Commit | Line | Data |
---|---|---|
a7f290da | 1 | /* |
a7f290da BH |
2 | * Copyright (C) 2004 Benjamin Herrenschmidt, IBM Corp. |
3 | * <benh@kernel.crashing.org> | |
4 | * | |
5 | * This program is free software; you can redistribute it and/or | |
6 | * modify it under the terms of the GNU General Public License | |
7 | * as published by the Free Software Foundation; either version | |
8 | * 2 of the License, or (at your option) any later version. | |
9 | */ | |
10 | ||
a7f290da BH |
11 | #include <linux/module.h> |
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> | |
17 | #include <linux/smp_lock.h> | |
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> | |
25 | ||
26 | #include <asm/pgtable.h> | |
27 | #include <asm/system.h> | |
28 | #include <asm/processor.h> | |
29 | #include <asm/mmu.h> | |
30 | #include <asm/mmu_context.h> | |
31 | #include <asm/lmb.h> | |
32 | #include <asm/machdep.h> | |
33 | #include <asm/cputable.h> | |
34 | #include <asm/sections.h> | |
e8222502 | 35 | #include <asm/firmware.h> |
a7f290da BH |
36 | #include <asm/vdso.h> |
37 | #include <asm/vdso_datapage.h> | |
38 | ||
0909c8c2 BH |
39 | #include "setup.h" |
40 | ||
a7f290da BH |
41 | #undef DEBUG |
42 | ||
43 | #ifdef DEBUG | |
44 | #define DBG(fmt...) printk(fmt) | |
45 | #else | |
46 | #define DBG(fmt...) | |
47 | #endif | |
48 | ||
49 | /* Max supported size for symbol names */ | |
50 | #define MAX_SYMNAME 64 | |
51 | ||
52 | extern char vdso32_start, vdso32_end; | |
53 | static void *vdso32_kbase = &vdso32_start; | |
54 | unsigned int vdso32_pages; | |
55 | unsigned long vdso32_sigtramp; | |
56 | unsigned long vdso32_rt_sigtramp; | |
57 | ||
58 | #ifdef CONFIG_PPC64 | |
59 | extern char vdso64_start, vdso64_end; | |
60 | static void *vdso64_kbase = &vdso64_start; | |
61 | unsigned int vdso64_pages; | |
62 | unsigned long vdso64_rt_sigtramp; | |
63 | #endif /* CONFIG_PPC64 */ | |
64 | ||
65 | /* | |
66 | * The vdso data page (aka. systemcfg for old ppc64 fans) is here. | |
67 | * Once the early boot kernel code no longer needs to muck around | |
68 | * with it, it will become dynamically allocated | |
69 | */ | |
70 | static union { | |
71 | struct vdso_data data; | |
72 | u8 page[PAGE_SIZE]; | |
73 | } vdso_data_store __attribute__((__section__(".data.page_aligned"))); | |
74 | struct vdso_data *vdso_data = &vdso_data_store.data; | |
75 | ||
76 | /* Format of the patch table */ | |
77 | struct vdso_patch_def | |
78 | { | |
79 | unsigned long ftr_mask, ftr_value; | |
80 | const char *gen_name; | |
81 | const char *fix_name; | |
82 | }; | |
83 | ||
84 | /* Table of functions to patch based on the CPU type/revision | |
85 | * | |
86 | * Currently, we only change sync_dicache to do nothing on processors | |
87 | * with a coherent icache | |
88 | */ | |
89 | static struct vdso_patch_def vdso_patches[] = { | |
90 | { | |
91 | CPU_FTR_COHERENT_ICACHE, CPU_FTR_COHERENT_ICACHE, | |
92 | "__kernel_sync_dicache", "__kernel_sync_dicache_p5" | |
93 | }, | |
94 | { | |
95 | CPU_FTR_USE_TB, 0, | |
96 | "__kernel_gettimeofday", NULL | |
97 | }, | |
98 | }; | |
99 | ||
100 | /* | |
101 | * Some infos carried around for each of them during parsing at | |
102 | * boot time. | |
103 | */ | |
104 | struct lib32_elfinfo | |
105 | { | |
106 | Elf32_Ehdr *hdr; /* ptr to ELF */ | |
107 | Elf32_Sym *dynsym; /* ptr to .dynsym section */ | |
108 | unsigned long dynsymsize; /* size of .dynsym section */ | |
