(pg, upg); | |
168 | } | |
169 | } | |
170 | if (!vma || !test_thread_flag(TIF_32BIT)) { | |
171 | printk("vDSO64 @ %016lx:\n", (unsigned long)vdso64_kbase); | |
172 | for (i=0; i<vdso64_pages; i++) { | |
173 | struct page *pg = virt_to_page(vdso64_kbase + | |
174 | i*PAGE_SIZE); | |
175 | struct page *upg = (vma && vma->vm_mm) ? | |
6aab341e | 176 | follow_page(vma, vma->vm_start + i*PAGE_SIZE, 0) |
a7f290da BH |
177 | : NULL; |
178 | dump_one_vdso_page(pg, upg); | |
179 | } | |
180 | } | |
181 | } | |
182 | #endif /* DEBUG */ | |
183 | ||
a7f290da BH |
184 | /* |
185 | * This is called from binfmt_elf, we create the special vma for the | |
186 | * vDSO and insert it into the mm struct tree | |
187 | */ | |
fc5243d9 | 188 | int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp) |
a7f290da BH |
189 | { |
190 | struct mm_struct *mm = current->mm; | |
c13e4ca2 | 191 | struct page **vdso_pagelist; |
a7f290da BH |
192 | unsigned long vdso_pages; |
193 | unsigned long vdso_base; | |
a5bba930 | 194 | int rc; |
a7f290da | 195 | |
7ac9a137 BH |
196 | if (!vdso_ready) |
197 | return 0; | |
198 | ||
a7f290da BH |
199 | #ifdef CONFIG_PPC64 |
200 | if (test_thread_flag(TIF_32BIT)) { | |
c13e4ca2 | 201 | vdso_pagelist = vdso32_pagelist; |
a7f290da BH |
202 | vdso_pages = vdso32_pages; |
203 | vdso_base = VDSO32_MBASE; | |
204 | } else { | |
c13e4ca2 | 205 | vdso_pagelist = vdso64_pagelist; |
a7f290da | 206 | vdso_pages = vdso64_pages; |
30d0b368 AB |
207 | /* |
208 | * On 64bit we don't have a preferred map address. This | |
209 | * allows get_unmapped_area to find an area near other mmaps | |
210 | * and most likely share a SLB entry. | |
211 | */ | |
212 | vdso_base = 0; | |
a7f290da BH |
213 | } |
214 | #else | |
c13e4ca2 | 215 | vdso_pagelist = vdso32_pagelist; |
a7f290da BH |
216 | vdso_pages = vdso32_pages; |
217 | vdso_base = VDSO32_MBASE; | |
218 | #endif | |
219 | ||
a5bba930 | 220 | current->mm->context.vdso_base = 0; |
a7f290da BH |
221 | |
222 | /* vDSO has a problem and was disabled, just don't "enable" it for the | |
223 | * process | |
224 | */ | |
225 | if (vdso_pages == 0) | |
226 | return 0; | |
a7f290da BH |
227 | /* Add a page to the vdso size for the data page */ |
228 | vdso_pages ++; | |
229 | ||
230 | /* | |
231 | * pick a base address for the vDSO in process space. We try to put it | |
232 | * at vdso_base which is the "natural" base for it, but we might fail | |
233 | * and end up putting it elsewhere. | |
234 | */ | |
a5bba930 | 235 | down_write(&mm->mmap_sem); |
a7f290da BH |
236 | vdso_base = get_unmapped_area(NULL, vdso_base, |
237 | vdso_pages << PAGE_SHIFT, 0, 0); | |
a5bba930 BH |
238 | if (IS_ERR_VALUE(vdso_base)) { |
239 | rc = vdso_base; | |
240 | goto fail_mmapsem; | |
a7f290da BH |
241 | } |
242 | ||
f2053f1a AB |
243 | /* |
244 | * Put vDSO base into mm struct. We need to do this before calling | |
245 | * install_special_mapping or the perf counter mmap tracking code | |
