blob: fd5b2ea5d2993bbafa4b370fee3aba3f172577c0 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * linux/fs/binfmt_elf.c
3 *
4 * These are the functions used to load ELF format executables as used
5 * on SVr4 machines. Information on the format may be found in the book
6 * "UNIX SYSTEM V RELEASE 4 Programmers Guide: Ansi C and Programming Support
7 * Tools".
8 *
9 * Copyright 1993, 1994: Eric Youngdale (ericy@cais.com).
10 */
11
12#include <linux/module.h>
13#include <linux/kernel.h>
14#include <linux/fs.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070015#include <linux/mm.h>
16#include <linux/mman.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070017#include <linux/errno.h>
18#include <linux/signal.h>
19#include <linux/binfmts.h>
20#include <linux/string.h>
21#include <linux/file.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070022#include <linux/slab.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070023#include <linux/personality.h>
24#include <linux/elfcore.h>
25#include <linux/init.h>
26#include <linux/highuid.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070027#include <linux/compiler.h>
28#include <linux/highmem.h>
29#include <linux/pagemap.h>
30#include <linux/security.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070031#include <linux/random.h>
Jesper Juhlf4e5cc22006-06-23 02:05:35 -070032#include <linux/elf.h>
Alexey Dobriyan7e80d0d2007-05-08 00:28:59 -070033#include <linux/utsname.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070034#include <asm/uaccess.h>
35#include <asm/param.h>
36#include <asm/page.h>
37
Jesper Juhlf4e5cc22006-06-23 02:05:35 -070038static int load_elf_binary(struct linux_binprm *bprm, struct pt_regs *regs);
39static int load_elf_library(struct file *);
Andrew Mortonbb1ad822008-01-30 13:31:07 +010040static unsigned long elf_map(struct file *, unsigned long, struct elf_phdr *,
41 int, int, unsigned long);
Linus Torvalds1da177e2005-04-16 15:20:36 -070042
Linus Torvalds1da177e2005-04-16 15:20:36 -070043/*
44 * If we don't support core dumping, then supply a NULL so we
45 * don't even try.
46 */
Christoph Hellwig698ba7b2009-12-15 16:47:37 -080047#ifdef CONFIG_ELF_CORE
Masami Hiramatsuf6151df2009-12-17 15:27:16 -080048static int elf_core_dump(struct coredump_params *cprm);
Linus Torvalds1da177e2005-04-16 15:20:36 -070049#else
50#define elf_core_dump NULL
51#endif
52
53#if ELF_EXEC_PAGESIZE > PAGE_SIZE
Jesper Juhlf4e5cc22006-06-23 02:05:35 -070054#define ELF_MIN_ALIGN ELF_EXEC_PAGESIZE
Linus Torvalds1da177e2005-04-16 15:20:36 -070055#else
Jesper Juhlf4e5cc22006-06-23 02:05:35 -070056#define ELF_MIN_ALIGN PAGE_SIZE
Linus Torvalds1da177e2005-04-16 15:20:36 -070057#endif
58
59#ifndef ELF_CORE_EFLAGS
60#define ELF_CORE_EFLAGS 0
61#endif
62
63#define ELF_PAGESTART(_v) ((_v) & ~(unsigned long)(ELF_MIN_ALIGN-1))
64#define ELF_PAGEOFFSET(_v) ((_v) & (ELF_MIN_ALIGN-1))
65#define ELF_PAGEALIGN(_v) (((_v) + ELF_MIN_ALIGN - 1) & ~(ELF_MIN_ALIGN - 1))
66
67static struct linux_binfmt elf_format = {
68 .module = THIS_MODULE,
69 .load_binary = load_elf_binary,
70 .load_shlib = load_elf_library,
71 .core_dump = elf_core_dump,
Andi Kleen9fbbd4d2007-02-13 13:26:26 +010072 .min_coredump = ELF_EXEC_PAGESIZE,
73 .hasvdso = 1
Linus Torvalds1da177e2005-04-16 15:20:36 -070074};
75
Andrew Mortond4e3cc32007-07-21 04:37:32 -070076#define BAD_ADDR(x) ((unsigned long)(x) >= TASK_SIZE)
Linus Torvalds1da177e2005-04-16 15:20:36 -070077
78static int set_brk(unsigned long start, unsigned long end)
79{
80 start = ELF_PAGEALIGN(start);
81 end = ELF_PAGEALIGN(end);
82 if (end > start) {
83 unsigned long addr;
84 down_write(&current->mm->mmap_sem);
85 addr = do_brk(start, end - start);
86 up_write(&current->mm->mmap_sem);
87 if (BAD_ADDR(addr))
88 return addr;
89 }
90 current->mm->start_brk = current->mm->brk = end;
91 return 0;
92}
93
Linus Torvalds1da177e2005-04-16 15:20:36 -070094/* We need to explicitly zero any fractional pages
95 after the data section (i.e. bss). This would
96 contain the junk from the file that should not
Jesper Juhlf4e5cc22006-06-23 02:05:35 -070097 be in memory
98 */
Linus Torvalds1da177e2005-04-16 15:20:36 -070099static int padzero(unsigned long elf_bss)
100{
101 unsigned long nbyte;
102
103 nbyte = ELF_PAGEOFFSET(elf_bss);
104 if (nbyte) {
105 nbyte = ELF_MIN_ALIGN - nbyte;
106 if (clear_user((void __user *) elf_bss, nbyte))
107 return -EFAULT;
108 }
109 return 0;
110}
111
Ohad Ben-Cohen09c6dd32008-02-03 18:05:15 +0200112/* Let's use some macros to make this stack manipulation a little clearer */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700113#ifdef CONFIG_STACK_GROWSUP
114#define STACK_ADD(sp, items) ((elf_addr_t __user *)(sp) + (items))
115#define STACK_ROUND(sp, items) \
116 ((15 + (unsigned long) ((sp) + (items))) &~ 15UL)
Jesper Juhlf4e5cc22006-06-23 02:05:35 -0700117#define STACK_ALLOC(sp, len) ({ \
118 elf_addr_t __user *old_sp = (elf_addr_t __user *)sp; sp += len; \
119 old_sp; })
Linus Torvalds1da177e2005-04-16 15:20:36 -0700120#else
121#define STACK_ADD(sp, items) ((elf_addr_t __user *)(sp) - (items))
122#define STACK_ROUND(sp, items) \
123 (((unsigned long) (sp - items)) &~ 15UL)
124#define STACK_ALLOC(sp, len) ({ sp -= len ; sp; })
125#endif
126
Nathan Lynch483fad12008-07-22 04:48:46 +1000127#ifndef ELF_BASE_PLATFORM
128/*
129 * AT_BASE_PLATFORM indicates the "real" hardware/microarchitecture.
130 * If the arch defines ELF_BASE_PLATFORM (in asm/elf.h), the value
131 * will be copied to the user stack in the same manner as AT_PLATFORM.
132 */
133#define ELF_BASE_PLATFORM NULL
134#endif
135
Linus Torvalds1da177e2005-04-16 15:20:36 -0700136static int
Jesper Juhlf4e5cc22006-06-23 02:05:35 -0700137create_elf_tables(struct linux_binprm *bprm, struct elfhdr *exec,
Andi Kleend20894a2008-02-08 04:21:54 -0800138 unsigned long load_addr, unsigned long interp_load_addr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700139{
140 unsigned long p = bprm->p;
141 int argc = bprm->argc;
142 int envc = bprm->envc;
143 elf_addr_t __user *argv;
144 elf_addr_t __user *envp;
145 elf_addr_t __user *sp;
146 elf_addr_t __user *u_platform;
Nathan Lynch483fad12008-07-22 04:48:46 +1000147 elf_addr_t __user *u_base_platform;
Kees Cookf06295b2009-01-07 18:08:52 -0800148 elf_addr_t __user *u_rand_bytes;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700149 const char *k_platform = ELF_PLATFORM;
Nathan Lynch483fad12008-07-22 04:48:46 +1000150 const char *k_base_platform = ELF_BASE_PLATFORM;
Kees Cookf06295b2009-01-07 18:08:52 -0800151 unsigned char k_rand_bytes[16];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700152 int items;
153 elf_addr_t *elf_info;
154 int ei_index = 0;
David Howells86a264a2008-11-14 10:39:18 +1100155 const struct cred *cred = current_cred();
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700156 struct vm_area_struct *vma;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700157
158 /*
Franck Bui-Huud68c9d62007-10-16 23:30:24 -0700159 * In some cases (e.g. Hyper-Threading), we want to avoid L1
160 * evictions by the processes running on the same package. One
161 * thing we can do is to shuffle the initial stack for them.
162 */
163
164 p = arch_align_stack(p);
165
166 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700167 * If this architecture has a platform capability string, copy it
168 * to userspace. In some cases (Sparc), this info is impossible
169 * for userspace to get any other way, in others (i386) it is
170 * merely difficult.
171 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700172 u_platform = NULL;
173 if (k_platform) {
174 size_t len = strlen(k_platform) + 1;
175
Linus Torvalds1da177e2005-04-16 15:20:36 -0700176 u_platform = (elf_addr_t __user *)STACK_ALLOC(p, len);
177 if (__copy_to_user(u_platform, k_platform, len))
178 return -EFAULT;
179 }
180
Nathan Lynch483fad12008-07-22 04:48:46 +1000181 /*
182 * If this architecture has a "base" platform capability
183 * string, copy it to userspace.
184 */
185 u_base_platform = NULL;
186 if (k_base_platform) {
187 size_t len = strlen(k_base_platform) + 1;
188
189 u_base_platform = (elf_addr_t __user *)STACK_ALLOC(p, len);
190 if (__copy_to_user(u_base_platform, k_base_platform, len))
191 return -EFAULT;
192 }
193
Kees Cookf06295b2009-01-07 18:08:52 -0800194 /*
195 * Generate 16 random bytes for userspace PRNG seeding.
196 */
197 get_random_bytes(k_rand_bytes, sizeof(k_rand_bytes));
198 u_rand_bytes = (elf_addr_t __user *)
199 STACK_ALLOC(p, sizeof(k_rand_bytes));
200 if (__copy_to_user(u_rand_bytes, k_rand_bytes, sizeof(k_rand_bytes)))
201 return -EFAULT;
202
Linus Torvalds1da177e2005-04-16 15:20:36 -0700203 /* Create the ELF interpreter info */
Jesper Juhl785d5572006-06-23 02:05:35 -0700204 elf_info = (elf_addr_t *)current->mm->saved_auxv;
Olaf Hering4f9a58d2007-10-16 23:30:12 -0700205 /* update AT_VECTOR_SIZE_BASE if the number of NEW_AUX_ENT() changes */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700206#define NEW_AUX_ENT(id, val) \
Jesper Juhlf4e5cc22006-06-23 02:05:35 -0700207 do { \
Jesper Juhl785d5572006-06-23 02:05:35 -0700208 elf_info[ei_index++] = id; \
209 elf_info[ei_index++] = val; \
Jesper Juhlf4e5cc22006-06-23 02:05:35 -0700210 } while (0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700211
212#ifdef ARCH_DLINFO
213 /*
214 * ARCH_DLINFO must come first so PPC can do its special alignment of
215 * AUXV.