109 | char *dynstr; /* ptr to .dynstr section */ | |
110 | unsigned long text; /* offset of .text section in .so */ | |
111 | }; | |
112 | ||
113 | struct lib64_elfinfo | |
114 | { | |
115 | Elf64_Ehdr *hdr; | |
116 | Elf64_Sym *dynsym; | |
117 | unsigned long dynsymsize; | |
118 | char *dynstr; | |
119 | unsigned long text; | |
120 | }; | |
121 | ||
122 | ||
123 | #ifdef __DEBUG | |
124 | static void dump_one_vdso_page(struct page *pg, struct page *upg) | |
125 | { | |
126 | printk("kpg: %p (c:%d,f:%08lx)", __va(page_to_pfn(pg) << PAGE_SHIFT), | |
127 | page_count(pg), | |
128 | pg->flags); | |
129 | if (upg/* && pg != upg*/) { | |
130 | printk(" upg: %p (c:%d,f:%08lx)", __va(page_to_pfn(upg) | |
131 | << PAGE_SHIFT), | |
132 | page_count(upg), | |
133 | upg->flags); | |
134 | } | |
135 | printk("\n"); | |
136 | } | |
137 | ||
138 | static void dump_vdso_pages(struct vm_area_struct * vma) | |
139 | { | |
140 | int i; | |
141 | ||
142 | if (!vma || test_thread_flag(TIF_32BIT)) { | |
143 | printk("vDSO32 @ %016lx:\n", (unsigned long)vdso32_kbase); | |
144 | for (i=0; i<vdso32_pages; i++) { | |
145 | struct page *pg = virt_to_page(vdso32_kbase + | |
146 | i*PAGE_SIZE); | |
147 | struct page *upg = (vma && vma->vm_mm) ? | |
6aab341e | 148 | follow_page(vma, vma->vm_start + i*PAGE_SIZE, 0) |
a7f290da BH |
149 | : NULL; |
150 | dump_one_vdso_page(pg, upg); | |
151 | } | |
152 | } | |
153 | if (!vma || !test_thread_flag(TIF_32BIT)) { | |
154 | printk("vDSO64 @ %016lx:\n", (unsigned long)vdso64_kbase); | |
155 | for (i=0; i<vdso64_pages; i++) { | |
156 | struct page *pg = virt_to_page(vdso64_kbase + | |
157 | i*PAGE_SIZE); | |
158 | struct page *upg = (vma && vma->vm_mm) ? | |
6aab341e | 159 | follow_page(vma, vma->vm_start + i*PAGE_SIZE, 0) |
a7f290da BH |
160 | : NULL; |
161 | dump_one_vdso_page(pg, upg); | |
162 | } | |
163 | } | |
164 | } | |
165 | #endif /* DEBUG */ | |
166 | ||
167 | /* | |
168 | * Keep a dummy vma_close for now, it will prevent VMA merging. | |
169 | */ | |
170 | static void vdso_vma_close(struct vm_area_struct * vma) | |
171 | { | |
172 | } | |
173 | ||
174 | /* | |
175 | * Our nopage() function, maps in the actual vDSO kernel pages, they will | |
176 | * be mapped read-only by do_no_page(), and eventually COW'ed, either | |
177 | * right away for an initial write access, or by do_wp_page(). | |
178 | */ | |
179 | static struct page * vdso_vma_nopage(struct vm_area_struct * vma, | |
180 | unsigned long address, int *type) | |
181 | { | |
182 | unsigned long offset = address - vma->vm_start; | |
183 | struct page *pg; | |
184 | #ifdef CONFIG_PPC64 | |
0551fbd2 BH |
185 | void *vbase = (vma->vm_mm->task_size > TASK_SIZE_USER32) ? |
186 | vdso64_kbase : vdso32_kbase; | |
a7f290da BH |
187 | #else |
188 | void *vbase = vdso32_kbase; | |
189 | #endif | |
190 | ||
191 | DBG("vdso_vma_nopage(current: %s, address: %016lx, off: %lx)\n", | |
192 | current->comm, address, offset); | |
193 | ||
194 | if (address < vma->vm_start || address > vma->vm_end) | |
195 | return NOPAGE_SIGBUS; | |
196 | ||
197 | /* | |
198 | * Last page is systemcfg. | |
199 | */ | |
200 | if ((vma->vm_end - address) <= PAGE_SIZE) | |
201 | pg = virt_to_page(vdso_data); | |
202 | else | |
203 | pg = virt_to_page(vbase + offset); | |
204 | ||
205 | get_page(pg); | |
206 | DBG(" ->page count: %d\n", page_count(pg)); | |
207 | ||
208 | return pg; | |
209 | } | |
210 | ||