246 | * will fail to recognise it as a vDSO (since arch_vma_name fails). | |
247 | */ | |
248 | current->mm->context.vdso_base = vdso_base; | |
249 | ||
a7f290da BH |
250 | /* |
251 | * our vma flags don't have VM_WRITE so by default, the process isn't | |
252 | * allowed to write those pages. | |
253 | * gdb can break that with ptrace interface, and thus trigger COW on | |
254 | * those pages but it's then your responsibility to never do that on | |
255 | * the "data" page of the vDSO or you'll stop getting kernel updates | |
256 | * and your nice userland gettimeofday will be totally dead. | |
257 | * It's fine to use that for setting breakpoints in the vDSO code | |
258 | * pages though | |
c13e4ca2 | 259 | * |
3a0cfadb RM |
260 | * Make sure the vDSO gets into every core dump. |
261 | * Dumping its contents makes post-mortem fully interpretable later | |
262 | * without matching up the same kernel and hardware config to see | |
263 | * what PC values meant. | |
264 | */ | |
c13e4ca2 RM |
265 | rc = install_special_mapping(mm, vdso_base, vdso_pages << PAGE_SHIFT, |
266 | VM_READ|VM_EXEC| | |
267 | VM_MAYREAD|VM_MAYWRITE|VM_MAYEXEC| | |
268 | VM_ALWAYSDUMP, | |
269 | vdso_pagelist); | |
f2053f1a AB |
270 | if (rc) { |
271 | current->mm->context.vdso_base = 0; | |
c13e4ca2 | 272 | goto fail_mmapsem; |
f2053f1a | 273 | } |
c13e4ca2 | 274 | |
a7f290da | 275 | up_write(&mm->mmap_sem); |
a7f290da | 276 | return 0; |
a5bba930 | 277 | |
a5bba930 BH |
278 | fail_mmapsem: |
279 | up_write(&mm->mmap_sem); | |
280 | return rc; | |
281 | } | |
282 | ||
283 | const char *arch_vma_name(struct vm_area_struct *vma) | |
284 | { | |
285 | if (vma->vm_mm && vma->vm_start == vma->vm_mm->context.vdso_base) | |
286 | return "[vdso]"; | |
287 | return NULL; | |
a7f290da BH |
288 | } |
289 | ||
a5bba930 BH |
290 | |
291 | ||
a7f290da BH |
292 | static void * __init find_section32(Elf32_Ehdr *ehdr, const char *secname, |
293 | unsigned long *size) | |
294 | { | |
295 | Elf32_Shdr *sechdrs; | |
296 | unsigned int i; | |
297 | char *secnames; | |
298 | ||
299 | /* Grab section headers and strings so we can tell who is who */ | |
300 | sechdrs = (void *)ehdr + ehdr->e_shoff; | |
301 | secnames = (void *)ehdr + sechdrs[ehdr->e_shstrndx].sh_offset; | |
302 | ||
303 | /* Find the section they want */ | |
304 | for (i = 1; i < ehdr->e_shnum; i++) { | |
305 | if (strcmp(secnames+sechdrs[i].sh_name, secname) == 0) { | |
306 | if (size) | |
307 | *size = sechdrs[i].sh_size; | |
308 | return (void *)ehdr + sechdrs[i].sh_offset; | |
309 | } | |
310 | } | |
311 | *size = 0; | |
312 | return NULL; | |
313 | } | |
314 | ||
315 | static Elf32_Sym * __init find_symbol32(struct lib32_elfinfo *lib, | |
316 | const char *symname) | |
317 | { | |
318 | unsigned int i; | |
319 | char name[MAX_SYMNAME], *c; | |
320 | ||
321 | for (i = 0; i < (lib->dynsymsize / sizeof(Elf32_Sym)); i++) { | |
322 | if (lib->dynsym[i].st_name == 0) | |