Olaf Hering4f9a58d2007-10-16 23:30:12 -0700216 * update AT_VECTOR_SIZE_ARCH if the number of NEW_AUX_ENT() in
217 * ARCH_DLINFO changes
Linus Torvalds1da177e2005-04-16 15:20:36 -0700218 */
219 ARCH_DLINFO;
220#endif
221 NEW_AUX_ENT(AT_HWCAP, ELF_HWCAP);
222 NEW_AUX_ENT(AT_PAGESZ, ELF_EXEC_PAGESIZE);
223 NEW_AUX_ENT(AT_CLKTCK, CLOCKS_PER_SEC);
224 NEW_AUX_ENT(AT_PHDR, load_addr + exec->e_phoff);
Jesper Juhlf4e5cc22006-06-23 02:05:35 -0700225 NEW_AUX_ENT(AT_PHENT, sizeof(struct elf_phdr));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700226 NEW_AUX_ENT(AT_PHNUM, exec->e_phnum);
227 NEW_AUX_ENT(AT_BASE, interp_load_addr);
228 NEW_AUX_ENT(AT_FLAGS, 0);
229 NEW_AUX_ENT(AT_ENTRY, exec->e_entry);
David Howells86a264a2008-11-14 10:39:18 +1100230 NEW_AUX_ENT(AT_UID, cred->uid);
231 NEW_AUX_ENT(AT_EUID, cred->euid);
232 NEW_AUX_ENT(AT_GID, cred->gid);
233 NEW_AUX_ENT(AT_EGID, cred->egid);
Jesper Juhl785d5572006-06-23 02:05:35 -0700234 NEW_AUX_ENT(AT_SECURE, security_bprm_secureexec(bprm));
Kees Cookf06295b2009-01-07 18:08:52 -0800235 NEW_AUX_ENT(AT_RANDOM, (elf_addr_t)(unsigned long)u_rand_bytes);
John Reiser65191082008-07-21 14:21:32 -0700236 NEW_AUX_ENT(AT_EXECFN, bprm->exec);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700237 if (k_platform) {
Jesper Juhlf4e5cc22006-06-23 02:05:35 -0700238 NEW_AUX_ENT(AT_PLATFORM,
Jesper Juhl785d5572006-06-23 02:05:35 -0700239 (elf_addr_t)(unsigned long)u_platform);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700240 }
Nathan Lynch483fad12008-07-22 04:48:46 +1000241 if (k_base_platform) {
242 NEW_AUX_ENT(AT_BASE_PLATFORM,
243 (elf_addr_t)(unsigned long)u_base_platform);
244 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700245 if (bprm->interp_flags & BINPRM_FLAGS_EXECFD) {
Jesper Juhl785d5572006-06-23 02:05:35 -0700246 NEW_AUX_ENT(AT_EXECFD, bprm->interp_data);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700247 }
248#undef NEW_AUX_ENT
249 /* AT_NULL is zero; clear the rest too */
250 memset(&elf_info[ei_index], 0,
251 sizeof current->mm->saved_auxv - ei_index * sizeof elf_info[0]);
252
253 /* And advance past the AT_NULL entry. */
254 ei_index += 2;
255
256 sp = STACK_ADD(p, ei_index);
257
Andi Kleend20894a2008-02-08 04:21:54 -0800258 items = (argc + 1) + (envc + 1) + 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700259 bprm->p = STACK_ROUND(sp, items);
260
261 /* Point sp at the lowest address on the stack */
262#ifdef CONFIG_STACK_GROWSUP
263 sp = (elf_addr_t __user *)bprm->p - items - ei_index;
Jesper Juhlf4e5cc22006-06-23 02:05:35 -0700264 bprm->exec = (unsigned long)sp; /* XXX: PARISC HACK */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700265#else
266 sp = (elf_addr_t __user *)bprm->p;
267#endif
268
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700269
270 /*
271 * Grow the stack manually; some architectures have a limit on how
272 * far ahead a user-space access may be in order to grow the stack.
273 */
274 vma = find_extend_vma(current->mm, bprm->p);
275 if (!vma)
276 return -EFAULT;
277
Linus Torvalds1da177e2005-04-16 15:20:36 -0700278 /* Now, let's put argc (and argv, envp if appropriate) on the stack */
279 if (__put_user(argc, sp++))
280 return -EFAULT;
Andi Kleend20894a2008-02-08 04:21:54 -0800281 argv = sp;
282 envp = argv + argc + 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700283
284 /* Populate argv and envp */
Greg Kroah-Hartmana84a5052005-05-11 00:10:44 -0700285 p = current->mm->arg_end = current->mm->arg_start;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700286 while (argc-- > 0) {
287 size_t len;
Heiko Carstens841d5fb2006-12-06 20:36:35 -0800288 if (__put_user((elf_addr_t)p, argv++))
289 return -EFAULT;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700290 len = strnlen_user((void __user *)p, MAX_ARG_STRLEN);
291 if (!len || len > MAX_ARG_STRLEN)
WANG Cong23c49712008-05-08 21:52:33 +0800292 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700293 p += len;
294 }
295 if (__put_user(0, argv))
296 return -EFAULT;
297 current->mm->arg_end = current->mm->env_start = p;
298 while (envc-- > 0) {
299 size_t len;
Heiko Carstens841d5fb2006-12-06 20:36:35 -0800300 if (__put_user((elf_addr_t)p, envp++))
301 return -EFAULT;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700302 len = strnlen_user((void __user *)p, MAX_ARG_STRLEN);
303 if (!len || len > MAX_ARG_STRLEN)
WANG Cong23c49712008-05-08 21:52:33 +0800304 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700305 p += len;
306 }
307 if (__put_user(0, envp))
308 return -EFAULT;
309 current->mm->env_end = p;
310
311 /* Put the elf_info on the stack in the right place. */
312 sp = (elf_addr_t __user *)envp + 1;
313 if (copy_to_user(sp, elf_info, ei_index * sizeof(elf_addr_t)))
314 return -EFAULT;
315 return 0;
316}
317
318#ifndef elf_map
319
320static unsigned long elf_map(struct file *filep, unsigned long addr,
Jiri Kosinacc503c12008-01-30 13:31:07 +0100321 struct elf_phdr *eppnt, int prot, int type,
322 unsigned long total_size)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700323{
324 unsigned long map_addr;
Jiri Kosinacc503c12008-01-30 13:31:07 +0100325 unsigned long size = eppnt->p_filesz + ELF_PAGEOFFSET(eppnt->p_vaddr);
326 unsigned long off = eppnt->p_offset - ELF_PAGEOFFSET(eppnt->p_vaddr);
327 addr = ELF_PAGESTART(addr);
328 size = ELF_PAGEALIGN(size);
Jan Kratochvil60bfba72007-07-15 23:40:06 -0700329
Andrew Mortond4e3cc32007-07-21 04:37:32 -0700330 /* mmap() will return -EINVAL if given a zero size, but a
331 * segment with zero filesize is perfectly valid */
Jiri Kosinacc503c12008-01-30 13:31:07 +0100332 if (!size)
333 return addr;
334
335 down_write(&current->mm->mmap_sem);
336 /*
337 * total_size is the size of the ELF (interpreter) image.
338 * The _first_ mmap needs to know the full size, otherwise
339 * randomization might put this image into an overlapping
340 * position with the ELF binary image. (since size < total_size)
341 * So we first map the 'big' image - and unmap the remainder at
342 * the end. (which unmap is needed for ELF images with holes.)
343 */
344 if (total_size) {
345 total_size = ELF_PAGEALIGN(total_size);
346 map_addr = do_mmap(filep, addr, total_size, prot, type, off);
347 if (!BAD_ADDR(map_addr))
348 do_munmap(current->mm, map_addr+size, total_size-size);
349 } else
350 map_addr = do_mmap(filep, addr, size, prot, type, off);
351
Linus Torvalds1da177e2005-04-16 15:20:36 -0700352 up_write(&current->mm->mmap_sem);
353 return(map_addr);
354}
355
356#endif /* !elf_map */
357
Jiri Kosinacc503c12008-01-30 13:31:07 +0100358static unsigned long total_mapping_size(struct elf_phdr *cmds, int nr)
359{
360 int i, first_idx = -1, last_idx = -1;
361
362 for (i = 0; i < nr; i++) {
363 if (cmds[i].p_type == PT_LOAD) {
364 last_idx = i;
365 if (first_idx == -1)
366 first_idx = i;
367 }
368 }
369 if (first_idx == -1)
370 return 0;
371
372 return cmds[last_idx].p_vaddr + cmds[last_idx].p_memsz -
373 ELF_PAGESTART(cmds[first_idx].p_vaddr);
374}
375
376
Linus Torvalds1da177e2005-04-16 15:20:36 -0700377/* This is much more generalized than the library routine read function,
378 so we keep this separate. Technically the library read function
379 is only provided so that we can read a.out libraries that have
380 an ELF header */
381
Jesper Juhlf4e5cc22006-06-23 02:05:35 -0700382static unsigned long load_elf_interp(struct elfhdr *interp_elf_ex,
Jiri Kosinacc503c12008-01-30 13:31:07 +0100383 struct file *interpreter, unsigned long *interp_map_addr,
384 unsigned long no_base)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700385{
386 struct elf_phdr *elf_phdata;
387 struct elf_phdr *eppnt;
388 unsigned long load_addr = 0;
389 int load_addr_set = 0;
390 unsigned long last_bss = 0, elf_bss = 0;
391 unsigned long error = ~0UL;
Jiri Kosinacc503c12008-01-30 13:31:07 +0100392 unsigned long total_size;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700393 int retval, i, size;
394
395 /* First of all, some simple consistency checks */
396 if (interp_elf_ex->e_type != ET_EXEC &&
397 interp_elf_ex->e_type != ET_DYN)
398 goto out;
399 if (!elf_check_arch(interp_elf_ex))
400 goto out;
401 if (!interpreter->f_op || !interpreter->f_op->mmap)
402 goto out;
403
404 /*
405 * If the size of this structure has changed, then punt, since
406 * we will be doing the wrong thing.
407 */
408 if (interp_elf_ex->e_phentsize != sizeof(struct elf_phdr))
409 goto out;
410 if (interp_elf_ex->e_phnum < 1 ||
411 interp_elf_ex->e_phnum > 65536U / sizeof(struct elf_phdr))
412 goto out;
413
414 /* Now read in all of the header information */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700415 size = sizeof(struct elf_phdr) * interp_elf_ex->e_phnum;
416 if (size > ELF_MIN_ALIGN)
417 goto out;
Jesper Juhlf4e5cc22006-06-23 02:05:35 -0700418 elf_phdata = kmalloc(size, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700419 if (!elf_phdata)
420 goto out;
421
Jesper Juhlf4e5cc22006-06-23 02:05:35 -0700422 retval = kernel_read(interpreter, interp_elf_ex->e_phoff,
423 (char *)elf_phdata,size);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700424 error = -EIO;
425 if (retval != size) {
426 if (retval < 0)
427 error = retval;
428 goto out_close;
429 }
430
Jiri Kosinacc503c12008-01-30 13:31:07 +0100431 total_size = total_mapping_size(elf_phdata, interp_elf_ex->e_phnum);
432 if (!total_size) {
433 error = -EINVAL;
434 goto out_close;
435 }
436
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437 eppnt = elf_phdata;
Jesper Juhlf4e5cc22006-06-23 02:05:35 -0700438 for (i = 0; i < interp_elf_ex->e_phnum; i++, eppnt++) {
439 if (eppnt->p_type == PT_LOAD) {
440 int elf_type = MAP_PRIVATE | MAP_DENYWRITE;
441 int elf_prot = 0;
442 unsigned long vaddr = 0;
443 unsigned long k, map_addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700444
Jesper Juhlf4e5cc22006-06-23 02:05:35 -0700445 if (eppnt->p_flags & PF_R)
446 elf_prot = PROT_READ;
447 if (eppnt->p_flags & PF_W)
448 elf_prot |= PROT_WRITE;
449 if (eppnt->p_flags & PF_X)
450 elf_prot |= PROT_EXEC;
451 vaddr = eppnt->p_vaddr;
452 if (interp_elf_ex->e_type == ET_EXEC || load_addr_set)
453 elf_type |= MAP_FIXED;
Jiri Kosinacc503c12008-01-30 13:31:07 +0100454 else if (no_base && interp_elf_ex->e_type == ET_DYN)
455 load_addr = -vaddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700456
Jesper Juhlf4e5cc22006-06-23 02:05:35 -0700457 map_addr = elf_map(interpreter, load_addr + vaddr,
Andrew Mortonbb1ad822008-01-30 13:31:07 +0100458 eppnt, elf_prot, elf_type, total_size);
Jiri Kosinacc503c12008-01-30 13:31:07 +0100459 total_size = 0;
460 if (!*interp_map_addr)
461 *interp_map_addr = map_addr;
Jesper Juhlf4e5cc22006-06-23 02:05:35 -0700462 error = map_addr;
463 if (BAD_ADDR(map_addr))
464 goto out_close;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700465
Jesper Juhlf4e5cc22006-06-23 02:05:35 -0700466 if (!load_addr_set &&
467 interp_elf_ex->e_type == ET_DYN) {
468 load_addr = map_addr - ELF_PAGESTART(vaddr);
469 load_addr_set = 1;
470 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700471
Jesper Juhlf4e5cc22006-06-23 02:05:35 -0700472 /*
473 * Check to see if the section's size will overflow the
474 * allowed task size. Note that p_filesz must always be
475 * <= p_memsize so it's only necessary to check p_memsz.