211 | static struct vm_operations_struct vdso_vmops = { | |
212 | .close = vdso_vma_close, | |
213 | .nopage = vdso_vma_nopage, | |
214 | }; | |
215 | ||
216 | /* | |
217 | * This is called from binfmt_elf, we create the special vma for the | |
218 | * vDSO and insert it into the mm struct tree | |
219 | */ | |
220 | int arch_setup_additional_pages(struct linux_binprm *bprm, | |
221 | int executable_stack) | |
222 | { | |
223 | struct mm_struct *mm = current->mm; | |
224 | struct vm_area_struct *vma; | |
225 | unsigned long vdso_pages; | |
226 | unsigned long vdso_base; | |
a5bba930 | 227 | int rc; |
a7f290da BH |
228 | |
229 | #ifdef CONFIG_PPC64 | |
230 | if (test_thread_flag(TIF_32BIT)) { | |
231 | vdso_pages = vdso32_pages; | |
232 | vdso_base = VDSO32_MBASE; | |
233 | } else { | |
234 | vdso_pages = vdso64_pages; | |
235 | vdso_base = VDSO64_MBASE; | |
236 | } | |
237 | #else | |
238 | vdso_pages = vdso32_pages; | |
239 | vdso_base = VDSO32_MBASE; | |
240 | #endif | |
241 | ||
a5bba930 | 242 | current->mm->context.vdso_base = 0; |
a7f290da BH |
243 | |
244 | /* vDSO has a problem and was disabled, just don't "enable" it for the | |
245 | * process | |
246 | */ | |
247 | if (vdso_pages == 0) | |
248 | return 0; | |
a7f290da BH |
249 | /* Add a page to the vdso size for the data page */ |
250 | vdso_pages ++; | |
251 | ||
252 | /* | |
253 | * pick a base address for the vDSO in process space. We try to put it | |
254 | * at vdso_base which is the "natural" base for it, but we might fail | |
255 | * and end up putting it elsewhere. | |
256 | */ | |
a5bba930 | 257 | down_write(&mm->mmap_sem); |
a7f290da BH |
258 | vdso_base = get_unmapped_area(NULL, vdso_base, |
259 | vdso_pages << PAGE_SHIFT, 0, 0); | |
a5bba930 BH |
260 | if (IS_ERR_VALUE(vdso_base)) { |
261 | rc = vdso_base; | |
262 | goto fail_mmapsem; | |
a7f290da BH |
263 | } |
264 | ||
a7f290da | 265 | |
a5bba930 | 266 | /* Allocate a VMA structure and fill it up */ |
e94b1766 | 267 | vma = kmem_cache_zalloc(vm_area_cachep, GFP_KERNEL); |
a5bba930 BH |
268 | if (vma == NULL) { |
269 | rc = -ENOMEM; | |
270 | goto fail_mmapsem; | |
271 | } | |
a7f290da | 272 | vma->vm_mm = mm; |
a5bba930 | 273 | vma->vm_start = vdso_base; |
a7f290da BH |
274 | vma->vm_end = vma->vm_start + (vdso_pages << PAGE_SHIFT); |
275 | ||
276 | /* | |
277 | * our vma flags don't have VM_WRITE so by default, the process isn't | |
278 | * allowed to write those pages. | |
279 | * gdb can break that with ptrace interface, and thus trigger COW on | |
280 | * those pages but it's then your responsibility to never do that on | |
281 | * the "data" page of the vDSO or you'll stop getting kernel updates | |
282 | * and your nice userland gettimeofday will be totally dead. | |
283 | * It's fine to use that for setting breakpoints in the vDSO code | |
284 | * pages though | |
285 | */ | |
a5bba930 | 286 | vma->vm_flags = VM_READ|VM_EXEC|VM_MAYREAD|VM_MAYWRITE|VM_MAYEXEC; |
3a0cfadb RM |
287 | /* |
288 | * Make sure the vDSO gets into every core dump. | |
289 | * Dumping its contents makes post-mortem fully interpretable later | |
290 | * without matching up the same kernel and hardware config to see | |
291 | * what PC values meant. | |
292 | */ | |
293 | vma->vm_flags |= VM_ALWAYSDUMP; | |
a7f290da BH |
294 | vma->vm_flags |= mm->def_flags; |
295 | vma->vm_page_prot = protection_map[vma->vm_flags & 0x7]; | |
296 | vma->vm_ops = &vdso_vmops; | |
