| 1 |
1627 |
jcastillo |
/*
|
| 2 |
|
|
* linux/fs/exec.c
|
| 3 |
|
|
*
|
| 4 |
|
|
* Copyright (C) 1991, 1992 Linus Torvalds
|
| 5 |
|
|
*/
|
| 6 |
|
|
|
| 7 |
|
|
/*
|
| 8 |
|
|
* #!-checking implemented by tytso.
|
| 9 |
|
|
*/
|
| 10 |
|
|
/*
|
| 11 |
|
|
* Demand-loading implemented 01.12.91 - no need to read anything but
|
| 12 |
|
|
* the header into memory. The inode of the executable is put into
|
| 13 |
|
|
* "current->executable", and page faults do the actual loading. Clean.
|
| 14 |
|
|
*
|
| 15 |
|
|
* Once more I can proudly say that linux stood up to being changed: it
|
| 16 |
|
|
* was less than 2 hours work to get demand-loading completely implemented.
|
| 17 |
|
|
*
|
| 18 |
|
|
* Demand loading changed July 1993 by Eric Youngdale. Use mmap instead,
|
| 19 |
|
|
* current->executable is only used by the procfs. This allows a dispatch
|
| 20 |
|
|
* table to check for several different types of binary formats. We keep
|
| 21 |
|
|
* trying until we recognize the file or we run out of supported binary
|
| 22 |
|
|
* formats.
|
| 23 |
|
|
*/
|
| 24 |
|
|
|
| 25 |
|
|
/*
|
| 26 |
|
|
* uClinux revisions for NO_MM
|
| 27 |
|
|
* Copyright (C) 1998 Kenneth Albanowski <kjahds@kjahds.com>,
|
| 28 |
|
|
* The Silver Hammer Group, Ltd.
|
| 29 |
|
|
*/
|
| 30 |
|
|
|
| 31 |
|
|
#include <linux/fs.h>
|
| 32 |
|
|
#include <linux/sched.h>
|
| 33 |
|
|
#include <linux/kernel.h>
|
| 34 |
|
|
#include <linux/mm.h>
|
| 35 |
|
|
#include <linux/mman.h>
|
| 36 |
|
|
#include <linux/a.out.h>
|
| 37 |
|
|
#include <linux/errno.h>
|
| 38 |
|
|
#include <linux/signal.h>
|
| 39 |
|
|
#include <linux/string.h>
|
| 40 |
|
|
#include <linux/stat.h>
|
| 41 |
|
|
#include <linux/fcntl.h>
|
| 42 |
|
|
#include <linux/ptrace.h>
|
| 43 |
|
|
#include <linux/user.h>
|
| 44 |
|
|
#include <linux/malloc.h>
|
| 45 |
|
|
#include <linux/binfmts.h>
|
| 46 |
|
|
#include <linux/personality.h>
|
| 47 |
|
|
|
| 48 |
|
|
#include <asm/system.h>
|
| 49 |
|
|
#include <asm/segment.h>
|
| 50 |
|
|
#include <asm/pgtable.h>
|
| 51 |
|
|
|
| 52 |
|
|
#include <linux/config.h>
|
| 53 |
|
|
#ifdef CONFIG_KERNELD
|
| 54 |
|
|
#include <linux/kerneld.h>
|
| 55 |
|
|
#endif
|
| 56 |
|
|
|
| 57 |
|
|
asmlinkage int sys_exit(int exit_code);
|
| 58 |
|
|
asmlinkage int sys_brk(unsigned long);
|
| 59 |
|
|
|
| 60 |
|
|
/*
|
| 61 |
|
|
* Here are the actual binaries that will be accepted:
|
| 62 |
|
|
* add more with "register_binfmt()" if using modules...
|
| 63 |
|
|
*
|
| 64 |
|
|
* These are defined again for the 'real' modules if you are using a
|
| 65 |
|
|
* module definition for these routines.
