[PATCH] vmscan: balancing fix
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / fs / binfmt_aout.c
blob72011826f0cb340aabc54f65ee7d31653fc7d364
1 /*
2 * linux/fs/binfmt_aout.c
4 * Copyright (C) 1991, 1992, 1996 Linus Torvalds
5 */
7 #include <linux/module.h>
9 #include <linux/time.h>
10 #include <linux/kernel.h>
11 #include <linux/mm.h>
12 #include <linux/mman.h>
13 #include <linux/a.out.h>
14 #include <linux/errno.h>
15 #include <linux/signal.h>
16 #include <linux/string.h>
17 #include <linux/fs.h>
18 #include <linux/file.h>
19 #include <linux/stat.h>
20 #include <linux/fcntl.h>
21 #include <linux/ptrace.h>
22 #include <linux/user.h>
23 #include <linux/slab.h>
24 #include <linux/binfmts.h>
25 #include <linux/personality.h>
26 #include <linux/init.h>
28 #include <asm/system.h>
29 #include <asm/uaccess.h>
30 #include <asm/cacheflush.h>
32 static int load_aout_binary(struct linux_binprm *, struct pt_regs * regs);
33 static int load_aout_library(struct file*);
34 static int aout_core_dump(long signr, struct pt_regs * regs, struct file *file);
36 extern void dump_thread(struct pt_regs *, struct user *);
38 static struct linux_binfmt aout_format = {
39 .module = THIS_MODULE,
40 .load_binary = load_aout_binary,
41 .load_shlib = load_aout_library,
42 .core_dump = aout_core_dump,
43 .min_coredump = PAGE_SIZE
46 #define BAD_ADDR(x) ((unsigned long)(x) >= TASK_SIZE)
48 static int set_brk(unsigned long start, unsigned long end)
50 start = PAGE_ALIGN(start);
51 end = PAGE_ALIGN(end);
52 if (end > start) {
53 unsigned long addr;
54 down_write(&current->mm->mmap_sem);
55 addr = do_brk(start, end - start);
56 up_write(&current->mm->mmap_sem);
57 if (BAD_ADDR(addr))
58 return addr;
60 return 0;
64 * These are the only things you should do on a core-file: use only these
65 * macros to write out all the necessary info.
68 static int dump_write(struct file *file, const void *addr, int nr)
70 return file->f_op->write(file, addr, nr, &file->f_pos) == nr;
73 #define DUMP_WRITE(addr, nr) \
74 if (!dump_write(file, (void *)(addr), (nr))) \
75 goto end_coredump;
77 #define DUMP_SEEK(offset) \
78 if (file->f_op->llseek) { \
79 if (file->f_op->llseek(file,(offset),0) != (offset)) \
80 goto end_coredump; \
81 } else file->f_pos = (offset)
84 * Routine writes a core dump image in the current directory.
85 * Currently only a stub-function.
87 * Note that setuid/setgid files won't make a core-dump if the uid/gid
88 * changed due to the set[u|g]id. It's enforced by the "current->mm->dumpable"
89 * field, which also makes sure the core-dumps won't be recursive if the
90 * dumping of the process results in another error..