297 | ||
a5bba930 BH |
298 | /* Insert new VMA */ |
299 | rc = insert_vm_struct(mm, vma); | |
300 | if (rc) | |
301 | goto fail_vma; | |
302 | ||
303 | /* Put vDSO base into mm struct and account for memory usage */ | |
304 | current->mm->context.vdso_base = vdso_base; | |
a7f290da BH |
305 | mm->total_vm += (vma->vm_end - vma->vm_start) >> PAGE_SHIFT; |
306 | up_write(&mm->mmap_sem); | |
a7f290da | 307 | return 0; |
a5bba930 BH |
308 | |
309 | fail_vma: | |
310 | kmem_cache_free(vm_area_cachep, vma); | |
311 | fail_mmapsem: | |
312 | up_write(&mm->mmap_sem); | |
313 | return rc; | |
314 | } | |
315 | ||
316 | const char *arch_vma_name(struct vm_area_struct *vma) | |
317 | { | |
318 | if (vma->vm_mm && vma->vm_start == vma->vm_mm->context.vdso_base) | |
319 | return "[vdso]"; | |
320 | return NULL; | |
a7f290da BH |
321 | } |
322 | ||
a5bba930 BH |
323 | |
324 | ||
a7f290da BH |
325 | static void * __init find_section32(Elf32_Ehdr *ehdr, const char *secname, |
326 | unsigned long *size) | |
327 | { | |
328 | Elf32_Shdr *sechdrs; | |
329 | unsigned int i; | |
330 | char *secnames; | |
331 | ||
332 | /* Grab section headers and strings so we can tell who is who */ | |
333 | sechdrs = (void *)ehdr + ehdr->e_shoff; | |
334 | secnames = (void *)ehdr + sechdrs[ehdr->e_shstrndx].sh_offset; | |
335 | ||
336 | /* Find the section they want */ | |
337 | for (i = 1; i < ehdr->e_shnum; i++) { | |
338 | if (strcmp(secnames+sechdrs[i].sh_name, secname) == 0) { | |
339 | if (size) | |
340 | *size = sechdrs[i].sh_size; | |
341 | return (void *)ehdr + sechdrs[i].sh_offset; | |
342 | } | |
343 | } | |
344 | *size = 0; | |
345 | return NULL; | |
346 | } | |
347 | ||
348 | static Elf32_Sym * __init find_symbol32(struct lib32_elfinfo *lib, | |
349 | const char *symname) | |
350 | { | |
351 | unsigned int i; | |
352 | char name[MAX_SYMNAME], *c; | |
353 | ||
354 | for (i = 0; i < (lib->dynsymsize / sizeof(Elf32_Sym)); i++) { | |
355 | if (lib->dynsym[i].st_name == 0) | |
356 | continue; | |
357 | strlcpy(name, lib->dynstr + lib->dynsym[i].st_name, | |
358 | MAX_SYMNAME); | |
359 | c = strchr(name, '@'); | |
360 | if (c) | |
361 | *c = 0; | |
362 | if (strcmp(symname, name) == 0) | |
363 | return &lib->dynsym[i]; | |
364 | } | |
365 | return NULL; | |
366 | } | |
367 | ||
368 | /* Note that we assume the section is .text and the symbol is relative to | |
369 | * the library base | |
370 | */ | |
371 | static unsigned long __init find_function32(struct lib32_elfinfo *lib, | |
372 | const char *symname) | |
373 | { | |
374 | Elf32_Sym *sym = find_symbol32(lib, symname); | |
375 | ||
376 | if (sym == NULL) { | |
377 | printk(KERN_WARNING "vDSO32: function %s not found !\n", | |
378 | symname); | |
379 | return 0; | |
380 | } | |
381 | return sym->st_value - VDSO32_LBASE; | |
382 | } | |
383 | ||
384 | static int vdso_do_func_patch32(struct lib32_elfinfo *v32, | |
385 | struct lib64_elfinfo *v64, | |
386 | const char *orig, const char *fix) | |
387 | { | |
388 | Elf32_Sym *sym32_gen, *sym32_fix; | |
389 | ||
390 | sym32_gen = find_symbol32(v32, orig); | |
391 | if (sym32_gen == NULL) { | |
392 | printk(KERN_ERR "vDSO32: Can't find symbol %s !\n", orig); | |
393 | return -1; | |
394 | } | |
395 | if (fix == NULL) { | |
396 | sym32_gen->st_name = 0; | |
397 | return 0; | |
398 | } | |
399 | sym32_fix = find_symbol32(v32, fix); | |
400 | if (sym32_fix == NULL) { | |