323 | continue; | |
324 | strlcpy(name, lib->dynstr + lib->dynsym[i].st_name, | |
325 | MAX_SYMNAME); | |
326 | c = strchr(name, '@'); | |
327 | if (c) | |
328 | *c = 0; | |
329 | if (strcmp(symname, name) == 0) | |
330 | return &lib->dynsym[i]; | |
331 | } | |
332 | return NULL; | |
333 | } | |
334 | ||
335 | /* Note that we assume the section is .text and the symbol is relative to | |
336 | * the library base | |
337 | */ | |
338 | static unsigned long __init find_function32(struct lib32_elfinfo *lib, | |
339 | const char *symname) | |
340 | { | |
341 | Elf32_Sym *sym = find_symbol32(lib, symname); | |
342 | ||
343 | if (sym == NULL) { | |
344 | printk(KERN_WARNING "vDSO32: function %s not found !\n", | |
345 | symname); | |
346 | return 0; | |
347 | } | |
348 | return sym->st_value - VDSO32_LBASE; | |
349 | } | |
350 | ||
cf8918fe AB |
351 | static int __init vdso_do_func_patch32(struct lib32_elfinfo *v32, |
352 | struct lib64_elfinfo *v64, | |
353 | const char *orig, const char *fix) | |
a7f290da BH |
354 | { |
355 | Elf32_Sym *sym32_gen, *sym32_fix; | |
356 | ||
357 | sym32_gen = find_symbol32(v32, orig); | |
358 | if (sym32_gen == NULL) { | |
359 | printk(KERN_ERR "vDSO32: Can't find symbol %s !\n", orig); | |
360 | return -1; | |
361 | } | |
362 | if (fix == NULL) { | |
363 | sym32_gen->st_name = 0; | |
364 | return 0; | |
365 | } | |
366 | sym32_fix = find_symbol32(v32, fix); | |
367 | if (sym32_fix == NULL) { | |
368 | printk(KERN_ERR "vDSO32: Can't find symbol %s !\n", fix); | |
369 | return -1; | |
370 | } | |
371 | sym32_gen->st_value = sym32_fix->st_value; | |
372 | sym32_gen->st_size = sym32_fix->st_size; | |
373 | sym32_gen->st_info = sym32_fix->st_info; | |
374 | sym32_gen->st_other = sym32_fix->st_other; | |
375 | sym32_gen->st_shndx = sym32_fix->st_shndx; | |
376 | ||
377 | return 0; | |
378 | } | |
379 | ||
380 | ||
381 | #ifdef CONFIG_PPC64 | |
382 | ||
383 | static void * __init find_section64(Elf64_Ehdr *ehdr, const char *secname, | |
384 | unsigned long *size) | |
385 | { | |
386 | Elf64_Shdr *sechdrs; | |
387 | unsigned int i; | |
388 | char *secnames; | |
389 | ||
390 | /* Grab section headers and strings so we can tell who is who */ | |
391 | sechdrs = (void *)ehdr + ehdr->e_shoff; | |
392 | secnames = (void *)ehdr + sechdrs[ehdr->e_shstrndx].sh_offset; | |
393 | ||
394 | /* Find the section they want */ | |
395 | for (i = 1; i < ehdr->e_shnum; i++) { | |
396 | if (strcmp(secnames+sechdrs[i].sh_name, secname) == 0) { | |
397 | if (size) | |
398 | *size = sechdrs[i].sh_size; | |
399 | return (void *)ehdr + sechdrs[i].sh_offset; | |
400 | } | |
401 | } | |
402 | if (size) | |
403 | *size = 0; | |
404 | return NULL; | |
405 | } | |
406 | ||
407 | static Elf64_Sym * __init find_symbol64(struct lib64_elfinfo *lib, | |
408 | const char *symname) | |
409 | { | |
410 | unsigned int i; | |
411 | char name[MAX_SYMNAME], *c; | |
412 | ||
413 | for (i = 0; i < (lib->dynsymsize / sizeof(Elf64_Sym)); i++) { | |