93 static int aout_core_dump(long signr, struct pt_regs * regs, struct file *file)
95 mm_segment_t fs;
96 int has_dumped = 0;
97 unsigned long dump_start, dump_size;
98 struct user dump;
99 #if defined(__alpha__)
100 # define START_DATA(u) (u.start_data)
101 #elif defined(__arm__)
102 # define START_DATA(u) ((u.u_tsize << PAGE_SHIFT) + u.start_code)
103 #elif defined(__sparc__)
104 # define START_DATA(u) (u.u_tsize)
105 #elif defined(__i386__) || defined(__mc68000__) || defined(__arch_um__)
106 # define START_DATA(u) (u.u_tsize << PAGE_SHIFT)
107 #endif
108 #ifdef __sparc__
109 # define START_STACK(u) ((regs->u_regs[UREG_FP]) & ~(PAGE_SIZE - 1))
110 #else
111 # define START_STACK(u) (u.start_stack)
112 #endif
114 fs = get_fs();
115 set_fs(KERNEL_DS);
116 has_dumped = 1;
117 current->flags |= PF_DUMPCORE;
118 strncpy(dump.u_comm, current->comm, sizeof(dump.u_comm));
119 #ifndef __sparc__
120 dump.u_ar0 = (void *)(((unsigned long)(&dump.regs)) - ((unsigned long)(&dump)));
121 #endif
122 dump.signal = signr;
123 dump_thread(regs, &dump);
125 /* If the size of the dump file exceeds the rlimit, then see what would happen
126 if we wrote the stack, but not the data area. */
127 #ifdef __sparc__
128 if ((dump.u_dsize+dump.u_ssize) >
129 current->signal->rlim[RLIMIT_CORE].rlim_cur)
130 dump.u_dsize = 0;
131 #else
132 if ((dump.u_dsize+dump.u_ssize+1) * PAGE_SIZE >
133 current->signal->rlim[RLIMIT_CORE].rlim_cur)
134 dump.u_dsize = 0;
135 #endif
137 /* Make sure we have enough room to write the stack and data areas. */
138 #ifdef __sparc__
139 if ((dump.u_ssize) >
140 current->signal->rlim[RLIMIT_CORE].rlim_cur)
141 dump.u_ssize = 0;
142 #else
143 if ((dump.u_ssize+1) * PAGE_SIZE >
144 current->signal->rlim[RLIMIT_CORE].rlim_cur)
145 dump.u_ssize = 0;
146 #endif
148 /* make sure we actually have a data and stack area to dump */
149 set_fs(USER_DS);
150 #ifdef __sparc__
151 if (!access_ok(VERIFY_READ, (void __user *)START_DATA(dump), dump.u_dsize))
152 dump.u_dsize = 0;
153 if (!access_ok(VERIFY_READ, (void __user *)START_STACK(dump), dump.u_ssize))
154 dump.u_ssize = 0;
155 #else
156 if (!access_ok(VERIFY_READ, (void __user *)START_DATA(dump), dump.u_dsize << PAGE_SHIFT))
157 dump.u_dsize = 0;
158 if (!access_ok(VERIFY_READ, (void __user *)START_STACK(dump), dump.u_ssize << PAGE_SHIFT))
159 dump.u_ssize = 0;
160 #endif
162 set_fs(KERNEL_DS);
163 /* struct user */
164 DUMP_WRITE(&dump,sizeof(dump));
165 /* Now dump all of the user data. Include malloced stuff as well */
166 #ifndef __sparc__
167 DUMP_SEEK(PAGE_SIZE);
168 #endif
169 /* now we start writing out the user space info */
170 set_fs(USER_DS);
171 /* Dump the data area */
172 if (dump.u_dsize != 0) {
173 dump_start = START_DATA(dump);
174 #ifdef __sparc__
175 dump_size = dump.u_dsize;
176 #else
177 dump_size = dump.u_dsize << PAGE_SHIFT;
178 #endif
179 DUMP_WRITE(dump_start,dump_size);
181 /* Now prepare to dump the stack area */
182 if (dump.u_ssize != 0) {
183 dump_start = START_STACK(dump);
184 #ifdef __sparc__
185 dump_size = dump.u_ssize;
186 #else
187 dump_size = dump.u_ssize << PAGE_SHIFT;
188 #endif
189 DUMP_WRITE(dump_start,dump_size);
191 /* Finally dump the task struct. Not be used by gdb, but could be useful */
192 set_fs(KERNEL_DS);
193 DUMP_WRITE(current,sizeof(*current));
194 end_coredump:
195 set_fs(fs);
196 return has_dumped;
200 * create_aout_tables() parses the env- and arg-strings in new user
201 * memory and creates the pointer tables from them, and puts their
202 * addresses on the "stack", returning the new stack pointer value.