401 | printk(KERN_ERR "vDSO32: Can't find symbol %s !\n", fix); | |
402 | return -1; | |
403 | } | |
404 | sym32_gen->st_value = sym32_fix->st_value; | |
405 | sym32_gen->st_size = sym32_fix->st_size; | |
406 | sym32_gen->st_info = sym32_fix->st_info; | |
407 | sym32_gen->st_other = sym32_fix->st_other; | |
408 | sym32_gen->st_shndx = sym32_fix->st_shndx; | |
409 | ||
410 | return 0; | |
411 | } | |
412 | ||
413 | ||
414 | #ifdef CONFIG_PPC64 | |
415 | ||
416 | static void * __init find_section64(Elf64_Ehdr *ehdr, const char *secname, | |
417 | unsigned long *size) | |
418 | { | |
419 | Elf64_Shdr *sechdrs; | |
420 | unsigned int i; | |
421 | char *secnames; | |
422 | ||
423 | /* Grab section headers and strings so we can tell who is who */ | |
424 | sechdrs = (void *)ehdr + ehdr->e_shoff; | |
425 | secnames = (void *)ehdr + sechdrs[ehdr->e_shstrndx].sh_offset; | |
426 | ||
427 | /* Find the section they want */ | |
428 | for (i = 1; i < ehdr->e_shnum; i++) { | |
429 | if (strcmp(secnames+sechdrs[i].sh_name, secname) == 0) { | |
430 | if (size) | |
431 | *size = sechdrs[i].sh_size; | |
432 | return (void *)ehdr + sechdrs[i].sh_offset; | |
433 | } | |
434 | } | |
435 | if (size) | |
436 | *size = 0; | |
437 | return NULL; | |
438 | } | |
439 | ||
440 | static Elf64_Sym * __init find_symbol64(struct lib64_elfinfo *lib, | |
441 | const char *symname) | |
442 | { | |
443 | unsigned int i; | |
444 | char name[MAX_SYMNAME], *c; | |
445 | ||
446 | for (i = 0; i < (lib->dynsymsize / sizeof(Elf64_Sym)); i++) { | |
447 | if (lib->dynsym[i].st_name == 0) | |
448 | continue; | |
449 | strlcpy(name, lib->dynstr + lib->dynsym[i].st_name, | |
450 | MAX_SYMNAME); | |
451 | c = strchr(name, '@'); | |
452 | if (c) | |
453 | *c = 0; | |
454 | if (strcmp(symname, name) == 0) | |
455 | return &lib->dynsym[i]; | |
456 | } | |
457 | return NULL; | |
458 | } | |
459 | ||
460 | /* Note that we assume the section is .text and the symbol is relative to | |
461 | * the library base | |
462 | */ | |
463 | static unsigned long __init find_function64(struct lib64_elfinfo *lib, | |
464 | const char *symname) | |
465 | { | |
466 | Elf64_Sym *sym = find_symbol64(lib, symname); | |
467 | ||
468 | if (sym == NULL) { | |
469 | printk(KERN_WARNING "vDSO64: function %s not found !\n", | |
470 | symname); | |
471 | return 0; | |
472 | } | |
473 | #ifdef VDS64_HAS_DESCRIPTORS | |
474 | return *((u64 *)(vdso64_kbase + sym->st_value - VDSO64_LBASE)) - | |
475 | VDSO64_LBASE; | |
476 | #else | |
477 | return sym->st_value - VDSO64_LBASE; | |
478 | #endif | |
479 | } | |
480 | ||
481 | static int vdso_do_func_patch64(struct lib32_elfinfo *v32, | |
482 | struct lib64_elfinfo *v64, | |
483 | const char *orig, const char *fix) | |
484 | { | |
485 | Elf64_Sym *sym64_gen, *sym64_fix; | |
486 | ||
487 | sym64_gen = find_symbol64(v64, orig); | |
488 | if (sym64_gen == NULL) { | |
489 | printk(KERN_ERR "vDSO64: Can't find symbol %s !\n", orig); | |
490 | return -1; | |
491 | } | |
492 | if (fix == NULL) { | |
493 | sym64_gen->st_name = 0; | |
494 | return 0; | |
495 | } | |
496 | sym64_fix = find_symbol64(v64, fix); | |
497 | if (sym64_fix == NULL) { | |
498 | printk(KERN_ERR "vDSO64: Can't find symbol %s !\n", fix); | |
499 | return -1; | |
500 | } | |
501 | sym64_gen->st_value = sym64_fix->st_value; | |
502 | sym64_gen->st_size = sym64_fix->st_size; | |
503 | sym64_gen->st_info = sym64_fix->st_info; | |
504 | sym64_gen->st_other = sym64_fix->st_other; | |