|
| 66 |
|
|
*/
|
| 67 |
|
|
|
| 68 |
|
|
static struct linux_binfmt *formats = (struct linux_binfmt *) NULL;
|
| 69 |
|
|
|
| 70 |
|
|
void binfmt_setup(void)
|
| 71 |
|
|
{
|
| 72 |
|
|
#ifdef CONFIG_BINFMT_FLAT
|
| 73 |
|
|
init_flat_binfmt();
|
| 74 |
|
|
|
| 75 |
|
|
#endif
|
| 76 |
|
|
|
| 77 |
|
|
#ifdef CONFIG_BINFMT_ELF
|
| 78 |
|
|
init_elf_binfmt();
|
| 79 |
|
|
#endif
|
| 80 |
|
|
|
| 81 |
|
|
#ifdef CONFIG_BINFMT_AOUT
|
| 82 |
|
|
init_aout_binfmt();
|
| 83 |
|
|
#endif
|
| 84 |
|
|
|
| 85 |
|
|
#ifdef CONFIG_BINFMT_JAVA
|
| 86 |
|
|
init_java_binfmt();
|
| 87 |
|
|
#endif
|
| 88 |
|
|
/* This cannot be configured out of the kernel */
|
| 89 |
|
|
init_script_binfmt();
|
| 90 |
|
|
}
|
| 91 |
|
|
|
| 92 |
|
|
int register_binfmt(struct linux_binfmt * fmt)
|
| 93 |
|
|
{
|
| 94 |
|
|
struct linux_binfmt ** tmp = &formats;
|
| 95 |
|
|
|
| 96 |
|
|
if (!fmt)
|
| 97 |
|
|
return -EINVAL;
|
| 98 |
|
|
if (fmt->next)
|
| 99 |
|
|
return -EBUSY;
|
| 100 |
|
|
while (*tmp) {
|
| 101 |
|
|
if (fmt == *tmp)
|
| 102 |
|
|
return -EBUSY;
|
| 103 |
|
|
tmp = &(*tmp)->next;
|
| 104 |
|
|
}
|
| 105 |
|
|
fmt->next = formats;
|
| 106 |
|
|
formats = fmt;
|
| 107 |
|
|
return 0;
|
| 108 |
|
|
}
|
| 109 |
|
|
|
| 110 |
|
|
#ifdef CONFIG_MODULES
|
| 111 |
|
|
int unregister_binfmt(struct linux_binfmt * fmt)
|
| 112 |
|
|
{
|
| 113 |
|
|
struct linux_binfmt ** tmp = &formats;
|
| 114 |
|
|
|
| 115 |
|
|
while (*tmp) {
|
| 116 |
|
|
if (fmt == *tmp) {
|
| 117 |
|
|
*tmp = fmt->next;
|
| 118 |
|
|
return 0;
|
| 119 |
|
|
}
|
| 120 |
|
|
tmp = &(*tmp)->next;
|
| 121 |
|
|
}
|
| 122 |
|
|
return -EINVAL;
|
| 123 |
|
|
}
|
| 124 |
|
|
#endif /* CONFIG_MODULES */
|
| 125 |
|
|
|
| 126 |
|
|
int open_inode(struct inode * inode, int mode)
|
| 127 |
|
|
{
|
| 128 |
|
|
int fd;
|
| 129 |
|
|
|
| 130 |
|
|
if (!inode->i_op || !inode->i_op->default_file_ops)
|
| 131 |
|
|
return -EINVAL;
|
| 132 |
|
|
fd = get_unused_fd();
|
| 133 |
|
|
if (fd >= 0) {
|
| 134 |
|
|
struct file * f = get_empty_filp();
|
| 135 |
|
|
if (!f) {
|
| 136 |
|
|
put_unused_fd(fd);
|
| 137 |
|
|
return -ENFILE;
|
| 138 |
|
|
}
|
| 139 |
|
|
f->f_flags = mode;
|
| 140 |
|
|
f->f_mode = (mode+1) & O_ACCMODE;
|
| 141 |
|
|
f->f_inode = inode;
|
| 142 |
|
|
f->f_pos = 0;
|
| 143 |
|
|
f->f_reada = 0;
|
| 144 |
|
|
f->f_op = inode->i_op->default_file_ops;
|
| 145 |
|
|
if (f->f_op->open) {
|
| 146 |
|
|
int error = f->f_op->open(inode,f);
|
| 147 |
|
|
if (error) {
|
| 148 |
|
|
f->f_count--;
|
| 149 |
|
|
put_unused_fd(fd);
|
| 150 |
|
|
return error;
|
| 151 |
|
|
}
|
| 152 |
|
|
}
|
| 153 |
|
|
current->files->fd[fd] = f;
|
| 154 |
|
|
inode->i_count++;
|
| 155 |
|
|
}
|
| 156 |
|
|
return fd;
|
| 157 |
|
|
}
|
| 158 |
|
|
|
| 159 |
|
|
/*
|
| 160 |
|
|
* Note that a shared library must be both readable and executable due to
|
| 161 |
|
|
* security reasons.