476 */
477 k = load_addr + eppnt->p_vaddr;
Chuck Ebbertce510592006-07-03 00:24:14 -0700478 if (BAD_ADDR(k) ||
Jesper Juhlf4e5cc22006-06-23 02:05:35 -0700479 eppnt->p_filesz > eppnt->p_memsz ||
480 eppnt->p_memsz > TASK_SIZE ||
481 TASK_SIZE - eppnt->p_memsz < k) {
482 error = -ENOMEM;
483 goto out_close;
484 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700485
Jesper Juhlf4e5cc22006-06-23 02:05:35 -0700486 /*
487 * Find the end of the file mapping for this phdr, and
488 * keep track of the largest address we see for this.
489 */
490 k = load_addr + eppnt->p_vaddr + eppnt->p_filesz;
491 if (k > elf_bss)
492 elf_bss = k;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700493
Jesper Juhlf4e5cc22006-06-23 02:05:35 -0700494 /*
495 * Do the same thing for the memory mapping - between
496 * elf_bss and last_bss is the bss section.
497 */
498 k = load_addr + eppnt->p_memsz + eppnt->p_vaddr;
499 if (k > last_bss)
500 last_bss = k;
501 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700502 }
503
Linus Torvalds1da177e2005-04-16 15:20:36 -0700504 if (last_bss > elf_bss) {
Roland McGrath752015d2009-09-08 19:49:40 -0700505 /*
506 * Now fill out the bss section. First pad the last page up
507 * to the page boundary, and then perform a mmap to make sure
508 * that there are zero-mapped pages up to and including the
509 * last bss page.
510 */
511 if (padzero(elf_bss)) {
512 error = -EFAULT;
513 goto out_close;
514 }
515
516 /* What we have mapped so far */
517 elf_bss = ELF_PAGESTART(elf_bss + ELF_MIN_ALIGN - 1);
518
519 /* Map the last of the bss segment */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700520 down_write(&current->mm->mmap_sem);
521 error = do_brk(elf_bss, last_bss - elf_bss);
522 up_write(&current->mm->mmap_sem);
523 if (BAD_ADDR(error))
524 goto out_close;
525 }
526
Jiri Kosinacc503c12008-01-30 13:31:07 +0100527 error = load_addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700528
529out_close:
530 kfree(elf_phdata);
531out:
532 return error;
533}
534
Linus Torvalds1da177e2005-04-16 15:20:36 -0700535/*
536 * These are the functions used to load ELF style executables and shared
537 * libraries. There is no binary dependent code anywhere else.
538 */
539
540#define INTERPRETER_NONE 0
Linus Torvalds1da177e2005-04-16 15:20:36 -0700541#define INTERPRETER_ELF 2
542
Andi Kleen913bd902006-03-25 16:29:09 +0100543#ifndef STACK_RND_MASK
James Bottomleyd1cabd62007-03-16 13:38:35 -0800544#define STACK_RND_MASK (0x7ff >> (PAGE_SHIFT - 12)) /* 8MB of VA */
Andi Kleen913bd902006-03-25 16:29:09 +0100545#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700546
547static unsigned long randomize_stack_top(unsigned long stack_top)
548{
549 unsigned int random_variable = 0;
550
Andi Kleenc16b63e2006-09-26 10:52:28 +0200551 if ((current->flags & PF_RANDOMIZE) &&
552 !(current->personality & ADDR_NO_RANDOMIZE)) {
Andi Kleen913bd902006-03-25 16:29:09 +0100553 random_variable = get_random_int() & STACK_RND_MASK;
554 random_variable <<= PAGE_SHIFT;
555 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700556#ifdef CONFIG_STACK_GROWSUP
Andi Kleen913bd902006-03-25 16:29:09 +0100557 return PAGE_ALIGN(stack_top) + random_variable;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700558#else
Andi Kleen913bd902006-03-25 16:29:09 +0100559 return PAGE_ALIGN(stack_top) - random_variable;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700560#endif
561}
562
Jesper Juhlf4e5cc22006-06-23 02:05:35 -0700563static int load_elf_binary(struct linux_binprm *bprm, struct pt_regs *regs)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700564{
565 struct file *interpreter = NULL; /* to shut gcc up */
566 unsigned long load_addr = 0, load_bias = 0;
567 int load_addr_set = 0;
568 char * elf_interpreter = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700569 unsigned long error;
Jesper Juhlf4e5cc22006-06-23 02:05:35 -0700570 struct elf_phdr *elf_ppnt, *elf_phdata;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700571 unsigned long elf_bss, elf_brk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700572 int retval, i;
573 unsigned int size;
Jiri Kosinacc503c12008-01-30 13:31:07 +0100574 unsigned long elf_entry;
575 unsigned long interp_load_addr = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700576 unsigned long start_code, end_code, start_data, end_data;
577 unsigned long reloc_func_desc = 0;
David Rientjes8de61e62006-12-06 20:40:16 -0800578 int executable_stack = EXSTACK_DEFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700579 unsigned long def_flags = 0;
580 struct {
581 struct elfhdr elf_ex;
582 struct elfhdr interp_elf_ex;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700583 } *loc;
584
585 loc = kmalloc(sizeof(*loc), GFP_KERNEL);
586 if (!loc) {
587 retval = -ENOMEM;
588 goto out_ret;
589 }
590
591 /* Get the exec-header */
Jesper Juhlf4e5cc22006-06-23 02:05:35 -0700592 loc->elf_ex = *((struct elfhdr *)bprm->buf);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700593
594 retval = -ENOEXEC;
595 /* First of all, some simple consistency checks */
596 if (memcmp(loc->elf_ex.e_ident, ELFMAG, SELFMAG) != 0)
597 goto out;
598
599 if (loc->elf_ex.e_type != ET_EXEC && loc->elf_ex.e_type != ET_DYN)
600 goto out;
601 if (!elf_check_arch(&loc->elf_ex))
602 goto out;
603 if (!bprm->file->f_op||!bprm->file->f_op->mmap)
604 goto out;
605
606 /* Now read in all of the header information */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700607 if (loc->elf_ex.e_phentsize != sizeof(struct elf_phdr))
608 goto out;
609 if (loc->elf_ex.e_phnum < 1 ||
610 loc->elf_ex.e_phnum > 65536U / sizeof(struct elf_phdr))
611 goto out;
612 size = loc->elf_ex.e_phnum * sizeof(struct elf_phdr);
613 retval = -ENOMEM;
Jesper Juhlf4e5cc22006-06-23 02:05:35 -0700614 elf_phdata = kmalloc(size, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700615 if (!elf_phdata)
616 goto out;
617
Jesper Juhlf4e5cc22006-06-23 02:05:35 -0700618 retval = kernel_read(bprm->file, loc->elf_ex.e_phoff,
619 (char *)elf_phdata, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700620 if (retval != size) {
621 if (retval >= 0)
622 retval = -EIO;
623 goto out_free_ph;
624 }
625
Linus Torvalds1da177e2005-04-16 15:20:36 -0700626 elf_ppnt = elf_phdata;
627 elf_bss = 0;
628 elf_brk = 0;
629
630 start_code = ~0UL;
631 end_code = 0;
632 start_data = 0;
633 end_data = 0;
634
635 for (i = 0; i < loc->elf_ex.e_phnum; i++) {
636 if (elf_ppnt->p_type == PT_INTERP) {
637 /* This is the program interpreter used for
638 * shared libraries - for now assume that this
639 * is an a.out format binary
640 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700641 retval = -ENOEXEC;
642 if (elf_ppnt->p_filesz > PATH_MAX ||
643 elf_ppnt->p_filesz < 2)
Al Viroe7b9b552009-03-29 16:31:16 -0400644 goto out_free_ph;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700645
646 retval = -ENOMEM;
Jesper Juhl792db3a2006-01-09 20:54:45 -0800647 elf_interpreter = kmalloc(elf_ppnt->p_filesz,
Jesper Juhlf4e5cc22006-06-23 02:05:35 -0700648 GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700649 if (!elf_interpreter)
Al Viroe7b9b552009-03-29 16:31:16 -0400650 goto out_free_ph;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700651
652 retval = kernel_read(bprm->file, elf_ppnt->p_offset,
Jesper Juhlf4e5cc22006-06-23 02:05:35 -0700653 elf_interpreter,
654 elf_ppnt->p_filesz);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700655 if (retval != elf_ppnt->p_filesz) {
656 if (retval >= 0)
657 retval = -EIO;
658 goto out_free_interp;
659 }
660 /* make sure path is NULL terminated */
661 retval = -ENOEXEC;
662 if (elf_interpreter[elf_ppnt->p_filesz - 1] != '\0')
663 goto out_free_interp;
664
Linus Torvalds1da177e2005-04-16 15:20:36 -0700665 interpreter = open_exec(elf_interpreter);
666 retval = PTR_ERR(interpreter);
667 if (IS_ERR(interpreter))
668 goto out_free_interp;
Alexey Dobriyan1fb84492007-01-26 00:57:16 -0800669
670 /*
671 * If the binary is not readable then enforce
672 * mm->dumpable = 0 regardless of the interpreter's
673 * permissions.
674 */
675 if (file_permission(interpreter, MAY_READ) < 0)
676 bprm->interp_flags |= BINPRM_FLAGS_ENFORCE_NONDUMP;
677
Jesper Juhlf4e5cc22006-06-23 02:05:35 -0700678 retval = kernel_read(interpreter, 0, bprm->buf,
679 BINPRM_BUF_SIZE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700680 if (retval != BINPRM_BUF_SIZE) {
681 if (retval >= 0)
682 retval = -EIO;
683 goto out_free_dentry;
684 }
685
686 /* Get the exec headers */
Jesper Juhlf4e5cc22006-06-23 02:05:35 -0700687 loc->interp_elf_ex = *((struct elfhdr *)bprm->buf);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700688 break;
689 }
690 elf_ppnt++;
691 }
692
693 elf_ppnt = elf_phdata;
694 for (i = 0; i < loc->elf_ex.e_phnum; i++, elf_ppnt++)
695 if (elf_ppnt->p_type == PT_GNU_STACK) {
696 if (elf_ppnt->p_flags & PF_X)
697 executable_stack = EXSTACK_ENABLE_X;
698 else
699 executable_stack = EXSTACK_DISABLE_X;
700 break;
701 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700702
703 /* Some simple consistency checks for the interpreter */
704 if (elf_interpreter) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700705 retval = -ELIBBAD;
Andi Kleend20894a2008-02-08 04:21:54 -0800706 /* Not an ELF interpreter */
707 if (memcmp(loc->interp_elf_ex.e_ident, ELFMAG, SELFMAG) != 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700708 goto out_free_dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700709 /* Verify the interpreter has a valid arch */
Andi Kleend20894a2008-02-08 04:21:54 -0800710 if (!elf_check_arch(&loc->interp_elf_ex))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700711 goto out_free_dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700712 }
713
Linus Torvalds1da177e2005-04-16 15:20:36 -0700714 /* Flush all traces of the currently running executable */
715 retval = flush_old_exec(bprm);
716 if (retval)
717 goto out_free_dentry;
718
Linus Torvalds1da177e2005-04-16 15:20:36 -0700719 /* OK, This is the point of no return */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700720 current->flags &= ~PF_FORKNOEXEC;
721 current->mm->def_flags = def_flags;
722
723 /* Do this immediately, since STACK_TOP as used in setup_arg_pages
724 may depend on the personality. */
Martin Schwidefsky0b592682008-10-16 15:39:57 +0200725 SET_PERSONALITY(loc->elf_ex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700726 if (elf_read_implies_exec(loc->elf_ex, executable_stack))
727 current->personality |= READ_IMPLIES_EXEC;
728
Jesper Juhlf4e5cc22006-06-23 02:05:35 -0700729 if (!(current->personality & ADDR_NO_RANDOMIZE) && randomize_va_space)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700730 current->flags |= PF_RANDOMIZE;
Linus Torvalds221af7f2010-01-28 22:14:42 -0800731
732 setup_new_exec(bprm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700733
734 /* Do this so that we can load the interpreter, if need be. We will
735 change some of these later */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700736 current->mm->free_area_cache = current->mm->mmap_base;
Wolfgang Wander1363c3c2005-06-21 17:14:49 -0700737 current->mm->cached_hole_size = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700738 retval = setup_arg_pages(bprm, randomize_stack_top(STACK_TOP),
739 executable_stack);
740 if (retval < 0) {
741 send_sig(SIGKILL, current, 0);
742 goto out_free_dentry;
743 }
744
Linus Torvalds1da177e2005-04-16 15:20:36 -0700745 current->mm->start_stack = bprm->p;
746
André Goddard Rosaaf901ca2009-11-14 13:09:05 -0200747 /* Now we do a little grungy work by mmapping the ELF image into
Jiri Kosinacc503c12008-01-30 13:31:07 +0100748 the correct location in memory. */
Jesper Juhlf4e5cc22006-06-23 02:05:35 -0700749 for(i = 0, elf_ppnt = elf_phdata;
750 i < loc->elf_ex.e_phnum; i++, elf_ppnt++) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700751 int elf_prot = 0, elf_flags;
752 unsigned long k, vaddr;
753
754 if (elf_ppnt->p_type != PT_LOAD)
755 continue;
756
757 if (unlikely (elf_brk > elf_bss)) {
758 unsigned long nbyte;
759
760 /* There was a PT_LOAD segment with p_memsz > p_filesz
761 before this one. Map anonymous pages, if needed,
762 and clear the area. */
763 retval = set_brk (elf_bss + load_bias,
764 elf_brk + load_bias);
765 if (retval) {
766 send_sig(SIGKILL, current, 0);
767 goto out_free_dentry;
768 }
769 nbyte = ELF_PAGEOFFSET(elf_bss);
770 if (nbyte) {
771 nbyte = ELF_MIN_ALIGN - nbyte;
772 if (nbyte > elf_brk - elf_bss)
773 nbyte = elf_brk - elf_bss;
774 if (clear_user((void __user *)elf_bss +
775 load_bias, nbyte)) {
776 /*
777 * This bss-zeroing can fail if the ELF
Jesper Juhlf4e5cc22006-06-23 02:05:35 -0700778 * file specifies odd protections. So
Linus Torvalds1da177e2005-04-16 15:20:36 -0700779 * we don't check the return value
780 */
781 }
782 }
783 }
784
Jesper Juhlf4e5cc22006-06-23 02:05:35 -0700785 if (elf_ppnt->p_flags & PF_R)
786 elf_prot |= PROT_READ;
787 if (elf_ppnt->p_flags & PF_W)
788 elf_prot |= PROT_WRITE;
789 if (elf_ppnt->p_flags & PF_X)
790 elf_prot |= PROT_EXEC;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700791
Jesper Juhlf4e5cc22006-06-23 02:05:35 -0700792 elf_flags = MAP_PRIVATE | MAP_DENYWRITE | MAP_EXECUTABLE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700793
794 vaddr = elf_ppnt->p_vaddr;
795 if (loc->elf_ex.e_type == ET_EXEC || load_addr_set) {
796 elf_flags |= MAP_FIXED;
797 } else if (loc->elf_ex.e_type == ET_DYN) {
Jesper Juhlf4e5cc22006-06-23 02:05:35 -0700798 /* Try and get dynamic programs out of the way of the
799 * default mmap base, as well as whatever program they
800 * might try to exec. This is because the brk will
801 * follow the loader, and is not movable. */
Jiri Kosinacc503c12008-01-30 13:31:07 +0100802#ifdef CONFIG_X86
803 load_bias = 0;
804#else
Linus Torvalds90cb28e2007-01-06 13:28:21 -0800805 load_bias = ELF_PAGESTART(ELF_ET_DYN_BASE - vaddr);
Jiri Kosinacc503c12008-01-30 13:31:07 +0100806#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700807 }
808
Jesper Juhlf4e5cc22006-06-23 02:05:35 -0700809 error = elf_map(bprm->file, load_bias + vaddr, elf_ppnt,
Andrew Mortonbb1ad822008-01-30 13:31:07 +0100810 elf_prot, elf_flags, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700811 if (BAD_ADDR(error)) {
812 send_sig(SIGKILL, current, 0);
Alexey Kuznetsovb140f252007-05-08 00:31:57 -0700813 retval = IS_ERR((void *)error) ?