505 | sym64_gen->st_shndx = sym64_fix->st_shndx; | |
506 | ||
507 | return 0; | |
508 | } | |
509 | ||
510 | #endif /* CONFIG_PPC64 */ | |
511 | ||
512 | ||
513 | static __init int vdso_do_find_sections(struct lib32_elfinfo *v32, | |
514 | struct lib64_elfinfo *v64) | |
515 | { | |
516 | void *sect; | |
517 | ||
518 | /* | |
519 | * Locate symbol tables & text section | |
520 | */ | |
521 | ||
522 | v32->dynsym = find_section32(v32->hdr, ".dynsym", &v32->dynsymsize); | |
523 | v32->dynstr = find_section32(v32->hdr, ".dynstr", NULL); | |
524 | if (v32->dynsym == NULL || v32->dynstr == NULL) { | |
525 | printk(KERN_ERR "vDSO32: required symbol section not found\n"); | |
526 | return -1; | |
527 | } | |
528 | sect = find_section32(v32->hdr, ".text", NULL); | |
529 | if (sect == NULL) { | |
530 | printk(KERN_ERR "vDSO32: the .text section was not found\n"); | |
531 | return -1; | |
532 | } | |
533 | v32->text = sect - vdso32_kbase; | |
534 | ||
535 | #ifdef CONFIG_PPC64 | |
536 | v64->dynsym = find_section64(v64->hdr, ".dynsym", &v64->dynsymsize); | |
537 | v64->dynstr = find_section64(v64->hdr, ".dynstr", NULL); | |
538 | if (v64->dynsym == NULL || v64->dynstr == NULL) { | |
539 | printk(KERN_ERR "vDSO64: required symbol section not found\n"); | |
540 | return -1; | |
541 | } | |
542 | sect = find_section64(v64->hdr, ".text", NULL); | |
543 | if (sect == NULL) { | |
544 | printk(KERN_ERR "vDSO64: the .text section was not found\n"); | |
545 | return -1; | |
546 | } | |
547 | v64->text = sect - vdso64_kbase; | |
548 | #endif /* CONFIG_PPC64 */ | |
549 | ||
550 | return 0; | |
551 | } | |
552 | ||
553 | static __init void vdso_setup_trampolines(struct lib32_elfinfo *v32, | |
554 | struct lib64_elfinfo *v64) | |
555 | { | |
556 | /* | |
557 | * Find signal trampolines | |
558 | */ | |
559 | ||
560 | #ifdef CONFIG_PPC64 | |
561 | vdso64_rt_sigtramp = find_function64(v64, "__kernel_sigtramp_rt64"); | |
562 | #endif | |
563 | vdso32_sigtramp = find_function32(v32, "__kernel_sigtramp32"); | |
564 | vdso32_rt_sigtramp = find_function32(v32, "__kernel_sigtramp_rt32"); | |
565 | } | |
566 | ||
567 | static __init int vdso_fixup_datapage(struct lib32_elfinfo *v32, | |
568 | struct lib64_elfinfo *v64) | |
569 | { | |
570 | Elf32_Sym *sym32; | |
571 | #ifdef CONFIG_PPC64 | |
572 | Elf64_Sym *sym64; | |
573 | ||
574 | sym64 = find_symbol64(v64, "__kernel_datapage_offset"); | |
575 | if (sym64 == NULL) { | |
576 | printk(KERN_ERR "vDSO64: Can't find symbol " | |
577 | "__kernel_datapage_offset !\n"); | |
578 | return -1; | |
579 | } | |
580 | *((int *)(vdso64_kbase + sym64->st_value - VDSO64_LBASE)) = | |
581 | (vdso64_pages << PAGE_SHIFT) - | |
582 | (sym64->st_value - VDSO64_LBASE); | |
583 | #endif /* CONFIG_PPC64 */ | |
584 | ||
585 | sym32 = find_symbol32(v32, "__kernel_datapage_offset"); | |
586 | if (sym32 == NULL) { | |
587 | printk(KERN_ERR "vDSO32: Can't find symbol " | |
588 | "__kernel_datapage_offset !\n"); | |
589 | return -1; | |
590 | } | |
591 | *((int *)(vdso32_kbase + (sym32->st_value - VDSO32_LBASE))) = | |
592 | (vdso32_pages << PAGE_SHIFT) - | |
593 | (sym32->st_value - VDSO32_LBASE); | |
594 | ||
595 | return 0; | |
596 | } | |
597 | ||
0909c8c2 BH |
598 | |
599 | static __init int vdso_fixup_features(struct lib32_elfinfo *v32, | |
600 | struct lib64_elfinfo *v64) | |
601 | { | |
602 | void *start32; | |
603 | unsigned long size32; | |