|
| 162 |
|
|
*
|
| 163 |
|
|
* Also note that we take the address to load from from the file itself.
|
| 164 |
|
|
*/
|
| 165 |
|
|
asmlinkage int sys_uselib(const char * library)
|
| 166 |
|
|
{
|
| 167 |
|
|
int fd, retval;
|
| 168 |
|
|
struct file * file;
|
| 169 |
|
|
struct linux_binfmt * fmt;
|
| 170 |
|
|
|
| 171 |
|
|
fd = sys_open(library, 0, 0);
|
| 172 |
|
|
if (fd < 0)
|
| 173 |
|
|
return fd;
|
| 174 |
|
|
file = current->files->fd[fd];
|
| 175 |
|
|
retval = -ENOEXEC;
|
| 176 |
|
|
if (file && file->f_inode && file->f_op && file->f_op->read) {
|
| 177 |
|
|
for (fmt = formats ; fmt ; fmt = fmt->next) {
|
| 178 |
|
|
int (*fn)(int) = fmt->load_shlib;
|
| 179 |
|
|
if (!fn)
|
| 180 |
|
|
continue;
|
| 181 |
|
|
retval = fn(fd);
|
| 182 |
|
|
if (retval != -ENOEXEC)
|
| 183 |
|
|
break;
|
| 184 |
|
|
}
|
| 185 |
|
|
}
|
| 186 |
|
|
sys_close(fd);
|
| 187 |
|
|
return retval;
|
| 188 |
|
|
}
|
| 189 |
|
|
|
| 190 |
|
|
/*
|
| 191 |
|
|
* count() counts the number of arguments/envelopes
|
| 192 |
|
|
*
|
| 193 |
|
|
* We also do some limited EFAULT checking: this isn't complete, but
|
| 194 |
|
|
* it does cover most cases. I'll have to do this correctly some day..
|
| 195 |
|
|
*/
|
| 196 |
|
|
static int count(char ** argv)
|
| 197 |
|
|
{
|
| 198 |
|
|
int error, i = 0;
|
| 199 |
|
|
char ** tmp, *p;
|
| 200 |
|
|
|
| 201 |
|
|
if ((tmp = argv) != NULL) {
|
| 202 |
|
|
error = verify_area(VERIFY_READ, tmp, sizeof(char *));
|
| 203 |
|
|
if (error)
|
| 204 |
|
|
return error;
|
| 205 |
|
|
while ((p = get_user(tmp++)) != NULL) {
|
| 206 |
|
|
i++;
|
| 207 |
|
|
error = verify_area(VERIFY_READ, p, 1);
|
| 208 |
|
|
if (error)
|
| 209 |
|
|
return error;
|
| 210 |
|
|
}
|
| 211 |
|
|
}
|
| 212 |
|
|
return i;
|
| 213 |
|
|
}
|
| 214 |
|
|
|
| 215 |
|
|
#ifndef NO_MM
|
| 216 |
|
|
/*
|
| 217 |
|
|
* 'copy_string()' copies argument/envelope strings from user
|
| 218 |
|
|
* memory to free pages in kernel mem. These are in a format ready
|
| 219 |
|
|
* to be put directly into the top of new user memory.