814 PTR_ERR((void*)error) : -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700815 goto out_free_dentry;
816 }
817
818 if (!load_addr_set) {
819 load_addr_set = 1;
820 load_addr = (elf_ppnt->p_vaddr - elf_ppnt->p_offset);
821 if (loc->elf_ex.e_type == ET_DYN) {
822 load_bias += error -
823 ELF_PAGESTART(load_bias + vaddr);
824 load_addr += load_bias;
825 reloc_func_desc = load_bias;
826 }
827 }
828 k = elf_ppnt->p_vaddr;
Jesper Juhlf4e5cc22006-06-23 02:05:35 -0700829 if (k < start_code)
830 start_code = k;
831 if (start_data < k)
832 start_data = k;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700833
834 /*
835 * Check to see if the section's size will overflow the
836 * allowed task size. Note that p_filesz must always be
837 * <= p_memsz so it is only necessary to check p_memsz.
838 */
Chuck Ebbertce510592006-07-03 00:24:14 -0700839 if (BAD_ADDR(k) || elf_ppnt->p_filesz > elf_ppnt->p_memsz ||
Linus Torvalds1da177e2005-04-16 15:20:36 -0700840 elf_ppnt->p_memsz > TASK_SIZE ||
841 TASK_SIZE - elf_ppnt->p_memsz < k) {
Jesper Juhlf4e5cc22006-06-23 02:05:35 -0700842 /* set_brk can never work. Avoid overflows. */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700843 send_sig(SIGKILL, current, 0);
Alexey Kuznetsovb140f252007-05-08 00:31:57 -0700844 retval = -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700845 goto out_free_dentry;
846 }
847
848 k = elf_ppnt->p_vaddr + elf_ppnt->p_filesz;
849
850 if (k > elf_bss)
851 elf_bss = k;
852 if ((elf_ppnt->p_flags & PF_X) && end_code < k)
853 end_code = k;
854 if (end_data < k)
855 end_data = k;
856 k = elf_ppnt->p_vaddr + elf_ppnt->p_memsz;
857 if (k > elf_brk)
858 elf_brk = k;
859 }
860
861 loc->elf_ex.e_entry += load_bias;
862 elf_bss += load_bias;
863 elf_brk += load_bias;
864 start_code += load_bias;
865 end_code += load_bias;
866 start_data += load_bias;
867 end_data += load_bias;
868
869 /* Calling set_brk effectively mmaps the pages that we need
870 * for the bss and break sections. We must do this before
871 * mapping in the interpreter, to make sure it doesn't wind
872 * up getting placed where the bss needs to go.
873 */
874 retval = set_brk(elf_bss, elf_brk);
875 if (retval) {
876 send_sig(SIGKILL, current, 0);
877 goto out_free_dentry;
878 }
akpm@osdl.org6de50512005-10-11 08:29:08 -0700879 if (likely(elf_bss != elf_brk) && unlikely(padzero(elf_bss))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700880 send_sig(SIGSEGV, current, 0);
881 retval = -EFAULT; /* Nobody gets to see this, but.. */
882 goto out_free_dentry;
883 }
884
885 if (elf_interpreter) {
Andi Kleend20894a2008-02-08 04:21:54 -0800886 unsigned long uninitialized_var(interp_map_addr);
Jiri Kosinacc503c12008-01-30 13:31:07 +0100887
Andi Kleend20894a2008-02-08 04:21:54 -0800888 elf_entry = load_elf_interp(&loc->interp_elf_ex,
889 interpreter,
890 &interp_map_addr,
891 load_bias);
892 if (!IS_ERR((void *)elf_entry)) {
893 /*
894 * load_elf_interp() returns relocation
895 * adjustment
896 */
897 interp_load_addr = elf_entry;
898 elf_entry += loc->interp_elf_ex.e_entry;
Jiri Kosinacc503c12008-01-30 13:31:07 +0100899 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700900 if (BAD_ADDR(elf_entry)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700901 force_sig(SIGSEGV, current);
Chuck Ebbertce510592006-07-03 00:24:14 -0700902 retval = IS_ERR((void *)elf_entry) ?
903 (int)elf_entry : -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700904 goto out_free_dentry;
905 }
906 reloc_func_desc = interp_load_addr;
907
908 allow_write_access(interpreter);
909 fput(interpreter);
910 kfree(elf_interpreter);
911 } else {
912 elf_entry = loc->elf_ex.e_entry;
Suresh Siddha5342fba2006-02-26 04:18:28 +0100913 if (BAD_ADDR(elf_entry)) {
Chuck Ebbertce510592006-07-03 00:24:14 -0700914 force_sig(SIGSEGV, current);
915 retval = -EINVAL;
Suresh Siddha5342fba2006-02-26 04:18:28 +0100916 goto out_free_dentry;
917 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700918 }
919
920 kfree(elf_phdata);
921
Linus Torvalds1da177e2005-04-16 15:20:36 -0700922 set_binfmt(&elf_format);
923
Benjamin Herrenschmidt547ee842005-04-16 15:24:35 -0700924#ifdef ARCH_HAS_SETUP_ADDITIONAL_PAGES
Martin Schwidefskyfc5243d2008-12-25 13:38:35 +0100925 retval = arch_setup_additional_pages(bprm, !!elf_interpreter);
Benjamin Herrenschmidt547ee842005-04-16 15:24:35 -0700926 if (retval < 0) {
927 send_sig(SIGKILL, current, 0);
Roland McGrath18c8baf2005-04-28 15:17:19 -0700928 goto out;
Benjamin Herrenschmidt547ee842005-04-16 15:24:35 -0700929 }
930#endif /* ARCH_HAS_SETUP_ADDITIONAL_PAGES */
931
David Howellsa6f76f22008-11-14 10:39:24 +1100932 install_exec_creds(bprm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700933 current->flags &= ~PF_FORKNOEXEC;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700934 retval = create_elf_tables(bprm, &loc->elf_ex,
Jesper Juhlf4e5cc22006-06-23 02:05:35 -0700935 load_addr, interp_load_addr);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700936 if (retval < 0) {
937 send_sig(SIGKILL, current, 0);
938 goto out;
939 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700940 /* N.B. passed_fileno might not be initialized? */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700941 current->mm->end_code = end_code;
942 current->mm->start_code = start_code;
943 current->mm->start_data = start_data;
944 current->mm->end_data = end_data;
945 current->mm->start_stack = bprm->p;
946
Jiri Kosinac1d171a2008-01-30 13:30:40 +0100947#ifdef arch_randomize_brk
Ingo Molnar32a93232008-02-06 22:39:44 +0100948 if ((current->flags & PF_RANDOMIZE) && (randomize_va_space > 1))
Jiri Kosinac1d171a2008-01-30 13:30:40 +0100949 current->mm->brk = current->mm->start_brk =
950 arch_randomize_brk(current->mm);
951#endif
952
Linus Torvalds1da177e2005-04-16 15:20:36 -0700953 if (current->personality & MMAP_PAGE_ZERO) {
954 /* Why this, you ask??? Well SVr4 maps page 0 as read-only,
955 and some applications "depend" upon this behavior.
956 Since we do not have the power to recompile these, we
Jesper Juhlf4e5cc22006-06-23 02:05:35 -0700957 emulate the SVr4 behavior. Sigh. */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700958 down_write(&current->mm->mmap_sem);
959 error = do_mmap(NULL, 0, PAGE_SIZE, PROT_READ | PROT_EXEC,
960 MAP_FIXED | MAP_PRIVATE, 0);
961 up_write(&current->mm->mmap_sem);
962 }
963
964#ifdef ELF_PLAT_INIT
965 /*
966 * The ABI may specify that certain registers be set up in special
967 * ways (on i386 %edx is the address of a DT_FINI function, for
968 * example. In addition, it may also specify (eg, PowerPC64 ELF)
969 * that the e_entry field is the address of the function descriptor
970 * for the startup routine, rather than the address of the startup
971 * routine itself. This macro performs whatever initialization to
972 * the regs structure is required as well as any relocations to the
973 * function descriptor entries when executing dynamically links apps.