604 | ||
605 | #ifdef CONFIG_PPC64 | |
606 | void *start64; | |
607 | unsigned long size64; | |
608 | ||
609 | start64 = find_section64(v64->hdr, "__ftr_fixup", &size64); | |
610 | if (start64) | |
611 | do_feature_fixups(cur_cpu_spec->cpu_features, | |
612 | start64, start64 + size64); | |
613 | ||
614 | start64 = find_section64(v64->hdr, "__fw_ftr_fixup", &size64); | |
615 | if (start64) | |
616 | do_feature_fixups(powerpc_firmware_features, | |
617 | start64, start64 + size64); | |
618 | #endif /* CONFIG_PPC64 */ | |
619 | ||
620 | start32 = find_section32(v32->hdr, "__ftr_fixup", &size32); | |
621 | if (start32) | |
622 | do_feature_fixups(cur_cpu_spec->cpu_features, | |
623 | start32, start32 + size32); | |
624 | ||
625 | #ifdef CONFIG_PPC64 | |
626 | start32 = find_section32(v32->hdr, "__fw_ftr_fixup", &size32); | |
627 | if (start32) | |
628 | do_feature_fixups(powerpc_firmware_features, | |
629 | start32, start32 + size32); | |
630 | #endif /* CONFIG_PPC64 */ | |
631 | ||
632 | return 0; | |
633 | } | |
634 | ||
a7f290da BH |
635 | static __init int vdso_fixup_alt_funcs(struct lib32_elfinfo *v32, |
636 | struct lib64_elfinfo *v64) | |
637 | { | |
638 | int i; | |
639 | ||
640 | for (i = 0; i < ARRAY_SIZE(vdso_patches); i++) { | |
641 | struct vdso_patch_def *patch = &vdso_patches[i]; | |
642 | int match = (cur_cpu_spec->cpu_features & patch->ftr_mask) | |
643 | == patch->ftr_value; | |
644 | if (!match) | |
645 | continue; | |
646 | ||
647 | DBG("replacing %s with %s...\n", patch->gen_name, | |
648 | patch->fix_name ? "NONE" : patch->fix_name); | |
649 | ||
650 | /* | |
651 | * Patch the 32 bits and 64 bits symbols. Note that we do not | |
652 | * patch the "." symbol on 64 bits. | |
653 | * It would be easy to do, but doesn't seem to be necessary, | |
654 | * patching the OPD symbol is enough. | |
655 | */ | |
656 | vdso_do_func_patch32(v32, v64, patch->gen_name, | |
657 | patch->fix_name); | |
658 | #ifdef CONFIG_PPC64 | |
659 | vdso_do_func_patch64(v32, v64, patch->gen_name, | |
660 | patch->fix_name); | |
661 | #endif /* CONFIG_PPC64 */ | |
662 | } | |
663 | ||
664 | return 0; | |
665 | } | |
666 | ||
667 | ||
668 | static __init int vdso_setup(void) | |
669 | { | |
670 | struct lib32_elfinfo v32; | |
671 | struct lib64_elfinfo v64; | |
672 | ||
673 | v32.hdr = vdso32_kbase; | |
674 | #ifdef CONFIG_PPC64 | |
675 | v64.hdr = vdso64_kbase; | |
676 | #endif | |
677 | if (vdso_do_find_sections(&v32, &v64)) | |
678 | return -1; | |
679 | ||
680 | if (vdso_fixup_datapage(&v32, &v64)) | |
681 | return -1; | |
682 | ||
0909c8c2 BH |
683 | if (vdso_fixup_features(&v32, &v64)) |
684 | return -1; | |
685 | ||
a7f290da BH |
686 | if (vdso_fixup_alt_funcs(&v32, &v64)) |
687 | return -1; | |
688 | ||
689 | vdso_setup_trampolines(&v32, &v64); | |
690 | ||
691 | return 0; | |
692 | } | |
693 | ||
694 | /* | |
695 | * Called from setup_arch to initialize the bitmap of available | |
696 | * syscalls in the systemcfg page | |
697 | */ | |
698 | static void __init vdso_setup_syscall_map(void) | |
699 | { | |
700 | unsigned int i; | |
701 | extern unsigned long *sys_call_table; | |
702 | extern unsigned long sys_ni_syscall; | |
703 | ||
704 | ||
705 | for (i = 0; i < __NR_syscalls; i++) { | |
706 | #ifdef CONFIG_PPC64 | |
707 | if (sys_call_table[i*2] != sys_ni_syscall) | |
708 | vdso_data->syscall_map_64[i >> 5] |= | |
709 | 0x80000000UL >> (i & 0x1f); | |