|
| 220 |
|
|
*
|
| 221 |
|
|
* Modified by TYT, 11/24/91 to add the from_kmem argument, which specifies
|
| 222 |
|
|
* whether the string and the string array are from user or kernel segments:
|
| 223 |
|
|
*
|
| 224 |
|
|
* from_kmem argv * argv **
|
| 225 |
|
|
* 0 user space user space
|
| 226 |
|
|
* 1 kernel space user space
|
| 227 |
|
|
* 2 kernel space kernel space
|
| 228 |
|
|
*
|
| 229 |
|
|
* We do this by playing games with the fs segment register. Since it
|
| 230 |
|
|
* is expensive to load a segment register, we try to avoid calling
|
| 231 |
|
|
* set_fs() unless we absolutely have to.
|
| 232 |
|
|
*/
|
| 233 |
|
|
unsigned long copy_strings(int argc,char ** argv,unsigned long *page,
|
| 234 |
|
|
unsigned long p, int from_kmem)
|
| 235 |
|
|
{
|
| 236 |
|
|
char *tmp, *tmp1, *pag = NULL;
|
| 237 |
|
|
int len, offset = 0;
|
| 238 |
|
|
unsigned long old_fs, new_fs;
|
| 239 |
|
|
|
| 240 |
|
|
if (!p)
|
| 241 |
|
|
return 0; /* bullet-proofing */
|
| 242 |
|
|
new_fs = get_ds();
|
| 243 |
|
|
old_fs = get_fs();
|
| 244 |
|
|
if (from_kmem==2)
|
| 245 |
|
|
set_fs(new_fs);
|
| 246 |
|
|
while (argc-- > 0) {
|
| 247 |
|
|
if (from_kmem == 1)
|
| 248 |
|
|
set_fs(new_fs);
|
| 249 |
|
|
if (!(tmp1 = tmp = get_user(argv+argc)))
|
| 250 |
|
|
panic("VFS: argc is wrong");
|
| 251 |
|
|
if (from_kmem == 1)
|
| 252 |
|
|
set_fs(old_fs);
|
| 253 |
|
|
while (get_user(tmp++));
|
| 254 |
|
|
len = tmp - tmp1;
|
| 255 |
|
|
if (p < len) { /* this shouldn't happen - 128kB */
|
| 256 |
|
|
set_fs(old_fs);
|
| 257 |
|
|
return 0;
|
| 258 |
|
|
}
|
| 259 |
|
|
while (len) {
|
| 260 |
|
|
--p; --tmp; --len;
|
| 261 |
|
|
if (--offset < 0) {
|
| 262 |
|
|
offset = p % PAGE_SIZE;
|
| 263 |
|
|
if (from_kmem==2)
|
| 264 |
|
|
set_fs(old_fs);
|
| 265 |
|
|
if (!(pag = (char *) page[p/PAGE_SIZE]) &&
|
| 266 |
|
|
!(pag = (char *) page[p/PAGE_SIZE] =
|
| 267 |
|
|
(unsigned long *) get_free_page(GFP_USER)))
|
| 268 |
|
|
return 0;
|
| 269 |
|
|
if (from_kmem==2)
|
| 270 |
|
|
set_fs(new_fs);
|
| 271 |
|
|
|
| 272 |
|
|
}
|
| 273 |
|
|
if (len == 0 || offset == 0)
|
| 274 |
|
|
*(pag + offset) = get_user(tmp);
|
| 275 |
|
|
else {
|
| 276 |
|