414 | if (lib->dynsym[i].st_name == 0) | |
415 | continue; | |
416 | strlcpy(name, lib->dynstr + lib->dynsym[i].st_name, | |
417 | MAX_SYMNAME); | |
418 | c = strchr(name, '@'); | |
419 | if (c) | |
420 | *c = 0; | |
421 | if (strcmp(symname, name) == 0) | |
422 | return &lib->dynsym[i]; | |
423 | } | |
424 | return NULL; | |
425 | } | |
426 | ||
427 | /* Note that we assume the section is .text and the symbol is relative to | |
428 | * the library base | |
429 | */ | |
430 | static unsigned long __init find_function64(struct lib64_elfinfo *lib, | |
431 | const char *symname) | |
432 | { | |
433 | Elf64_Sym *sym = find_symbol64(lib, symname); | |
434 | ||
435 | if (sym == NULL) { | |
436 | printk(KERN_WARNING "vDSO64: function %s not found !\n", | |
437 | symname); | |
438 | return 0; | |
439 | } | |
440 | #ifdef VDS64_HAS_DESCRIPTORS | |
441 | return *((u64 *)(vdso64_kbase + sym->st_value - VDSO64_LBASE)) - | |
442 | VDSO64_LBASE; | |
443 | #else | |
444 | return sym->st_value - VDSO64_LBASE; | |
445 | #endif | |
446 | } | |
447 | ||
cf8918fe AB |
448 | static int __init vdso_do_func_patch64(struct lib32_elfinfo *v32, |
449 | struct lib64_elfinfo *v64, | |
450 | const char *orig, const char *fix) | |
a7f290da BH |
451 | { |
452 | Elf64_Sym *sym64_gen, *sym64_fix; | |
453 | ||
454 | sym64_gen = find_symbol64(v64, orig); | |
455 | if (sym64_gen == NULL) { | |
456 | printk(KERN_ERR "vDSO64: Can't find symbol %s !\n", orig); | |
457 | return -1; | |
458 | } | |
459 | if (fix == NULL) { | |
460 | sym64_gen->st_name = 0; | |
461 | return 0; | |
462 | } | |
463 | sym64_fix = find_symbol64(v64, fix); | |
464 | if (sym64_fix == NULL) { | |
465 | printk(KERN_ERR "vDSO64: Can't find symbol %s !\n", fix); | |
466 | return -1; | |
467 | } | |
468 | sym64_gen->st_value = sym64_fix->st_value; | |
469 | sym64_gen->st_size = sym64_fix->st_size; | |
470 | sym64_gen->st_info = sym64_fix->st_info; | |
471 | sym64_gen->st_other = sym64_fix->st_other; | |
472 | sym64_gen->st_shndx = sym64_fix->st_shndx; | |
473 | ||
474 | return 0; | |
475 | } | |
476 | ||
477 | #endif /* CONFIG_PPC64 */ | |
478 | ||
479 | ||
480 | static __init int vdso_do_find_sections(struct lib32_elfinfo *v32, | |
481 | struct lib64_elfinfo *v64) | |
482 | { | |
483 | void *sect; | |
484 | ||
485 | /* | |
486 | * Locate symbol tables & text section | |
487 | */ | |
488 | ||
489 | v32->dynsym = find_section32(v32->hdr, ".dynsym", &v32->dynsymsize); | |
490 | v32->dynstr = find_section32(v32->hdr, ".dynstr", NULL); | |
491 | if (v32->dynsym == NULL || v32->dynstr == NULL) { | |
492 | printk(KERN_ERR "vDSO32: required symbol section not found\n"); | |
493 | return -1; | |
494 | } | |
495 | sect = find_section32(v32->hdr, ".text", NULL); | |
496 | if (sect == NULL) { | |
497 | printk(KERN_ERR "vDSO32: the .text section was not found\n"); | |
498 | return -1; | |
499 | } | |
500 | v32->text = sect - vdso32_kbase; | |