974 */
975 ELF_PLAT_INIT(regs, reloc_func_desc);
976#endif
977
978 start_thread(regs, elf_entry, bprm->p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700979 retval = 0;
980out:
981 kfree(loc);
982out_ret:
983 return retval;
984
985 /* error cleanup */
986out_free_dentry:
987 allow_write_access(interpreter);
988 if (interpreter)
989 fput(interpreter);
990out_free_interp:
Jesper Juhlf99d49a2005-11-07 01:01:34 -0800991 kfree(elf_interpreter);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700992out_free_ph:
993 kfree(elf_phdata);
994 goto out;
995}
996
997/* This is really simpleminded and specialized - we are loading an
998 a.out library that is given an ELF header. */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700999static int load_elf_library(struct file *file)
1000{
1001 struct elf_phdr *elf_phdata;
1002 struct elf_phdr *eppnt;
1003 unsigned long elf_bss, bss, len;
1004 int retval, error, i, j;
1005 struct elfhdr elf_ex;
1006
1007 error = -ENOEXEC;
Jesper Juhlf4e5cc22006-06-23 02:05:35 -07001008 retval = kernel_read(file, 0, (char *)&elf_ex, sizeof(elf_ex));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001009 if (retval != sizeof(elf_ex))
1010 goto out;
1011
1012 if (memcmp(elf_ex.e_ident, ELFMAG, SELFMAG) != 0)
1013 goto out;
1014
1015 /* First of all, some simple consistency checks */
1016 if (elf_ex.e_type != ET_EXEC || elf_ex.e_phnum > 2 ||
Jesper Juhlf4e5cc22006-06-23 02:05:35 -07001017 !elf_check_arch(&elf_ex) || !file->f_op || !file->f_op->mmap)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001018 goto out;
1019
1020 /* Now read in all of the header information */
1021
1022 j = sizeof(struct elf_phdr) * elf_ex.e_phnum;
1023 /* j < ELF_MIN_ALIGN because elf_ex.e_phnum <= 2 */
1024
1025 error = -ENOMEM;
1026 elf_phdata = kmalloc(j, GFP_KERNEL);
1027 if (!elf_phdata)
1028 goto out;
1029
1030 eppnt = elf_phdata;
1031 error = -ENOEXEC;
1032 retval = kernel_read(file, elf_ex.e_phoff, (char *)eppnt, j);
1033 if (retval != j)
1034 goto out_free_ph;
1035
1036 for (j = 0, i = 0; i<elf_ex.e_phnum; i++)
1037 if ((eppnt + i)->p_type == PT_LOAD)
1038 j++;
1039 if (j != 1)
1040 goto out_free_ph;
1041
1042 while (eppnt->p_type != PT_LOAD)
1043 eppnt++;
1044
1045 /* Now use mmap to map the library into memory. */
1046 down_write(&current->mm->mmap_sem);
1047 error = do_mmap(file,
1048 ELF_PAGESTART(eppnt->p_vaddr),
1049 (eppnt->p_filesz +
1050 ELF_PAGEOFFSET(eppnt->p_vaddr)),
1051 PROT_READ | PROT_WRITE | PROT_EXEC,
1052 MAP_FIXED | MAP_PRIVATE | MAP_DENYWRITE,
1053 (eppnt->p_offset -
1054 ELF_PAGEOFFSET(eppnt->p_vaddr)));
1055 up_write(&current->mm->mmap_sem);
1056 if (error != ELF_PAGESTART(eppnt->p_vaddr))
1057 goto out_free_ph;
1058
1059 elf_bss = eppnt->p_vaddr + eppnt->p_filesz;
1060 if (padzero(elf_bss)) {
1061 error = -EFAULT;
1062 goto out_free_ph;
1063 }
1064
Jesper Juhlf4e5cc22006-06-23 02:05:35 -07001065 len = ELF_PAGESTART(eppnt->p_filesz + eppnt->p_vaddr +
1066 ELF_MIN_ALIGN - 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001067 bss = eppnt->p_memsz + eppnt->p_vaddr;
1068 if (bss > len) {
1069 down_write(&current->mm->mmap_sem);
1070 do_brk(len, bss - len);
1071 up_write(&current->mm->mmap_sem);
1072 }
1073 error = 0;
1074
1075out_free_ph:
1076 kfree(elf_phdata);
1077out:
1078 return error;
1079}
1080
Christoph Hellwig698ba7b2009-12-15 16:47:37 -08001081#ifdef CONFIG_ELF_CORE
Linus Torvalds1da177e2005-04-16 15:20:36 -07001082/*
1083 * ELF core dumper
1084 *
1085 * Modelled on fs/exec.c:aout_core_dump()
1086 * Jeremy Fitzhardinge <jeremy@sw.oz.au>
1087 */
1088/*
1089 * These are the only things you should do on a core-file: use only these
1090 * functions to write out all the necessary info.
1091 */
1092static int dump_write(struct file *file, const void *addr, int nr)
1093{
1094 return file->f_op->write(file, addr, nr, &file->f_pos) == nr;
1095}
1096
Daniel Jacobowitz5db92852005-06-15 22:26:34 -07001097static int dump_seek(struct file *file, loff_t off)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001098{
Andi Kleend025c9d2006-09-30 23:29:28 -07001099 if (file->f_op->llseek && file->f_op->llseek != no_llseek) {
Petr Vandrovec7f14daa2006-10-13 04:13:16 +02001100 if (file->f_op->llseek(file, off, SEEK_CUR) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001101 return 0;
Andi Kleend025c9d2006-09-30 23:29:28 -07001102 } else {
1103 char *buf = (char *)get_zeroed_page(GFP_KERNEL);
1104 if (!buf)
1105 return 0;
1106 while (off > 0) {
1107 unsigned long n = off;
1108 if (n > PAGE_SIZE)
1109 n = PAGE_SIZE;
1110 if (!dump_write(file, buf, n))
1111 return 0;
1112 off -= n;
1113 }
1114 free_page((unsigned long)buf);
1115 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001116 return 1;
1117}
1118
1119/*
Roland McGrath82df3972007-10-16 23:27:02 -07001120 * Decide what to dump of a segment, part, all or none.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001121 */
Roland McGrath82df3972007-10-16 23:27:02 -07001122static unsigned long vma_dump_size(struct vm_area_struct *vma,
1123 unsigned long mm_flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001124{
KOSAKI Motohiroe575f112008-10-18 20:27:08 -07001125#define FILTER(type) (mm_flags & (1UL << MMF_DUMP_##type))
1126
Roland McGrathe5b97dd2007-01-26 00:56:48 -08001127 /* The vma can be set up to tell us the answer directly. */
1128 if (vma->vm_flags & VM_ALWAYSDUMP)
Roland McGrath82df3972007-10-16 23:27:02 -07001129 goto whole;
Roland McGrathe5b97dd2007-01-26 00:56:48 -08001130
KOSAKI Motohiroe575f112008-10-18 20:27:08 -07001131 /* Hugetlb memory check */
1132 if (vma->vm_flags & VM_HUGETLB) {
1133 if ((vma->vm_flags & VM_SHARED) && FILTER(HUGETLB_SHARED))
1134 goto whole;
1135 if (!(vma->vm_flags & VM_SHARED) && FILTER(HUGETLB_PRIVATE))
1136 goto whole;
1137 }
1138
Linus Torvalds1da177e2005-04-16 15:20:36 -07001139 /* Do not dump I/O mapped devices or special mappings */
1140 if (vma->vm_flags & (VM_IO | VM_RESERVED))
1141 return 0;
1142
Kawai, Hidehiroa1b59e82007-07-19 01:48:29 -07001143 /* By default, dump shared memory if mapped from an anonymous file. */
1144 if (vma->vm_flags & VM_SHARED) {
Roland McGrath82df3972007-10-16 23:27:02 -07001145 if (vma->vm_file->f_path.dentry->d_inode->i_nlink == 0 ?
1146 FILTER(ANON_SHARED) : FILTER(MAPPED_SHARED))
1147 goto whole;
1148 return 0;
Kawai, Hidehiroa1b59e82007-07-19 01:48:29 -07001149 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001150
Roland McGrath82df3972007-10-16 23:27:02 -07001151 /* Dump segments that have been written to. */
1152 if (vma->anon_vma && FILTER(ANON_PRIVATE))
1153 goto whole;
1154 if (vma->vm_file == NULL)
1155 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001156
Roland McGrath82df3972007-10-16 23:27:02 -07001157 if (FILTER(MAPPED_PRIVATE))
1158 goto whole;
1159
1160 /*
1161 * If this looks like the beginning of a DSO or executable mapping,
1162 * check for an ELF header. If we find one, dump the first page to
1163 * aid in determining what was mapped here.
1164 */
Roland McGrath92dc07b2009-02-06 17:34:07 -08001165 if (FILTER(ELF_HEADERS) &&
1166 vma->vm_pgoff == 0 && (vma->vm_flags & VM_READ)) {
Roland McGrath82df3972007-10-16 23:27:02 -07001167 u32 __user *header = (u32 __user *) vma->vm_start;
1168 u32 word;
Roland McGrath92dc07b2009-02-06 17:34:07 -08001169 mm_segment_t fs = get_fs();
Roland McGrath82df3972007-10-16 23:27:02 -07001170 /*
1171 * Doing it this way gets the constant folded by GCC.
1172 */
1173 union {
1174 u32 cmp;
1175 char elfmag[SELFMAG];
1176 } magic;
1177 BUILD_BUG_ON(SELFMAG != sizeof word);
1178 magic.elfmag[EI_MAG0] = ELFMAG0;
1179 magic.elfmag[EI_MAG1] = ELFMAG1;
1180 magic.elfmag[EI_MAG2] = ELFMAG2;
1181 magic.elfmag[EI_MAG3] = ELFMAG3;
Roland McGrath92dc07b2009-02-06 17:34:07 -08001182 /*
1183 * Switch to the user "segment" for get_user(),
1184 * then put back what elf_core_dump() had in place.