|
int bytes_to_copy = (len > offset) ? offset : len;
|
| 277 |
|
|
tmp -= bytes_to_copy;
|
| 278 |
|
|
p -= bytes_to_copy;
|
| 279 |
|
|
offset -= bytes_to_copy;
|
| 280 |
|
|
len -= bytes_to_copy;
|
| 281 |
|
|
memcpy_fromfs(pag + offset, tmp, bytes_to_copy + 1);
|
| 282 |
|
|
}
|
| 283 |
|
|
}
|
| 284 |
|
|
}
|
| 285 |
|
|
if (from_kmem==2)
|
| 286 |
|
|
set_fs(old_fs);
|
| 287 |
|
|
return p;
|
| 288 |
|
|
}
|
| 289 |
|
|
|
| 290 |
|
|
unsigned long setup_arg_pages(unsigned long p, struct linux_binprm * bprm)
|
| 291 |
|
|
{
|
| 292 |
|
|
unsigned long stack_base;
|
| 293 |
|
|
struct vm_area_struct *mpnt;
|
| 294 |
|
|
int i;
|
| 295 |
|
|
|
| 296 |
|
|
stack_base = STACK_TOP - MAX_ARG_PAGES*PAGE_SIZE;
|
| 297 |
|
|
|
| 298 |
|
|
p += stack_base;
|
| 299 |
|
|
if (bprm->loader)
|
| 300 |
|
|
bprm->loader += stack_base;
|
| 301 |
|
|
bprm->exec += stack_base;
|
| 302 |
|
|
|
| 303 |
|
|
mpnt = (struct vm_area_struct *)kmalloc(sizeof(*mpnt), GFP_KERNEL);
|
| 304 |
|
|
if (mpnt) {
|
| 305 |
|
|
mpnt->vm_mm = current->mm;
|
| 306 |
|
|
mpnt->vm_start = PAGE_MASK & (unsigned long) p;
|
| 307 |
|
|
mpnt->vm_end = STACK_TOP;
|
| 308 |
|
|
mpnt->vm_page_prot = PAGE_COPY;
|
| 309 |
|
|
mpnt->vm_flags = VM_STACK_FLAGS;
|
| 310 |
|
|
mpnt->vm_ops = NULL;
|
| 311 |
|
|
mpnt->vm_offset = 0;
|
| 312 |
|
|
mpnt->vm_inode = NULL;
|
| 313 |
|
|
mpnt->vm_pte = 0;
|
| 314 |
|
|
insert_vm_struct(current->mm, mpnt);
|
| 315 |
|
|
current->mm->total_vm = (mpnt->vm_end - mpnt->vm_start) >> PAGE_SHIFT;
|
| 316 |
|
|
|
| 317 |
|
|
for (i = 0 ; i < MAX_ARG_PAGES ; i++) {
|
| 318 |
|
|
if (bprm->page[i]) {
|
| 319 |
|
|
current->mm->rss++;
|
| 320 |
|
|
put_dirty_page(current,bprm->page[i],stack_base);
|
| 321 |
|
|
}
|
| 322 |
|
|
stack_base += PAGE_SIZE;
|
| 323 |
|
|
}
|
| 324 |
|
|
} else {
|
| 325 |
|
|
/*
|
| 326 |
|
|
* This one is tricky. We are already in the new context, so we cannot
|
| 327 |
|
|
* return with -ENOMEM. So we _have_ to deallocate argument pages here,
|
| 328 |
|
|
* if there is no VMA, they wont be freed at exit_mmap() -> memory leak.
|
| 329 |
|
|
*
|
| 330 |
|
|
* User space then gets a SIGSEGV when it tries to access argument pages.