501 | ||
502 | #ifdef CONFIG_PPC64 | |
503 | v64->dynsym = find_section64(v64->hdr, ".dynsym", &v64->dynsymsize); | |
504 | v64->dynstr = find_section64(v64->hdr, ".dynstr", NULL); | |
505 | if (v64->dynsym == NULL || v64->dynstr == NULL) { | |
506 | printk(KERN_ERR "vDSO64: required symbol section not found\n"); | |
507 | return -1; | |
508 | } | |
509 | sect = find_section64(v64->hdr, ".text", NULL); | |
510 | if (sect == NULL) { | |
511 | printk(KERN_ERR "vDSO64: the .text section was not found\n"); | |
512 | return -1; | |
513 | } | |
514 | v64->text = sect - vdso64_kbase; | |
515 | #endif /* CONFIG_PPC64 */ | |
516 | ||
517 | return 0; | |
518 | } | |
519 | ||
520 | static __init void vdso_setup_trampolines(struct lib32_elfinfo *v32, | |
521 | struct lib64_elfinfo *v64) | |
522 | { | |
523 | /* | |
524 | * Find signal trampolines | |
525 | */ | |
526 | ||
527 | #ifdef CONFIG_PPC64 | |
528 | vdso64_rt_sigtramp = find_function64(v64, "__kernel_sigtramp_rt64"); | |
529 | #endif | |
530 | vdso32_sigtramp = find_function32(v32, "__kernel_sigtramp32"); | |
531 | vdso32_rt_sigtramp = find_function32(v32, "__kernel_sigtramp_rt32"); | |
532 | } | |
533 | ||
534 | static __init int vdso_fixup_datapage(struct lib32_elfinfo *v32, | |
535 | struct lib64_elfinfo *v64) | |
536 | { | |
537 | Elf32_Sym *sym32; | |
538 | #ifdef CONFIG_PPC64 | |
539 | Elf64_Sym *sym64; | |
540 | ||
541 | sym64 = find_symbol64(v64, "__kernel_datapage_offset"); | |
542 | if (sym64 == NULL) { | |
543 | printk(KERN_ERR "vDSO64: Can't find symbol " | |
544 | "__kernel_datapage_offset !\n"); | |
545 | return -1; | |
546 | } | |
547 | *((int *)(vdso64_kbase + sym64->st_value - VDSO64_LBASE)) = | |
548 | (vdso64_pages << PAGE_SHIFT) - | |
549 | (sym64->st_value - VDSO64_LBASE); | |
550 | #endif /* CONFIG_PPC64 */ | |
551 | ||
552 | sym32 = find_symbol32(v32, "__kernel_datapage_offset"); | |
553 | if (sym32 == NULL) { | |
554 | printk(KERN_ERR "vDSO32: Can't find symbol " | |
555 | "__kernel_datapage_offset !\n"); | |
556 | return -1; | |
557 | } | |
558 | *((int *)(vdso32_kbase + (sym32->st_value - VDSO32_LBASE))) = | |
559 | (vdso32_pages << PAGE_SHIFT) - | |
560 | (sym32->st_value - VDSO32_LBASE); | |
561 | ||
562 | return 0; | |
563 | } | |
564 | ||
0909c8c2 BH |
565 | |
566 | static __init int vdso_fixup_features(struct lib32_elfinfo *v32, | |
567 | struct lib64_elfinfo *v64) | |
568 | { | |
569 | void *start32; | |
570 | unsigned long size32; | |
571 | ||
572 | #ifdef CONFIG_PPC64 | |
573 | void *start64; | |
574 | unsigned long size64; | |
575 | ||
576 | start64 = find_section64(v64->hdr, "__ftr_fixup", &size64); | |
577 | if (start64) | |
578 | do_feature_fixups(cur_cpu_spec->cpu_features, | |
579 | start64, start64 + size64); | |
580 | ||
7c03d653 BH |
581 | start64 = find_section64(v64->hdr, "__mmu_ftr_fixup", &size64); |
582 | if (start64) | |
583 | do_feature_fixups(cur_cpu_spec->mmu_features, | |
584 | start64, start64 + size64); | |