1185 */
1186 set_fs(USER_DS);
1187 if (unlikely(get_user(word, header)))
1188 word = 0;
1189 set_fs(fs);
1190 if (word == magic.cmp)
Roland McGrath82df3972007-10-16 23:27:02 -07001191 return PAGE_SIZE;
1192 }
1193
1194#undef FILTER
1195
1196 return 0;
1197
1198whole:
1199 return vma->vm_end - vma->vm_start;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001200}
1201
Linus Torvalds1da177e2005-04-16 15:20:36 -07001202/* An ELF note in memory */
1203struct memelfnote
1204{
1205 const char *name;
1206 int type;
1207 unsigned int datasz;
1208 void *data;
1209};
1210
1211static int notesize(struct memelfnote *en)
1212{
1213 int sz;
1214
1215 sz = sizeof(struct elf_note);
1216 sz += roundup(strlen(en->name) + 1, 4);
1217 sz += roundup(en->datasz, 4);
1218
1219 return sz;
1220}
1221
Andi Kleend025c9d2006-09-30 23:29:28 -07001222#define DUMP_WRITE(addr, nr, foffset) \
1223 do { if (!dump_write(file, (addr), (nr))) return 0; *foffset += (nr); } while(0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001224
Andi Kleend025c9d2006-09-30 23:29:28 -07001225static int alignfile(struct file *file, loff_t *foffset)
1226{
Petr Vandroveca7a0d862006-10-13 18:42:07 +02001227 static const char buf[4] = { 0, };
Andi Kleend025c9d2006-09-30 23:29:28 -07001228 DUMP_WRITE(buf, roundup(*foffset, 4) - *foffset, foffset);
1229 return 1;
1230}
1231
1232static int writenote(struct memelfnote *men, struct file *file,
1233 loff_t *foffset)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001234{
1235 struct elf_note en;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001236 en.n_namesz = strlen(men->name) + 1;
1237 en.n_descsz = men->datasz;
1238 en.n_type = men->type;
1239
Andi Kleend025c9d2006-09-30 23:29:28 -07001240 DUMP_WRITE(&en, sizeof(en), foffset);
1241 DUMP_WRITE(men->name, en.n_namesz, foffset);
1242 if (!alignfile(file, foffset))
1243 return 0;
1244 DUMP_WRITE(men->data, men->datasz, foffset);
1245 if (!alignfile(file, foffset))
1246 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001247
1248 return 1;
1249}
1250#undef DUMP_WRITE
Linus Torvalds1da177e2005-04-16 15:20:36 -07001251
Masami Hiramatsuf6151df2009-12-17 15:27:16 -08001252#define DUMP_WRITE(addr, nr) \
1253 if ((size += (nr)) > cprm->limit || \
1254 !dump_write(cprm->file, (addr), (nr))) \
Linus Torvalds1da177e2005-04-16 15:20:36 -07001255 goto end_coredump;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001256
Roland McGrath3aba4812008-01-30 13:31:44 +01001257static void fill_elf_header(struct elfhdr *elf, int segs,
1258 u16 machine, u32 flags, u8 osabi)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001259{
Cyrill Gorcunov6970c8e2008-04-29 01:01:18 -07001260 memset(elf, 0, sizeof(*elf));
1261
Linus Torvalds1da177e2005-04-16 15:20:36 -07001262 memcpy(elf->e_ident, ELFMAG, SELFMAG);
1263 elf->e_ident[EI_CLASS] = ELF_CLASS;
1264 elf->e_ident[EI_DATA] = ELF_DATA;
1265 elf->e_ident[EI_VERSION] = EV_CURRENT;
1266 elf->e_ident[EI_OSABI] = ELF_OSABI;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001267
1268 elf->e_type = ET_CORE;
Roland McGrath3aba4812008-01-30 13:31:44 +01001269 elf->e_machine = machine;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001270 elf->e_version = EV_CURRENT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001271 elf->e_phoff = sizeof(struct elfhdr);
Roland McGrath3aba4812008-01-30 13:31:44 +01001272 elf->e_flags = flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001273 elf->e_ehsize = sizeof(struct elfhdr);
1274 elf->e_phentsize = sizeof(struct elf_phdr);
1275 elf->e_phnum = segs;
Cyrill Gorcunov6970c8e2008-04-29 01:01:18 -07001276
Linus Torvalds1da177e2005-04-16 15:20:36 -07001277 return;
1278}
1279
Andrew Morton8d6b5eee2006-09-25 23:32:04 -07001280static void fill_elf_note_phdr(struct elf_phdr *phdr, int sz, loff_t offset)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001281{
1282 phdr->p_type = PT_NOTE;
1283 phdr->p_offset = offset;
1284 phdr->p_vaddr = 0;
1285 phdr->p_paddr = 0;
1286 phdr->p_filesz = sz;
1287 phdr->p_memsz = 0;
1288 phdr->p_flags = 0;
1289 phdr->p_align = 0;
1290 return;
1291}
1292
1293static void fill_note(struct memelfnote *note, const char *name, int type,
1294 unsigned int sz, void *data)
1295{
1296 note->name = name;
1297 note->type = type;
1298 note->datasz = sz;
1299 note->data = data;
1300 return;
1301}
1302
1303/*
Jesper Juhlf4e5cc22006-06-23 02:05:35 -07001304 * fill up all the fields in prstatus from the given task struct, except
1305 * registers which need to be filled up separately.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001306 */
1307static void fill_prstatus(struct elf_prstatus *prstatus,
Jesper Juhlf4e5cc22006-06-23 02:05:35 -07001308 struct task_struct *p, long signr)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001309{
1310 prstatus->pr_info.si_signo = prstatus->pr_cursig = signr;
1311 prstatus->pr_sigpend = p->pending.signal.sig[0];
1312 prstatus->pr_sighold = p->blocked.sig[0];
Oleg Nesterov3b34fc52009-06-17 16:27:38 -07001313 rcu_read_lock();
1314 prstatus->pr_ppid = task_pid_vnr(rcu_dereference(p->real_parent));
1315 rcu_read_unlock();
Pavel Emelyanovb4888932007-10-18 23:40:14 -07001316 prstatus->pr_pid = task_pid_vnr(p);
Pavel Emelyanovb4888932007-10-18 23:40:14 -07001317 prstatus->pr_pgrp = task_pgrp_vnr(p);
1318 prstatus->pr_sid = task_session_vnr(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001319 if (thread_group_leader(p)) {
Frank Mayharf06febc2008-09-12 09:54:39 -07001320 struct task_cputime cputime;
1321
Linus Torvalds1da177e2005-04-16 15:20:36 -07001322 /*
Frank Mayharf06febc2008-09-12 09:54:39 -07001323 * This is the record for the group leader. It shows the
1324 * group-wide total, not its individual thread total.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001325 */
Frank Mayharf06febc2008-09-12 09:54:39 -07001326 thread_group_cputime(p, &cputime);
1327 cputime_to_timeval(cputime.utime, &prstatus->pr_utime);
1328 cputime_to_timeval(cputime.stime, &prstatus->pr_stime);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001329 } else {
1330 cputime_to_timeval(p->utime, &prstatus->pr_utime);
1331 cputime_to_timeval(p->stime, &prstatus->pr_stime);
1332 }
1333 cputime_to_timeval(p->signal->cutime, &prstatus->pr_cutime);
1334 cputime_to_timeval(p->signal->cstime, &prstatus->pr_cstime);
1335}
1336
1337static int fill_psinfo(struct elf_prpsinfo *psinfo, struct task_struct *p,
1338 struct mm_struct *mm)
1339{
David Howellsc69e8d92008-11-14 10:39:19 +11001340 const struct cred *cred;
Greg Kroah-Hartmana84a5052005-05-11 00:10:44 -07001341 unsigned int i, len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001342
1343 /* first copy the parameters from user space */
1344 memset(psinfo, 0, sizeof(struct elf_prpsinfo));
1345
1346 len = mm->arg_end - mm->arg_start;
1347 if (len >= ELF_PRARGSZ)
1348 len = ELF_PRARGSZ-1;
1349 if (copy_from_user(&psinfo->pr_psargs,
1350 (const char __user *)mm->arg_start, len))
1351 return -EFAULT;
1352 for(i = 0; i < len; i++)
1353 if (psinfo->pr_psargs[i] == 0)
1354 psinfo->pr_psargs[i] = ' ';
1355 psinfo->pr_psargs[len] = 0;
1356
Oleg Nesterov3b34fc52009-06-17 16:27:38 -07001357 rcu_read_lock();
1358 psinfo->pr_ppid = task_pid_vnr(rcu_dereference(p->real_parent));
1359 rcu_read_unlock();
Pavel Emelyanovb4888932007-10-18 23:40:14 -07001360 psinfo->pr_pid = task_pid_vnr(p);
Pavel Emelyanovb4888932007-10-18 23:40:14 -07001361 psinfo->pr_pgrp = task_pgrp_vnr(p);
1362 psinfo->pr_sid = task_session_vnr(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001363
1364 i = p->state ? ffz(~p->state) + 1 : 0;
1365 psinfo->pr_state = i;
Carsten Otte55148542006-03-25 03:08:22 -08001366 psinfo->pr_sname = (i > 5) ? '.' : "RSDTZW"[i];
Linus Torvalds1da177e2005-04-16 15:20:36 -07001367 psinfo->pr_zomb = psinfo->pr_sname == 'Z';
1368 psinfo->pr_nice = task_nice(p);
1369 psinfo->pr_flag = p->flags;
David Howellsc69e8d92008-11-14 10:39:19 +11001370 rcu_read_lock();
1371 cred = __task_cred(p);
1372 SET_UID(psinfo->pr_uid, cred->uid);
1373 SET_GID(psinfo->pr_gid, cred->gid);
1374 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001375 strncpy(psinfo->pr_fname, p->comm, sizeof(psinfo->pr_fname));
1376
1377 return 0;
1378}
1379
Roland McGrath3aba4812008-01-30 13:31:44 +01001380static void fill_auxv_note(struct memelfnote *note, struct mm_struct *mm)
1381{
1382 elf_addr_t *auxv = (elf_addr_t *) mm->saved_auxv;
1383 int i = 0;
1384 do
1385 i += 2;
1386 while (auxv[i - 2] != AT_NULL);
1387 fill_note(note, "CORE", NT_AUXV, i * sizeof(elf_addr_t), auxv);
1388}
1389
Roland McGrath4206d3a2008-01-30 13:31:45 +01001390#ifdef CORE_DUMP_USE_REGSET
1391#include <linux/regset.h>
1392
1393struct elf_thread_core_info {
1394 struct elf_thread_core_info *next;
1395 struct task_struct *task;
1396 struct elf_prstatus prstatus;
1397 struct memelfnote notes[0];
1398};
1399
1400struct elf_note_info {
1401 struct elf_thread_core_info *thread;
1402 struct memelfnote psinfo;
1403 struct memelfnote auxv;
1404 size_t size;
1405 int thread_notes;
1406};
1407
Roland McGrathd31472b2008-03-04 14:28:30 -08001408/*
1409 * When a regset has a writeback hook, we call it on each thread before
1410 * dumping user memory. On register window machines, this makes sure the
1411 * user memory backing the register data is up to date before we read it.
1412 */
1413static void do_thread_regset_writeback(struct task_struct *task,
1414 const struct user_regset *regset)
1415{
1416 if (regset->writeback)
1417 regset->writeback(task, regset, 1);
1418}
1419
Roland McGrath4206d3a2008-01-30 13:31:45 +01001420static int fill_thread_core_info(struct elf_thread_core_info *t,
1421 const struct user_regset_view *view,
1422 long signr, size_t *total)
1423{
1424 unsigned int i;
1425
1426 /*
1427 * NT_PRSTATUS is the one special case, because the regset data
1428 * goes into the pr_reg field inside the note contents, rather
1429 * than being the whole note contents. We fill the reset in here.
1430 * We assume that regset 0 is NT_PRSTATUS.
1431 */
1432 fill_prstatus(&t->prstatus, t->task, signr);
1433 (void) view->regsets[0].get(t->task, &view->regsets[0],
1434 0, sizeof(t->prstatus.pr_reg),
1435 &t->prstatus.pr_reg, NULL);
1436
1437 fill_note(&t->notes[0], "CORE", NT_PRSTATUS,
1438 sizeof(t->prstatus), &t->prstatus);
1439 *total += notesize(&t->notes[0]);
1440
Roland McGrathd31472b2008-03-04 14:28:30 -08001441 do_thread_regset_writeback(t->task, &view->regsets[0]);
1442
Roland McGrath4206d3a2008-01-30 13:31:45 +01001443 /*
1444 * Each other regset might generate a note too. For each regset
1445 * that has no core_note_type or is inactive, we leave t->notes[i]
1446 * all zero and we'll know to skip writing it later.
1447 */
1448 for (i = 1; i < view->n; ++i) {
1449 const struct user_regset *regset = &view->regsets[i];
Roland McGrathd31472b2008-03-04 14:28:30 -08001450 do_thread_regset_writeback(t->task, regset);
Roland McGrath4206d3a2008-01-30 13:31:45 +01001451 if (regset->core_note_type &&
1452 (!regset->active || regset->active(t->task, regset))) {
1453 int ret;
1454 size_t size = regset->n * regset->size;
1455 void *data = kmalloc(size, GFP_KERNEL);
1456 if (unlikely(!data))
1457 return 0;
1458 ret = regset->get(t->task, regset,
1459 0, size, data, NULL);
1460 if (unlikely(ret))
1461 kfree(data);
1462 else {
1463 if (regset->core_note_type != NT_PRFPREG)
1464 fill_note(&t->notes[i], "LINUX",
1465 regset->core_note_type,
1466 size, data);
1467 else {
1468 t->prstatus.pr_fpvalid = 1;
1469 fill_note(&t->notes[i], "CORE",
1470 NT_PRFPREG, size, data);
1471 }
1472 *total += notesize(&t->notes[i]);
1473 }
1474 }
1475 }
1476
1477 return 1;
1478}
1479
1480static int fill_note_info(struct elfhdr *elf, int phdrs,
1481 struct elf_note_info *info,
1482 long signr, struct pt_regs *regs)
1483{
1484 struct task_struct *dump_task = current;
1485 const struct user_regset_view *view = task_user_regset_view(dump_task);
1486 struct elf_thread_core_info *t;
1487 struct elf_prpsinfo *psinfo;
Oleg Nesterov83914442008-07-25 01:47:45 -07001488 struct core_thread *ct;
Roland McGrath4206d3a2008-01-30 13:31:45 +01001489 unsigned int i;
1490
1491 info->size = 0;
1492 info->thread = NULL;
1493
1494 psinfo = kmalloc(sizeof(*psinfo), GFP_KERNEL);
Roland McGrath4206d3a2008-01-30 13:31:45 +01001495 if (psinfo == NULL)
1496 return 0;
1497
Amerigo Wange2dbe122009-07-01 01:06:26 -04001498 fill_note(&info->psinfo, "CORE", NT_PRPSINFO, sizeof(*psinfo), psinfo);
1499
Roland McGrath4206d3a2008-01-30 13:31:45 +01001500 /*
1501 * Figure out how many notes we're going to need for each thread.
1502 */
1503 info->thread_notes = 0;
1504 for (i = 0; i < view->n; ++i)
1505 if (view->regsets[i].core_note_type != 0)
1506 ++info->thread_notes;
1507
1508 /*
1509 * Sanity check. We rely on regset 0 being in NT_PRSTATUS,
1510 * since it is our one special case.
1511 */
1512 if (unlikely(info->thread_notes == 0) ||
1513 unlikely(view->regsets[0].core_note_type != NT_PRSTATUS)) {
1514 WARN_ON(1);
1515 return 0;
1516 }
1517
1518 /*
1519 * Initialize the ELF file header.
1520 */
1521 fill_elf_header(elf, phdrs,
1522 view->e_machine, view->e_flags, view->ei_osabi);
1523
1524 /*
1525 * Allocate a structure for each thread.