|
| 331 |
|
|
*/
|
| 332 |
|
|
for (i = 0 ; i < MAX_ARG_PAGES ; i++) {
|
| 333 |
|
|
if (bprm->page[i]) {
|
| 334 |
|
|
free_page(bprm->page[i]);
|
| 335 |
|
|
bprm->page[i] = 0;
|
| 336 |
|
|
}
|
| 337 |
|
|
}
|
| 338 |
|
|
}
|
| 339 |
|
|
|
| 340 |
|
|
return p;
|
| 341 |
|
|
}
|
| 342 |
|
|
#endif /* !NO_MM */
|
| 343 |
|
|
|
| 344 |
|
|
/*
|
| 345 |
|
|
* Read in the complete executable. This is used for "-N" files
|
| 346 |
|
|
* that aren't on a block boundary, and for files on filesystems
|
| 347 |
|
|
* without bmap support.
|
| 348 |
|
|
*/
|
| 349 |
|
|
int read_exec(struct inode *inode, unsigned long offset,
|
| 350 |
|
|
char * addr, unsigned long count, int to_kmem)
|
| 351 |
|
|
{
|
| 352 |
|
|
struct file file;
|
| 353 |
|
|
int result = -ENOEXEC;
|
| 354 |
|
|
|
| 355 |
|
|
if (!inode->i_op || !inode->i_op->default_file_ops)
|
| 356 |
|
|
goto end_readexec;
|
| 357 |
|
|
file.f_mode = 1;
|
| 358 |
|
|
file.f_flags = 0;
|
| 359 |
|
|
file.f_count = 1;
|
| 360 |
|
|
file.f_inode = inode;
|
| 361 |
|
|
file.f_pos = 0;
|
| 362 |
|
|
file.f_reada = 0;
|
| 363 |
|
|
file.f_op = inode->i_op->default_file_ops;
|
| 364 |
|
|
if (file.f_op->open)
|
| 365 |
|
|
if (file.f_op->open(inode,&file))
|
| 366 |
|
|
goto end_readexec;
|
| 367 |
|
|
if (!file.f_op || !file.f_op->read)
|
| 368 |
|
|
goto close_readexec;
|
| 369 |
|
|
if (file.f_op->lseek) {
|
| 370 |
|
|
if (file.f_op->lseek(inode,&file,offset,0) != offset)
|
| 371 |
|
|
goto close_readexec;
|
| 372 |
|
|
} else
|
| 373 |
|
|
file.f_pos = offset;
|
| 374 |
|
|
if (to_kmem) {
|
| 375 |
|
|
unsigned long old_fs = get_fs();
|
| 376 |
|
|
set_fs(get_ds());
|
| 377 |
|
|
result = file.f_op->read(inode, &file, addr, count);
|
| 378 |
|
|
set_fs(old_fs);
|
| 379 |
|
|
} else {
|
| 380 |
|
|
result = verify_area(VERIFY_WRITE, addr, count);
|
| 381 |
|
|
if (result)
|
| 382 |
|
|
goto close_readexec;
|
| 383 |
|
|
result = file.f_op->read(inode, &file, addr, count);
|
| 384 |
|
|
}
|
| 385 |
|
|
close_readexec:
|
| 386 |
|
|
if (file.f_op->release)
|
| 387 |
|
|
file.f_op->release(inode,&file);
|
| 388 |
|
|
end_readexec:
|
| 389 |
|
|
return result;
|
| 390 |
|
|
}
|
| 391 |
|
|
|
| 392 |
|
|
#ifndef NO_MM
|
| 393 |
|
|
static int exec_mmap(void)
|
| 394 |
|
|
{
|
| 395 |
|
|
/*
|
| 396 |
|
|
* The clear_page_tables done later on exec does the right thing
|
| 397 |
|
|
* to the page directory when shared, except for graceful abort
|
| 398 |
|
|
*/