710 | if (sys_call_table[i*2+1] != sys_ni_syscall) | |
711 | vdso_data->syscall_map_32[i >> 5] |= | |
712 | 0x80000000UL >> (i & 0x1f); | |
713 | #else /* CONFIG_PPC64 */ | |
714 | if (sys_call_table[i] != sys_ni_syscall) | |
715 | vdso_data->syscall_map_32[i >> 5] |= | |
716 | 0x80000000UL >> (i & 0x1f); | |
717 | #endif /* CONFIG_PPC64 */ | |
718 | } | |
719 | } | |
720 | ||
721 | ||
722 | void __init vdso_init(void) | |
723 | { | |
724 | int i; | |
725 | ||
726 | #ifdef CONFIG_PPC64 | |
727 | /* | |
728 | * Fill up the "systemcfg" stuff for backward compatiblity | |
729 | */ | |
730 | strcpy(vdso_data->eye_catcher, "SYSTEMCFG:PPC64"); | |
731 | vdso_data->version.major = SYSTEMCFG_MAJOR; | |
732 | vdso_data->version.minor = SYSTEMCFG_MINOR; | |
733 | vdso_data->processor = mfspr(SPRN_PVR); | |
e8222502 BH |
734 | /* |
735 | * Fake the old platform number for pSeries and iSeries and add | |
736 | * in LPAR bit if necessary | |
737 | */ | |
738 | vdso_data->platform = machine_is(iseries) ? 0x200 : 0x100; | |
739 | if (firmware_has_feature(FW_FEATURE_LPAR)) | |
740 | vdso_data->platform |= 1; | |
a7f290da BH |
741 | vdso_data->physicalMemorySize = lmb_phys_mem_size(); |
742 | vdso_data->dcache_size = ppc64_caches.dsize; | |
743 | vdso_data->dcache_line_size = ppc64_caches.dline_size; | |
744 | vdso_data->icache_size = ppc64_caches.isize; | |
745 | vdso_data->icache_line_size = ppc64_caches.iline_size; | |
746 | ||
747 | /* | |
748 | * Calculate the size of the 64 bits vDSO | |
749 | */ | |
750 | vdso64_pages = (&vdso64_end - &vdso64_start) >> PAGE_SHIFT; | |
751 | DBG("vdso64_kbase: %p, 0x%x pages\n", vdso64_kbase, vdso64_pages); | |
752 | #endif /* CONFIG_PPC64 */ | |
753 | ||
754 | ||
755 | /* | |
756 | * Calculate the size of the 32 bits vDSO | |
757 | */ | |
758 | vdso32_pages = (&vdso32_end - &vdso32_start) >> PAGE_SHIFT; | |
759 | DBG("vdso32_kbase: %p, 0x%x pages\n", vdso32_kbase, vdso32_pages); | |
760 | ||
761 | ||
762 | /* | |
763 | * Setup the syscall map in the vDOS | |
764 | */ | |
765 | vdso_setup_syscall_map(); | |
0909c8c2 | 766 | |
a7f290da BH |
767 | /* |
768 | * Initialize the vDSO images in memory, that is do necessary | |
769 | * fixups of vDSO symbols, locate trampolines, etc... | |
770 | */ | |
771 | if (vdso_setup()) { | |
772 | printk(KERN_ERR "vDSO setup failure, not enabled !\n"); | |
773 | vdso32_pages = 0; | |
774 | #ifdef CONFIG_PPC64 | |
775 | vdso64_pages = 0; | |
776 | #endif | |
777 | return; | |
778 | } | |
779 | ||
780 | /* Make sure pages are in the correct state */ | |
781 | for (i = 0; i < vdso32_pages; i++) { | |
782 | struct page *pg = virt_to_page(vdso32_kbase + i*PAGE_SIZE); | |
783 | ClearPageReserved(pg); | |
784 | get_page(pg); | |
785 | ||
786 | } | |
787 | #ifdef CONFIG_PPC64 | |
788 | for (i = 0; i < vdso64_pages; i++) { | |
789 | struct page *pg = virt_to_page(vdso64_kbase + i*PAGE_SIZE); | |
790 | ClearPageReserved(pg); | |
791 | get_page(pg); | |
792 | } | |
793 | #endif /* CONFIG_PPC64 */ | |
794 | ||
795 | get_page(virt_to_page(vdso_data)); | |
796 | } | |
797 | ||
798 | int in_gate_area_no_task(unsigned long addr) | |
799 | { | |
800 | return 0; | |
801 | } | |
802 | ||
803 | int in_gate_area(struct task_struct *task, unsigned long addr) | |
804 | { | |
805 | return 0; | |
806 | } | |
807 | ||
808 | struct vm_area_struct *get_gate_vma(struct task_struct *tsk) | |
809 | { | |
810 | return NULL; | |
811 | } | |
812 |