1526 */
Oleg Nesterov83914442008-07-25 01:47:45 -07001527 for (ct = &dump_task->mm->core_state->dumper; ct; ct = ct->next) {
1528 t = kzalloc(offsetof(struct elf_thread_core_info,
1529 notes[info->thread_notes]),
1530 GFP_KERNEL);
1531 if (unlikely(!t))
1532 return 0;
Oleg Nesterov24d52882008-07-25 01:47:40 -07001533
Oleg Nesterov83914442008-07-25 01:47:45 -07001534 t->task = ct->task;
1535 if (ct->task == dump_task || !info->thread) {
1536 t->next = info->thread;
1537 info->thread = t;
1538 } else {
1539 /*
1540 * Make sure to keep the original task at
1541 * the head of the list.
1542 */
1543 t->next = info->thread->next;
1544 info->thread->next = t;
Roland McGrath4206d3a2008-01-30 13:31:45 +01001545 }
Oleg Nesterov83914442008-07-25 01:47:45 -07001546 }
Roland McGrath4206d3a2008-01-30 13:31:45 +01001547
1548 /*
1549 * Now fill in each thread's information.
1550 */
1551 for (t = info->thread; t != NULL; t = t->next)
1552 if (!fill_thread_core_info(t, view, signr, &info->size))
1553 return 0;
1554
1555 /*
1556 * Fill in the two process-wide notes.
1557 */
1558 fill_psinfo(psinfo, dump_task->group_leader, dump_task->mm);
1559 info->size += notesize(&info->psinfo);
1560
1561 fill_auxv_note(&info->auxv, current->mm);
1562 info->size += notesize(&info->auxv);
1563
1564 return 1;
1565}
1566
1567static size_t get_note_info_size(struct elf_note_info *info)
1568{
1569 return info->size;
1570}
1571
1572/*
1573 * Write all the notes for each thread. When writing the first thread, the
1574 * process-wide notes are interleaved after the first thread-specific note.
1575 */
1576static int write_note_info(struct elf_note_info *info,
1577 struct file *file, loff_t *foffset)
1578{
1579 bool first = 1;
1580 struct elf_thread_core_info *t = info->thread;
1581
1582 do {
1583 int i;
1584
1585 if (!writenote(&t->notes[0], file, foffset))
1586 return 0;
1587
1588 if (first && !writenote(&info->psinfo, file, foffset))
1589 return 0;
1590 if (first && !writenote(&info->auxv, file, foffset))
1591 return 0;
1592
1593 for (i = 1; i < info->thread_notes; ++i)
1594 if (t->notes[i].data &&
1595 !writenote(&t->notes[i], file, foffset))
1596 return 0;
1597
1598 first = 0;
1599 t = t->next;
1600 } while (t);
1601
1602 return 1;
1603}
1604
1605static void free_note_info(struct elf_note_info *info)
1606{
1607 struct elf_thread_core_info *threads = info->thread;
1608 while (threads) {
1609 unsigned int i;
1610 struct elf_thread_core_info *t = threads;
1611 threads = t->next;
1612 WARN_ON(t->notes[0].data && t->notes[0].data != &t->prstatus);
1613 for (i = 1; i < info->thread_notes; ++i)
1614 kfree(t->notes[i].data);
1615 kfree(t);
1616 }
1617 kfree(info->psinfo.data);
1618}
1619
1620#else
1621
Linus Torvalds1da177e2005-04-16 15:20:36 -07001622/* Here is the structure in which status of each thread is captured. */
1623struct elf_thread_status
1624{
1625 struct list_head list;
1626 struct elf_prstatus prstatus; /* NT_PRSTATUS */
1627 elf_fpregset_t fpu; /* NT_PRFPREG */
1628 struct task_struct *thread;
1629#ifdef ELF_CORE_COPY_XFPREGS
Mark Nelson5b20cd82007-10-16 23:25:39 -07001630 elf_fpxregset_t xfpu; /* ELF_CORE_XFPREG_TYPE */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001631#endif
1632 struct memelfnote notes[3];
1633 int num_notes;
1634};
1635
1636/*
1637 * In order to add the specific thread information for the elf file format,
Jesper Juhlf4e5cc22006-06-23 02:05:35 -07001638 * we need to keep a linked list of every threads pr_status and then create
1639 * a single section for them in the final core file.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001640 */
1641static int elf_dump_thread_status(long signr, struct elf_thread_status *t)
1642{
1643 int sz = 0;
1644 struct task_struct *p = t->thread;
1645 t->num_notes = 0;
1646
1647 fill_prstatus(&t->prstatus, p, signr);
1648 elf_core_copy_task_regs(p, &t->prstatus.pr_reg);
1649
Jesper Juhlf4e5cc22006-06-23 02:05:35 -07001650 fill_note(&t->notes[0], "CORE", NT_PRSTATUS, sizeof(t->prstatus),
1651 &(t->prstatus));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001652 t->num_notes++;
1653 sz += notesize(&t->notes[0]);
1654
Jesper Juhlf4e5cc22006-06-23 02:05:35 -07001655 if ((t->prstatus.pr_fpvalid = elf_core_copy_task_fpregs(p, NULL,
1656 &t->fpu))) {
1657 fill_note(&t->notes[1], "CORE", NT_PRFPREG, sizeof(t->fpu),
1658 &(t->fpu));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001659 t->num_notes++;
1660 sz += notesize(&t->notes[1]);
1661 }
1662
1663#ifdef ELF_CORE_COPY_XFPREGS
1664 if (elf_core_copy_task_xfpregs(p, &t->xfpu)) {
Mark Nelson5b20cd82007-10-16 23:25:39 -07001665 fill_note(&t->notes[2], "LINUX", ELF_CORE_XFPREG_TYPE,
1666 sizeof(t->xfpu), &t->xfpu);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001667 t->num_notes++;
1668 sz += notesize(&t->notes[2]);
1669 }
1670#endif
1671 return sz;
1672}
1673
Roland McGrath3aba4812008-01-30 13:31:44 +01001674struct elf_note_info {
1675 struct memelfnote *notes;
1676 struct elf_prstatus *prstatus; /* NT_PRSTATUS */
1677 struct elf_prpsinfo *psinfo; /* NT_PRPSINFO */
1678 struct list_head thread_list;
1679 elf_fpregset_t *fpu;
1680#ifdef ELF_CORE_COPY_XFPREGS
1681 elf_fpxregset_t *xfpu;
1682#endif
1683 int thread_status_size;
1684 int numnote;
1685};
1686
Amerigo Wang0cf062d2009-09-23 15:57:05 -07001687static int elf_note_info_init(struct elf_note_info *info)
Roland McGrath3aba4812008-01-30 13:31:44 +01001688{
Amerigo Wang0cf062d2009-09-23 15:57:05 -07001689 memset(info, 0, sizeof(*info));
Roland McGrath3aba4812008-01-30 13:31:44 +01001690 INIT_LIST_HEAD(&info->thread_list);
1691
Amerigo Wang0cf062d2009-09-23 15:57:05 -07001692 /* Allocate space for six ELF notes */
1693 info->notes = kmalloc(6 * sizeof(struct memelfnote), GFP_KERNEL);
Roland McGrath3aba4812008-01-30 13:31:44 +01001694 if (!info->notes)
1695 return 0;
1696 info->psinfo = kmalloc(sizeof(*info->psinfo), GFP_KERNEL);
1697 if (!info->psinfo)
Amerigo Wang0cf062d2009-09-23 15:57:05 -07001698 goto notes_free;
Roland McGrath3aba4812008-01-30 13:31:44 +01001699 info->prstatus = kmalloc(sizeof(*info->prstatus), GFP_KERNEL);
1700 if (!info->prstatus)
Amerigo Wang0cf062d2009-09-23 15:57:05 -07001701 goto psinfo_free;
Roland McGrath3aba4812008-01-30 13:31:44 +01001702 info->fpu = kmalloc(sizeof(*info->fpu), GFP_KERNEL);
1703 if (!info->fpu)
Amerigo Wang0cf062d2009-09-23 15:57:05 -07001704 goto prstatus_free;
Roland McGrath3aba4812008-01-30 13:31:44 +01001705#ifdef ELF_CORE_COPY_XFPREGS
1706 info->xfpu = kmalloc(sizeof(*info->xfpu), GFP_KERNEL);
1707 if (!info->xfpu)
Amerigo Wang0cf062d2009-09-23 15:57:05 -07001708 goto fpu_free;
Roland McGrath3aba4812008-01-30 13:31:44 +01001709#endif
Amerigo Wang0cf062d2009-09-23 15:57:05 -07001710 return 1;
1711#ifdef ELF_CORE_COPY_XFPREGS
1712 fpu_free:
1713 kfree(info->fpu);
1714#endif
1715 prstatus_free:
1716 kfree(info->prstatus);
1717 psinfo_free:
1718 kfree(info->psinfo);
1719 notes_free:
1720 kfree(info->notes);
1721 return 0;
1722}
Roland McGrath3aba4812008-01-30 13:31:44 +01001723
Amerigo Wang0cf062d2009-09-23 15:57:05 -07001724static int fill_note_info(struct elfhdr *elf, int phdrs,
1725 struct elf_note_info *info,
1726 long signr, struct pt_regs *regs)
1727{
1728 struct list_head *t;
1729
1730 if (!elf_note_info_init(info))
1731 return 0;
1732
Roland McGrath3aba4812008-01-30 13:31:44 +01001733 if (signr) {
Oleg Nesterov83914442008-07-25 01:47:45 -07001734 struct core_thread *ct;
WANG Cong4220b7f2008-04-29 01:01:18 -07001735 struct elf_thread_status *ets;
Oleg Nesterov24d52882008-07-25 01:47:40 -07001736
Oleg Nesterov83914442008-07-25 01:47:45 -07001737 for (ct = current->mm->core_state->dumper.next;
1738 ct; ct = ct->next) {
1739 ets = kzalloc(sizeof(*ets), GFP_KERNEL);
1740 if (!ets)
1741 return 0;
1742
1743 ets->thread = ct->task;
1744 list_add(&ets->list, &info->thread_list);
1745 }
1746
Roland McGrath3aba4812008-01-30 13:31:44 +01001747 list_for_each(t, &info->thread_list) {
Roland McGrath3aba4812008-01-30 13:31:44 +01001748 int sz;
1749
WANG Cong4220b7f2008-04-29 01:01:18 -07001750 ets = list_entry(t, struct elf_thread_status, list);
1751 sz = elf_dump_thread_status(signr, ets);
Roland McGrath3aba4812008-01-30 13:31:44 +01001752 info->thread_status_size += sz;
1753 }
1754 }
1755 /* now collect the dump for the current */
1756 memset(info->prstatus, 0, sizeof(*info->prstatus));
1757 fill_prstatus(info->prstatus, current, signr);
1758 elf_core_copy_regs(&info->prstatus->pr_reg, regs);
1759
1760 /* Set up header */
1761 fill_elf_header(elf, phdrs, ELF_ARCH, ELF_CORE_EFLAGS, ELF_OSABI);
1762
1763 /*
1764 * Set up the notes in similar form to SVR4 core dumps made
1765 * with info from their /proc.