585 | ||
0909c8c2 BH |
586 | start64 = find_section64(v64->hdr, "__fw_ftr_fixup", &size64); |
587 | if (start64) | |
588 | do_feature_fixups(powerpc_firmware_features, | |
589 | start64, start64 + size64); | |
2d1b2027 KG |
590 | |
591 | start64 = find_section64(v64->hdr, "__lwsync_fixup", &size64); | |
592 | if (start64) | |
593 | do_lwsync_fixups(cur_cpu_spec->cpu_features, | |
594 | start64, start64 + size64); | |
0909c8c2 BH |
595 | #endif /* CONFIG_PPC64 */ |
596 | ||
597 | start32 = find_section32(v32->hdr, "__ftr_fixup", &size32); | |
598 | if (start32) | |
599 | do_feature_fixups(cur_cpu_spec->cpu_features, | |
600 | start32, start32 + size32); | |
601 | ||
7c03d653 BH |
602 | start32 = find_section32(v32->hdr, "__mmu_ftr_fixup", &size32); |
603 | if (start32) | |
604 | do_feature_fixups(cur_cpu_spec->mmu_features, | |
605 | start32, start32 + size32); | |
606 | ||
0909c8c2 BH |
607 | #ifdef CONFIG_PPC64 |
608 | start32 = find_section32(v32->hdr, "__fw_ftr_fixup", &size32); | |
609 | if (start32) | |
610 | do_feature_fixups(powerpc_firmware_features, | |
611 | start32, start32 + size32); | |
612 | #endif /* CONFIG_PPC64 */ | |
613 | ||
2d1b2027 KG |
614 | start32 = find_section32(v32->hdr, "__lwsync_fixup", &size32); |
615 | if (start32) | |
616 | do_lwsync_fixups(cur_cpu_spec->cpu_features, | |
617 | start32, start32 + size32); | |
618 | ||
0909c8c2 BH |
619 | return 0; |
620 | } | |
621 | ||
a7f290da BH |
622 | static __init int vdso_fixup_alt_funcs(struct lib32_elfinfo *v32, |
623 | struct lib64_elfinfo *v64) | |
624 | { | |
625 | int i; | |
626 | ||
627 | for (i = 0; i < ARRAY_SIZE(vdso_patches); i++) { | |
628 | struct vdso_patch_def *patch = &vdso_patches[i]; | |
629 | int match = (cur_cpu_spec->cpu_features & patch->ftr_mask) | |
630 | == patch->ftr_value; | |
631 | if (!match) | |
632 | continue; | |
633 | ||
634 | DBG("replacing %s with %s...\n", patch->gen_name, | |
635 | patch->fix_name ? "NONE" : patch->fix_name); | |
636 | ||
637 | /* | |
638 | * Patch the 32 bits and 64 bits symbols. Note that we do not | |
639 | * patch the "." symbol on 64 bits. | |
640 | * It would be easy to do, but doesn't seem to be necessary, | |
641 | * patching the OPD symbol is enough. | |
642 | */ | |
643 | vdso_do_func_patch32(v32, v64, patch->gen_name, | |
644 | patch->fix_name); | |
645 | #ifdef CONFIG_PPC64 | |
646 | vdso_do_func_patch64(v32, v64, patch->gen_name, | |
647 | patch->fix_name); | |
648 | #endif /* CONFIG_PPC64 */ | |
649 | } | |
650 | ||
651 | return 0; | |
652 | } | |
653 | ||
654 | ||
655 | static __init int vdso_setup(void) | |
656 | { | |
657 | struct lib32_elfinfo v32; | |
658 | struct lib64_elfinfo v64; | |
659 | ||
660 | v32.hdr = vdso32_kbase; | |
661 | #ifdef CONFIG_PPC64 | |
662 | v64.hdr = vdso64_kbase; | |
663 | #endif | |
664 | if (vdso_do_find_sections(&v32, &v64)) | |
665 | return -1; | |
666 | ||
667 | if (vdso_fixup_datapage(&v32, &v64)) | |