|
| 399 |
|
|
if (current->mm->count > 1) {
|
| 400 |
|
|
struct mm_struct *old_mm, *mm = kmalloc(sizeof(*mm), GFP_KERNEL);
|
| 401 |
|
|
if (!mm)
|
| 402 |
|
|
return -ENOMEM;
|
| 403 |
|
|
|
| 404 |
|
|
*mm = *current->mm;
|
| 405 |
|
|
mm->def_flags = 0; /* should future lockings be kept? */
|
| 406 |
|
|
mm->count = 1;
|
| 407 |
|
|
mm->mmap = NULL;
|
| 408 |
|
|
mm->mmap_avl = NULL;
|
| 409 |
|
|
mm->total_vm = 0;
|
| 410 |
|
|
mm->rss = 0;
|
| 411 |
|
|
|
| 412 |
|
|
old_mm = current->mm;
|
| 413 |
|
|
current->mm = mm;
|
| 414 |
|
|
if (new_page_tables(current)) {
|
| 415 |
|
|
/* The pgd belongs to the parent ... don't free it! */
|
| 416 |
|
|
mm->pgd = NULL;
|
| 417 |
|
|
current->mm = old_mm;
|
| 418 |
|
|
exit_mmap(mm);
|
| 419 |
|
|
kfree(mm);
|
| 420 |
|
|
return -ENOMEM;
|
| 421 |
|
|
}
|
| 422 |
|
|
|
| 423 |
|
|
if ((old_mm != &init_mm) && (!--old_mm->count)) {
|
| 424 |
|
|
/*
|
| 425 |
|
|
* all threads exited while we were sleeping, 'old_mm' is held
|
| 426 |
|
|
* by us exclusively, lets get rid of it:
|
| 427 |
|
|
*/
|
| 428 |
|
|
exit_mmap(old_mm);
|
| 429 |
|
|
free_page_tables(old_mm);
|
| 430 |
|
|
kfree(old_mm);
|
| 431 |
|
|
}
|
| 432 |
|
|
|
| 433 |
|
|
return 0;
|
| 434 |
|
|
}
|
| 435 |
|
|
flush_cache_mm(current->mm);
|
| 436 |
|
|
exit_mmap(current->mm);
|
| 437 |
|
|
clear_page_tables(current);
|
| 438 |
|
|
flush_tlb_mm(current->mm);
|
| 439 |
|
|
|
| 440 |
|
|
return 0;
|
| 441 |
|
|
}
|
| 442 |
|
|
#else /* NO_MM */
|
| 443 |
|
|
static int exec_mmap(void)
|
| 444 |
|
|
{
|
| 445 |
|
|
if (current->mm->count > 1) {
|
| 446 |
|
|
struct mm_struct *old_mm, *mm = kmalloc(sizeof(*mm), GFP_KERNEL);
|
| 447 |
|
|
if (!mm)
|
| 448 |
|
|
return -ENOMEM;
|
| 449 |
|
|
|
| 450 |
|
|
*mm = *current->mm;
|
| 451 |
|
|
mm->def_flags = 0; /* should future lockings be kept? */
|
| 452 |
|
|
mm->count = 1;
|
| 453 |
|
|
mm->total_vm = 0;
|
| 454 |
|
|
mm->rss = 0;
|
| 455 |
|
|
mm->executable = 0;
|
| 456 |
|
|
mm->tblock.rblock = 0;
|
| 457 |
|
|
mm->tblock.next = 0;
|
| 458 |
|
|
|
| 459 |
|
|
old_mm = current->mm;
|
| 460 |
|
|
current->mm = mm;
|
| 461 |
|
|
|
| 462 |
|
|
if ((old_mm != &init_mm) && (!--old_mm->count)) {
|
| 463 |
|
|
/*
|
| 464 |
|
|
* all threads exited while we were sleeping, 'old_mm' is held
|
| 465 |
|
|
* by us exclusively, lets get rid of it:
|
| 466 |
|
|
*/
|
| 467 |
|
|
exit_mmap(old_mm);
|
| 468 |
|
|
kfree(old_mm);
|
| 469 |
|
|
}
|
| 470 |
|
|
|