1766 */
1767
1768 fill_note(info->notes + 0, "CORE", NT_PRSTATUS,
1769 sizeof(*info->prstatus), info->prstatus);
1770 fill_psinfo(info->psinfo, current->group_leader, current->mm);
1771 fill_note(info->notes + 1, "CORE", NT_PRPSINFO,
1772 sizeof(*info->psinfo), info->psinfo);
1773
1774 info->numnote = 2;
1775
1776 fill_auxv_note(&info->notes[info->numnote++], current->mm);
1777
1778 /* Try to dump the FPU. */
1779 info->prstatus->pr_fpvalid = elf_core_copy_task_fpregs(current, regs,
1780 info->fpu);
1781 if (info->prstatus->pr_fpvalid)
1782 fill_note(info->notes + info->numnote++,
1783 "CORE", NT_PRFPREG, sizeof(*info->fpu), info->fpu);
1784#ifdef ELF_CORE_COPY_XFPREGS
1785 if (elf_core_copy_task_xfpregs(current, info->xfpu))
1786 fill_note(info->notes + info->numnote++,
1787 "LINUX", ELF_CORE_XFPREG_TYPE,
1788 sizeof(*info->xfpu), info->xfpu);
1789#endif
1790
1791 return 1;
Roland McGrath3aba4812008-01-30 13:31:44 +01001792}
1793
1794static size_t get_note_info_size(struct elf_note_info *info)
1795{
1796 int sz = 0;
1797 int i;
1798
1799 for (i = 0; i < info->numnote; i++)
1800 sz += notesize(info->notes + i);
1801
1802 sz += info->thread_status_size;
1803
1804 return sz;
1805}
1806
1807static int write_note_info(struct elf_note_info *info,
1808 struct file *file, loff_t *foffset)
1809{
1810 int i;
1811 struct list_head *t;
1812
1813 for (i = 0; i < info->numnote; i++)
1814 if (!writenote(info->notes + i, file, foffset))
1815 return 0;
1816
1817 /* write out the thread status notes section */
1818 list_for_each(t, &info->thread_list) {
1819 struct elf_thread_status *tmp =
1820 list_entry(t, struct elf_thread_status, list);
1821
1822 for (i = 0; i < tmp->num_notes; i++)
1823 if (!writenote(&tmp->notes[i], file, foffset))
1824 return 0;
1825 }
1826
1827 return 1;
1828}
1829
1830static void free_note_info(struct elf_note_info *info)
1831{
1832 while (!list_empty(&info->thread_list)) {
1833 struct list_head *tmp = info->thread_list.next;
1834 list_del(tmp);
1835 kfree(list_entry(tmp, struct elf_thread_status, list));
1836 }
1837
1838 kfree(info->prstatus);
1839 kfree(info->psinfo);
1840 kfree(info->notes);
1841 kfree(info->fpu);
1842#ifdef ELF_CORE_COPY_XFPREGS
1843 kfree(info->xfpu);
1844#endif
1845}
1846
Roland McGrath4206d3a2008-01-30 13:31:45 +01001847#endif
1848
Roland McGrathf47aef52007-01-26 00:56:49 -08001849static struct vm_area_struct *first_vma(struct task_struct *tsk,
1850 struct vm_area_struct *gate_vma)
1851{
1852 struct vm_area_struct *ret = tsk->mm->mmap;
1853
1854 if (ret)
1855 return ret;
1856 return gate_vma;
1857}
1858/*
1859 * Helper function for iterating across a vma list. It ensures that the caller
1860 * will visit `gate_vma' prior to terminating the search.
1861 */
1862static struct vm_area_struct *next_vma(struct vm_area_struct *this_vma,
1863 struct vm_area_struct *gate_vma)
1864{
1865 struct vm_area_struct *ret;
1866
1867 ret = this_vma->vm_next;
1868 if (ret)
1869 return ret;
1870 if (this_vma == gate_vma)
1871 return NULL;
1872 return gate_vma;
1873}
1874
Linus Torvalds1da177e2005-04-16 15:20:36 -07001875/*
1876 * Actual dumper
1877 *
1878 * This is a two-pass process; first we find the offsets of the bits,
1879 * and then they are actually written out. If we run out of core limit
1880 * we just truncate.
1881 */
Masami Hiramatsuf6151df2009-12-17 15:27:16 -08001882static int elf_core_dump(struct coredump_params *cprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001883{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001884 int has_dumped = 0;
1885 mm_segment_t fs;
1886 int segs;
1887 size_t size = 0;
Roland McGrathf47aef52007-01-26 00:56:49 -08001888 struct vm_area_struct *vma, *gate_vma;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001889 struct elfhdr *elf = NULL;
Andi Kleend025c9d2006-09-30 23:29:28 -07001890 loff_t offset = 0, dataoff, foffset;
Kawai, Hidehiroa1b59e82007-07-19 01:48:29 -07001891 unsigned long mm_flags;
Roland McGrath3aba4812008-01-30 13:31:44 +01001892 struct elf_note_info info;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001893
1894 /*
1895 * We no longer stop all VM operations.
1896 *
Jesper Juhlf4e5cc22006-06-23 02:05:35 -07001897 * This is because those proceses that could possibly change map_count
1898 * or the mmap / vma pages are now blocked in do_exit on current
1899 * finishing this core dump.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001900 *
1901 * Only ptrace can touch these memory addresses, but it doesn't change
Jesper Juhlf4e5cc22006-06-23 02:05:35 -07001902 * the map_count or the pages allocated. So no possibility of crashing
Linus Torvalds1da177e2005-04-16 15:20:36 -07001903 * exists while dumping the mm->vm_next areas to the core file.
1904 */
1905
1906 /* alloc memory for large data structures: too large to be on stack */
1907 elf = kmalloc(sizeof(*elf), GFP_KERNEL);
1908 if (!elf)
WANG Cong5f719552008-05-06 12:45:35 +08001909 goto out;
KAMEZAWA Hiroyuki341c87b2009-06-30 11:41:23 -07001910 /*
1911 * The number of segs are recored into ELF header as 16bit value.
1912 * Please check DEFAULT_MAX_MAP_COUNT definition when you modify here.
1913 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001914 segs = current->mm->map_count;
1915#ifdef ELF_CORE_EXTRA_PHDRS
1916 segs += ELF_CORE_EXTRA_PHDRS;
1917#endif
1918
Roland McGrathf47aef52007-01-26 00:56:49 -08001919 gate_vma = get_gate_vma(current);
1920 if (gate_vma != NULL)
1921 segs++;
1922
Roland McGrath3aba4812008-01-30 13:31:44 +01001923 /*
1924 * Collect all the non-memory information about the process for the
1925 * notes. This also sets up the file header.
1926 */
1927 if (!fill_note_info(elf, segs + 1, /* including notes section */
Masami Hiramatsuf6151df2009-12-17 15:27:16 -08001928 &info, cprm->signr, cprm->regs))
Roland McGrath3aba4812008-01-30 13:31:44 +01001929 goto cleanup;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001930
1931 has_dumped = 1;
1932 current->flags |= PF_DUMPCORE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001933
1934 fs = get_fs();
1935 set_fs(KERNEL_DS);
1936
1937 DUMP_WRITE(elf, sizeof(*elf));
1938 offset += sizeof(*elf); /* Elf header */
Petr Vandroveca7a0d862006-10-13 18:42:07 +02001939 offset += (segs + 1) * sizeof(struct elf_phdr); /* Program headers */
1940 foffset = offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001941
1942 /* Write notes phdr entry */
1943 {
1944 struct elf_phdr phdr;
Roland McGrath3aba4812008-01-30 13:31:44 +01001945 size_t sz = get_note_info_size(&info);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001946
Michael Ellermane5501492007-09-19 14:38:12 +10001947 sz += elf_coredump_extra_notes_size();
Dwayne Grant McConnellbf1ab972006-11-23 00:46:37 +01001948
Linus Torvalds1da177e2005-04-16 15:20:36 -07001949 fill_elf_note_phdr(&phdr, sz, offset);
1950 offset += sz;
1951 DUMP_WRITE(&phdr, sizeof(phdr));
1952 }
1953
Linus Torvalds1da177e2005-04-16 15:20:36 -07001954 dataoff = offset = roundup(offset, ELF_EXEC_PAGESIZE);
1955
Kawai, Hidehiroa1b59e82007-07-19 01:48:29 -07001956 /*
1957 * We must use the same mm->flags while dumping core to avoid
1958 * inconsistency between the program headers and bodies, otherwise an
1959 * unusable core file can be generated.
1960 */
1961 mm_flags = current->mm->flags;
1962
Linus Torvalds1da177e2005-04-16 15:20:36 -07001963 /* Write program headers for segments dump */
Roland McGrathf47aef52007-01-26 00:56:49 -08001964 for (vma = first_vma(current, gate_vma); vma != NULL;
1965 vma = next_vma(vma, gate_vma)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001966 struct elf_phdr phdr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001967
1968 phdr.p_type = PT_LOAD;
1969 phdr.p_offset = offset;
1970 phdr.p_vaddr = vma->vm_start;
1971 phdr.p_paddr = 0;
Roland McGrath82df3972007-10-16 23:27:02 -07001972 phdr.p_filesz = vma_dump_size(vma, mm_flags);
1973 phdr.p_memsz = vma->vm_end - vma->vm_start;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001974 offset += phdr.p_filesz;
1975 phdr.p_flags = vma->vm_flags & VM_READ ? PF_R : 0;
Jesper Juhlf4e5cc22006-06-23 02:05:35 -07001976 if (vma->vm_flags & VM_WRITE)
1977 phdr.p_flags |= PF_W;
1978 if (vma->vm_flags & VM_EXEC)
1979 phdr.p_flags |= PF_X;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001980 phdr.p_align = ELF_EXEC_PAGESIZE;
1981
1982 DUMP_WRITE(&phdr, sizeof(phdr));
1983 }
1984
1985#ifdef ELF_CORE_WRITE_EXTRA_PHDRS
1986 ELF_CORE_WRITE_EXTRA_PHDRS;
1987#endif
1988
1989 /* write out the notes section */
Masami Hiramatsuf6151df2009-12-17 15:27:16 -08001990 if (!write_note_info(&info, cprm->file, &foffset))
Roland McGrath3aba4812008-01-30 13:31:44 +01001991 goto end_coredump;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001992
Masami Hiramatsuf6151df2009-12-17 15:27:16 -08001993 if (elf_coredump_extra_notes_write(cprm->file, &foffset))
Michael Ellermane5501492007-09-19 14:38:12 +10001994 goto end_coredump;
Dwayne Grant McConnellbf1ab972006-11-23 00:46:37 +01001995
Andi Kleend025c9d2006-09-30 23:29:28 -07001996 /* Align to page */
Masami Hiramatsuf6151df2009-12-17 15:27:16 -08001997 if (!dump_seek(cprm->file, dataoff - foffset))
Hugh Dickinsf3e8fcc2009-09-21 17:03:25 -07001998 goto end_coredump;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001999
Roland McGrathf47aef52007-01-26 00:56:49 -08002000 for (vma = first_vma(current, gate_vma); vma != NULL;
2001 vma = next_vma(vma, gate_vma)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002002 unsigned long addr;
Roland McGrath82df3972007-10-16 23:27:02 -07002003 unsigned long end;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002004
Roland McGrath82df3972007-10-16 23:27:02 -07002005 end = vma->vm_start + vma_dump_size(vma, mm_flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002006
Roland McGrath82df3972007-10-16 23:27:02 -07002007 for (addr = vma->vm_start; addr < end; addr += PAGE_SIZE) {
Jesper Juhlf4e5cc22006-06-23 02:05:35 -07002008 struct page *page;
Hugh Dickinsf3e8fcc2009-09-21 17:03:25 -07002009 int stop;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002010
Hugh Dickinsf3e8fcc2009-09-21 17:03:25 -07002011 page = get_dump_page(addr);
2012 if (page) {
2013 void *kaddr = kmap(page);
Masami Hiramatsuf6151df2009-12-17 15:27:16 -08002014 stop = ((size += PAGE_SIZE) > cprm->limit) ||
2015 !dump_write(cprm->file, kaddr,
2016 PAGE_SIZE);
Hugh Dickinsf3e8fcc2009-09-21 17:03:25 -07002017 kunmap(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002018 page_cache_release(page);
Hugh Dickinsf3e8fcc2009-09-21 17:03:25 -07002019 } else
Masami Hiramatsuf6151df2009-12-17 15:27:16 -08002020 stop = !dump_seek(cprm->file, PAGE_SIZE);
Hugh Dickinsf3e8fcc2009-09-21 17:03:25 -07002021 if (stop)
2022 goto end_coredump;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002023 }
2024 }
2025
2026#ifdef ELF_CORE_WRITE_EXTRA_DATA
2027 ELF_CORE_WRITE_EXTRA_DATA;
2028#endif
2029
Linus Torvalds1da177e2005-04-16 15:20:36 -07002030end_coredump:
2031 set_fs(fs);
2032
2033cleanup:
Roland McGrath3aba4812008-01-30 13:31:44 +01002034 free_note_info(&info);
WANG Cong5f719552008-05-06 12:45:35 +08002035 kfree(elf);
2036out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002037 return has_dumped;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002038}
2039
Christoph Hellwig698ba7b2009-12-15 16:47:37 -08002040#endif /* CONFIG_ELF_CORE */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002041
2042static int __init init_elf_binfmt(void)
2043{
2044 return register_binfmt(&elf_format);
2045}
2046
2047static void __exit exit_elf_binfmt(void)
2048{
2049 /* Remove the COFF and ELF loaders. */
2050 unregister_binfmt(&elf_format);
2051}
2052
2053core_initcall(init_elf_binfmt);
2054module_exit(exit_elf_binfmt);
2055MODULE_LICENSE("GPL");