668 | return -1; | |
669 | ||
0909c8c2 BH |
670 | if (vdso_fixup_features(&v32, &v64)) |
671 | return -1; | |
672 | ||
a7f290da BH |
673 | if (vdso_fixup_alt_funcs(&v32, &v64)) |
674 | return -1; | |
675 | ||
676 | vdso_setup_trampolines(&v32, &v64); | |
677 | ||
678 | return 0; | |
679 | } | |
680 | ||
681 | /* | |
682 | * Called from setup_arch to initialize the bitmap of available | |
683 | * syscalls in the systemcfg page | |
684 | */ | |
685 | static void __init vdso_setup_syscall_map(void) | |
686 | { | |
687 | unsigned int i; | |
688 | extern unsigned long *sys_call_table; | |
689 | extern unsigned long sys_ni_syscall; | |
690 | ||
691 | ||
692 | for (i = 0; i < __NR_syscalls; i++) { | |
693 | #ifdef CONFIG_PPC64 | |
694 | if (sys_call_table[i*2] != sys_ni_syscall) | |
695 | vdso_data->syscall_map_64[i >> 5] |= | |
696 | 0x80000000UL >> (i & 0x1f); | |
697 | if (sys_call_table[i*2+1] != sys_ni_syscall) | |
698 | vdso_data->syscall_map_32[i >> 5] |= | |
699 | 0x80000000UL >> (i & 0x1f); | |
700 | #else /* CONFIG_PPC64 */ | |
701 | if (sys_call_table[i] != sys_ni_syscall) | |
702 | vdso_data->syscall_map_32[i >> 5] |= | |
703 | 0x80000000UL >> (i & 0x1f); | |
704 | #endif /* CONFIG_PPC64 */ | |
705 | } | |
706 | } | |
707 | ||
708 | ||
7ac9a137 | 709 | static int __init vdso_init(void) |
a7f290da BH |
710 | { |
711 | int i; | |
712 | ||
713 | #ifdef CONFIG_PPC64 | |
714 | /* | |
715 | * Fill up the "systemcfg" stuff for backward compatiblity | |
716 | */ | |
31fe5bf6 | 717 | strcpy((char *)vdso_data->eye_catcher, "SYSTEMCFG:PPC64"); |
a7f290da BH |
718 | vdso_data->version.major = SYSTEMCFG_MAJOR; |
719 | vdso_data->version.minor = SYSTEMCFG_MINOR; | |
720 | vdso_data->processor = mfspr(SPRN_PVR); | |
e8222502 BH |
721 | /* |
722 | * Fake the old platform number for pSeries and iSeries and add | |
723 | * in LPAR bit if necessary | |
724 | */ | |
725 | vdso_data->platform = machine_is(iseries) ? 0x200 : 0x100; | |
726 | if (firmware_has_feature(FW_FEATURE_LPAR)) | |
727 | vdso_data->platform |= 1; | |
a7f290da BH |
728 | vdso_data->physicalMemorySize = lmb_phys_mem_size(); |
729 | vdso_data->dcache_size = ppc64_caches.dsize; | |
730 | vdso_data->dcache_line_size = ppc64_caches.dline_size; | |
731 | vdso_data->icache_size = ppc64_caches.isize; | |
732 | vdso_data->icache_line_size = ppc64_caches.iline_size; | |
733 | ||
fbe48175 OJ |
734 | /* XXXOJN: Blocks should be added to ppc64_caches and used instead */ |
735 | vdso_data->dcache_block_size = ppc64_caches.dline_size; | |
736 | vdso_data->icache_block_size = ppc64_caches.iline_size; | |
737 | vdso_data->dcache_log_block_size = ppc64_caches.log_dline_size; | |
738 | vdso_data->icache_log_block_size = ppc64_caches.log_iline_size; | |
739 | ||
a7f290da BH |
740 | /* |
741 | * Calculate the size of the 64 bits vDSO | |
742 | */ | |
743 | vdso64_pages = (&vdso64_end - &vdso64_start) >> PAGE_SHIFT; | |
744 | DBG("vdso64_kbase: %p, 0x%x pages\n", vdso64_kbase, vdso64_pages); | |
