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