204 static unsigned long __user *create_aout_tables(char __user *p, struct linux_binprm * bprm)
206 char __user * __user *argv;
207 char __user * __user *envp;
208 unsigned long __user *sp;
209 int argc = bprm->argc;
210 int envc = bprm->envc;
212 sp = (void __user *)((-(unsigned long)sizeof(char *)) & (unsigned long) p);
213 #ifdef __sparc__
214 /* This imposes the proper stack alignment for a new process. */
215 sp = (void __user *) (((unsigned long) sp) & ~7);
216 if ((envc+argc+3)&1) --sp;
217 #endif
218 #ifdef __alpha__
219 /* whee.. test-programs are so much fun. */
220 put_user(0, --sp);
221 put_user(0, --sp);
222 if (bprm->loader) {
223 put_user(0, --sp);
224 put_user(0x3eb, --sp);
225 put_user(bprm->loader, --sp);
226 put_user(0x3ea, --sp);
228 put_user(bprm->exec, --sp);
229 put_user(0x3e9, --sp);
230 #endif
231 sp -= envc+1;
232 envp = (char __user * __user *) sp;
233 sp -= argc+1;
234 argv = (char __user * __user *) sp;
235 #if defined(__i386__) || defined(__mc68000__) || defined(__arm__) || defined(__arch_um__)
236 put_user((unsigned long) envp,--sp);
237 put_user((unsigned long) argv,--sp);
238 #endif
239 put_user(argc,--sp);
240 current->mm->arg_start = (unsigned long) p;
241 while (argc-->0) {
242 char c;
243 put_user(p,argv++);
244 do {
245 get_user(c,p++);
246 } while (c);
248 put_user(NULL,argv);
249 current->mm->arg_end = current->mm->env_start = (unsigned long) p;
250 while (envc-->0) {
251 char c;
252 put_user(p,envp++);
253 do {
254 get_user(c,p++);
255 } while (c);
257 put_user(NULL,envp);
258 current->mm->env_end = (unsigned long) p;
259 return sp;
263 * These are the functions used to load a.out style executables and shared
264 * libraries. There is no binary dependent code anywhere else.
267 static int load_aout_binary(struct linux_binprm * bprm, struct pt_regs * regs)
269 struct exec ex;
270 unsigned long error;
271 unsigned long fd_offset;
272 unsigned long rlim;
273 int retval;
275 ex = *((struct exec *) bprm->buf); /* exec-header */
276 if ((N_MAGIC(ex) != ZMAGIC && N_MAGIC(ex) != OMAGIC &&
277 N_MAGIC(ex) != QMAGIC && N_MAGIC(ex) != NMAGIC) ||
278 N_TRSIZE(ex) || N_DRSIZE(ex) ||
279 i_size_read(bprm->file->f_dentry->d_inode) < ex.a_text+ex.a_data+N_SYMSIZE(ex)+N_TXTOFF(ex)) {
280 return -ENOEXEC;
283 fd_offset = N_TXTOFF(ex);
285 /* Check initial limits. This avoids letting people circumvent
286 * size limits imposed on them by creating programs with large
287 * arrays in the data or bss.
289 rlim = current->signal->rlim[RLIMIT_DATA].rlim_cur;
290 if (rlim >= RLIM_INFINITY)
291 rlim = ~0;
292 if (ex.a_data + ex.a_bss > rlim)
293 return -ENOMEM;
295 /* Flush all traces of the currently running executable */
296 retval = flush_old_exec(bprm);
297 if (retval)
298 return retval;
300 /* OK, This is the point of no return */
301 #if defined(__alpha__)
302 SET_AOUT_PERSONALITY(bprm, ex);
303 #elif defined(__sparc__)
304 set_personality(PER_SUNOS);
305 #if !defined(__sparc_v9__)
306 memcpy(&current->thread.core_exec, &ex, sizeof(struct exec));
307 #endif
308 #else
309 set_personality(PER_LINUX);
310 #endif
312 current->mm->end_code = ex.a_text +
313 (current->mm->start_code = N_TXTADDR(ex));
314 current->mm->end_data = ex.a_data +
315 (current->mm->start_data = N_DATADDR(ex));
316 current->mm->brk = ex.a_bss +