fbe48175 OJ |
745 | #else |
746 | vdso_data->dcache_block_size = L1_CACHE_BYTES; | |
747 | vdso_data->dcache_log_block_size = L1_CACHE_SHIFT; | |
748 | vdso_data->icache_block_size = L1_CACHE_BYTES; | |
749 | vdso_data->icache_log_block_size = L1_CACHE_SHIFT; | |
a7f290da BH |
750 | #endif /* CONFIG_PPC64 */ |
751 | ||
752 | ||
753 | /* | |
754 | * Calculate the size of the 32 bits vDSO | |
755 | */ | |
756 | vdso32_pages = (&vdso32_end - &vdso32_start) >> PAGE_SHIFT; | |
757 | DBG("vdso32_kbase: %p, 0x%x pages\n", vdso32_kbase, vdso32_pages); | |
758 | ||
759 | ||
760 | /* | |
761 | * Setup the syscall map in the vDOS | |
762 | */ | |
763 | vdso_setup_syscall_map(); | |
0909c8c2 | 764 | |
a7f290da BH |
765 | /* |
766 | * Initialize the vDSO images in memory, that is do necessary | |
767 | * fixups of vDSO symbols, locate trampolines, etc... | |
768 | */ | |
769 | if (vdso_setup()) { | |
770 | printk(KERN_ERR "vDSO setup failure, not enabled !\n"); | |
771 | vdso32_pages = 0; | |
772 | #ifdef CONFIG_PPC64 | |
773 | vdso64_pages = 0; | |
774 | #endif | |
7ac9a137 | 775 | return 0; |
a7f290da BH |
776 | } |
777 | ||
778 | /* Make sure pages are in the correct state */ | |
7ac9a137 BH |
779 | vdso32_pagelist = kzalloc(sizeof(struct page *) * (vdso32_pages + 2), |
780 | GFP_KERNEL); | |
781 | BUG_ON(vdso32_pagelist == NULL); | |
a7f290da BH |
782 | for (i = 0; i < vdso32_pages; i++) { |
783 | struct page *pg = virt_to_page(vdso32_kbase + i*PAGE_SIZE); | |
784 | ClearPageReserved(pg); | |
785 | get_page(pg); | |
c13e4ca2 | 786 | vdso32_pagelist[i] = pg; |
a7f290da | 787 | } |
c13e4ca2 RM |
788 | vdso32_pagelist[i++] = virt_to_page(vdso_data); |
789 | vdso32_pagelist[i] = NULL; | |
790 | ||
a7f290da | 791 | #ifdef CONFIG_PPC64 |
7ac9a137 BH |
792 | vdso64_pagelist = kzalloc(sizeof(struct page *) * (vdso64_pages + 2), |
793 | GFP_KERNEL); | |
794 | BUG_ON(vdso64_pagelist == NULL); | |
a7f290da BH |
795 | for (i = 0; i < vdso64_pages; i++) { |
796 | struct page *pg = virt_to_page(vdso64_kbase + i*PAGE_SIZE); | |
797 | ClearPageReserved(pg); | |
798 | get_page(pg); | |
c13e4ca2 | 799 | vdso64_pagelist[i] = pg; |
a7f290da | 800 | } |
c13e4ca2 RM |
801 | vdso64_pagelist[i++] = virt_to_page(vdso_data); |
802 | vdso64_pagelist[i] = NULL; | |
a7f290da BH |
803 | #endif /* CONFIG_PPC64 */ |
804 | ||
805 | get_page(virt_to_page(vdso_data)); | |
7ac9a137 BH |
806 | |
807 | smp_wmb(); | |
808 | vdso_ready = 1; | |
809 | ||
810 | return 0; | |
a7f290da | 811 | } |
7ac9a137 | 812 | arch_initcall(vdso_init); |
a7f290da BH |
813 | |
814 | int in_gate_area_no_task(unsigned long addr) | |
815 | { | |
816 | return 0; | |
817 | } | |
818 | ||
819 | int in_gate_area(struct task_struct *task, unsigned long addr) | |
820 | { | |
821 | return 0; | |
822 | } | |
823 | ||
824 | struct vm_area_struct *get_gate_vma(struct task_struct *tsk) | |
825 | { | |
826 | return NULL; | |
827 | } | |
828 |