317 (current->mm->start_brk = N_BSSADDR(ex));
318 current->mm->free_area_cache = current->mm->mmap_base;
319 current->mm->cached_hole_size = 0;
321 current->mm->mmap = NULL;
322 compute_creds(bprm);
323 current->flags &= ~PF_FORKNOEXEC;
324 #ifdef __sparc__
325 if (N_MAGIC(ex) == NMAGIC) {
326 loff_t pos = fd_offset;
327 /* Fuck me plenty... */
328 /* <AOL></AOL> */
329 down_write(&current->mm->mmap_sem);
330 error = do_brk(N_TXTADDR(ex), ex.a_text);
331 up_write(&current->mm->mmap_sem);
332 bprm->file->f_op->read(bprm->file, (char *) N_TXTADDR(ex),
333 ex.a_text, &pos);
334 down_write(&current->mm->mmap_sem);
335 error = do_brk(N_DATADDR(ex), ex.a_data);
336 up_write(&current->mm->mmap_sem);
337 bprm->file->f_op->read(bprm->file, (char *) N_DATADDR(ex),
338 ex.a_data, &pos);
339 goto beyond_if;
341 #endif
343 if (N_MAGIC(ex) == OMAGIC) {
344 unsigned long text_addr, map_size;
345 loff_t pos;
347 text_addr = N_TXTADDR(ex);
349 #if defined(__alpha__) || defined(__sparc__)
350 pos = fd_offset;
351 map_size = ex.a_text+ex.a_data + PAGE_SIZE - 1;
352 #else
353 pos = 32;
354 map_size = ex.a_text+ex.a_data;
355 #endif
356 down_write(&current->mm->mmap_sem);
357 error = do_brk(text_addr & PAGE_MASK, map_size);
358 up_write(&current->mm->mmap_sem);
359 if (error != (text_addr & PAGE_MASK)) {
360 send_sig(SIGKILL, current, 0);
361 return error;
364 error = bprm->file->f_op->read(bprm->file,
365 (char __user *)text_addr,
366 ex.a_text+ex.a_data, &pos);
367 if ((signed long)error < 0) {
368 send_sig(SIGKILL, current, 0);
369 return error;
372 flush_icache_range(text_addr, text_addr+ex.a_text+ex.a_data);
373 } else {
374 static unsigned long error_time, error_time2;
375 if ((ex.a_text & 0xfff || ex.a_data & 0xfff) &&
376 (N_MAGIC(ex) != NMAGIC) && (jiffies-error_time2) > 5*HZ)
378 printk(KERN_NOTICE "executable not page aligned\n");
379 error_time2 = jiffies;
382 if ((fd_offset & ~PAGE_MASK) != 0 &&
383 (jiffies-error_time) > 5*HZ)
385 printk(KERN_WARNING
386 "fd_offset is not page aligned. Please convert program: %s\n",
387 bprm->file->f_dentry->d_name.name);
388 error_time = jiffies;
391 if (!bprm->file->f_op->mmap||((fd_offset & ~PAGE_MASK) != 0)) {
392 loff_t pos = fd_offset;
393 down_write(&current->mm->mmap_sem);
394 do_brk(N_TXTADDR(ex), ex.a_text+ex.a_data);
395 up_write(&current->mm->mmap_sem);
396 bprm->file->f_op->read(bprm->file,
397 (char __user *)N_TXTADDR(ex),
398 ex.a_text+ex.a_data, &pos);
399 flush_icache_range((unsigned long) N_TXTADDR(ex),
400 (unsigned long) N_TXTADDR(ex) +
401 ex.a_text+ex.a_data);
402 goto beyond_if;
405 down_write(&current->mm->mmap_sem);
406 error = do_mmap(bprm->file, N_TXTADDR(ex), ex.a_text,
407 PROT_READ | PROT_EXEC,
408 MAP_FIXED | MAP_PRIVATE | MAP_DENYWRITE | MAP_EXECUTABLE,
409 fd_offset);
410 up_write(&current->mm->mmap_sem);
412 if (error != N_TXTADDR(ex)) {
413 send_sig(SIGKILL, current, 0);
414 return error;
417 down_write(&current->mm->mmap_sem);
418 error = do_mmap(bprm->file, N_DATADDR(ex), ex.a_data,
419 PROT_READ | PROT_WRITE | PROT_EXEC,
420 MAP_FIXED | MAP_PRIVATE | MAP_DENYWRITE | MAP_EXECUTABLE,
421 fd_offset + ex.a_text);
422 up_write(&current->mm->mmap_sem);
423 if (error != N_DATADDR(ex)) {
424 send_sig(SIGKILL, current, 0);
425 return error;
428 beyond_if:
429 set_binfmt(&aout_format);
431 retval = set_brk(current->mm->start_brk, current->mm->brk);
432 if (retval < 0) {
433 send_sig(SIGKILL, current, 0);
434 return retval;
437 retval = setup_arg_pages(bprm, STACK_TOP, EXSTACK_DEFAULT);
438 if (retval < 0) {
439 /* Someone check-me: is this error path enough? */
440 send_sig(SIGKILL, current, 0);
441 return retval;
444 current->mm->start_stack =
445 (unsigned long) create_aout_tables((char __user *) bprm->p, bprm);
446 #ifdef __alpha__
447 regs->gp = ex.a_gpvalue;
448 #endif
449 start_thread(regs, ex.a_entry, current->mm->start_stack);
450 if (unlikely(current->ptrace & PT_PTRACED)) {
451 if (current->ptrace & PT_TRACE_EXEC)
452 ptrace_notify ((PTRACE_EVENT_EXEC << 8) | SIGTRAP);
453 else
454 send_sig(SIGTRAP, current, 0);
456 return 0;
459 static int load_aout_library(struct file *file)
461 struct inode * inode;
462 unsigned long bss, start_addr, len;
463 unsigned long error;
464 int retval;
465 struct exec ex;
467 inode = file->f_dentry->d_inode;
469 retval = -ENOEXEC;
470 error = kernel_read(file, 0, (char *) &ex, sizeof(ex));
471 if (error != sizeof(ex))
472 goto out;
474 /* We come in here for the regular a.out style of shared libraries */
475 if ((N_MAGIC(ex) != ZMAGIC && N_MAGIC(ex) != QMAGIC) || N_TRSIZE(ex) ||
476 N_DRSIZE(ex) || ((ex.a_entry & 0xfff) && N_MAGIC(ex) == ZMAGIC) ||
477 i_size_read(inode) < ex.a_text+ex.a_data+N_SYMSIZE(ex)+N_TXTOFF(ex)) {
478 goto out;
481 if (N_FLAGS(ex))
482 goto out;
484 /* For QMAGIC, the starting address is 0x20 into the page. We mask
485 this off to get the starting address for the page */
487 start_addr = ex.a_entry & 0xfffff000;
489 if ((N_TXTOFF(ex) & ~PAGE_MASK) != 0) {
490 static unsigned long error_time;
491 loff_t pos = N_TXTOFF(ex);
493 if ((jiffies-error_time) > 5*HZ)
495 printk(KERN_WARNING
496 "N_TXTOFF is not page aligned. Please convert library: %s\n",
497 file->f_dentry->d_name.name);
498 error_time = jiffies;
500 down_write(&current->mm->mmap_sem);
501 do_brk(start_addr, ex.a_text + ex.a_data + ex.a_bss);
502 up_write(&current->mm->mmap_sem);
504 file->f_op->read(file, (char __user *)start_addr,
505 ex.a_text + ex.a_data, &pos);
506 flush_icache_range((unsigned long) start_addr,
507 (unsigned long) start_addr + ex.a_text + ex.a_data);
509 retval = 0;
510 goto out;
512 /* Now use mmap to map the library into memory. */
513 down_write(&current->mm->mmap_sem);
514 error = do_mmap(file, start_addr, ex.a_text + ex.a_data,
515 PROT_READ | PROT_WRITE | PROT_EXEC,
516 MAP_FIXED | MAP_PRIVATE | MAP_DENYWRITE,
517 N_TXTOFF(ex));
518 up_write(&current->mm->mmap_sem);
519 retval = error;
520 if (error != start_addr)
521 goto out;
523 len = PAGE_ALIGN(ex.a_text + ex.a_data);
524 bss = ex.a_text + ex.a_data + ex.a_bss;
525 if (bss > len) {
526 down_write(&current->mm->mmap_sem);
527 error = do_brk(start_addr + len, bss - len);
528 up_write(&current->mm->mmap_sem);
529 retval = error;
530 if (error != start_addr + len)
531 goto out;
533 retval = 0;
534 out:
535 return retval;
538 static int __init init_aout_binfmt(void)
540 return register_binfmt(&aout_format);
543 static void __exit exit_aout_binfmt(void)
545 unregister_binfmt(&aout_format);
548 core_initcall(init_aout_binfmt);
549 module_exit(exit_aout_binfmt);
550 MODULE_LICENSE("GPL");