| 471 |
|
|
return 0;
|
| 472 |
|
|
}
|
| 473 |
|
|
exit_mmap(current->mm);
|
| 474 |
|
|
|
| 475 |
|
|
return 0;
|
| 476 |
|
|
}
|
| 477 |
|
|
|
| 478 |
|
|
#endif /* NO_MM */
|
| 479 |
|
|
|
| 480 |
|
|
/*
|
| 481 |
|
|
* These functions flushes out all traces of the currently running executable
|
| 482 |
|
|
* so that a new one can be started
|
| 483 |
|
|
*/
|
| 484 |
|
|
|
| 485 |
|
|
static inline void flush_old_signals(struct signal_struct *sig)
|
| 486 |
|
|
{
|
| 487 |
|
|
int i;
|
| 488 |
|
|
struct sigaction * sa = sig->action;
|
| 489 |
|
|
|
| 490 |
|
|
for (i=32 ; i != 0 ; i--) {
|
| 491 |
|
|
sa->sa_mask = 0;
|
| 492 |
|
|
sa->sa_flags = 0;
|
| 493 |
|
|
if (sa->sa_handler != SIG_IGN)
|
| 494 |
|
|
sa->sa_handler = NULL;
|
| 495 |
|
|
sa++;
|
| 496 |
|
|
}
|
| 497 |
|
|
}
|
| 498 |
|
|
|
| 499 |
|
|
static inline void flush_old_files(struct files_struct * files)
|
| 500 |
|
|
{
|
| 501 |
|
|
unsigned long j;
|
| 502 |
|
|
|
| 503 |
|
|
j = 0;
|
| 504 |
|
|
for (;;) {
|
| 505 |
|
|
unsigned long set, i;
|
| 506 |
|
|
|
| 507 |
|
|
i = j * __NFDBITS;
|
| 508 |
|
|
if (i >= NR_OPEN)
|
| 509 |
|
|
break;
|
| 510 |
|
|
set = files->close_on_exec.fds_bits[j];
|
| 511 |
|
|
files->close_on_exec.fds_bits[j] = 0;
|
| 512 |
|
|
j++;
|
| 513 |
|
|
for ( ; set ; i++,set >>= 1) {
|
| 514 |
|
|
if (set & 1)
|
| 515 |
|
|
sys_close(i);
|
| 516 |
|
|
}
|
| 517 |
|
|
}
|
| 518 |
|
|
}
|
| 519 |
|
|
|
| 520 |
|
|
int flush_old_exec(struct linux_binprm * bprm)
|
| 521 |
|
|
{
|
| 522 |
|
|
int i;
|
| 523 |
|
|
int ch;
|
| 524 |
|
|
char * name;
|
| 525 |
|
|
|
| 526 |
|
|
if (current->euid == current->uid && current->egid == current->gid)
|
| 527 |
|
|
current->dumpable = 1;
|
| 528 |
|
|
name = bprm->filename;
|
| 529 |
|
|
for (i=0; (ch = *(name++)) != '\0';) {
|
| 530 |
|
|
if (ch == '/')
|
| 531 |
|
|
i = 0;
|
| 532 |
|
|
else
|
| 533 |
|
|
if (i < 15)
|
| 534 |
|
|
current->comm[i++] = ch;
|
| 535 |
|
|
}
|
| 536 |
|
|
current->comm[i] = '\0';
|
| 537 |
|
|
|
| 538 |
|
|
/* Release all of the old mmap stuff. */
|
| 539 |
|
|
if (exec_mmap())
|
| 540 |
|
|
return -ENOMEM;
|
| 541 |
|
|
|
| 542 |
|
|
flush_thread();
|
| 543 |
|
|
|
| 544 |
|
|
if (bprm->e_uid != current->euid || bprm->e_gid != current->egid ||
|
| 545 |
|
|
permission(bprm->inode,MAY_READ))
|
| 546 |
|
|
current->dumpable = 0;
|
| 547 |
|
|
|
| 548 |
|
|
flush_old_signals(current->sig);
|
| 549 |
|
|
flush_old_files(current->files);
|
| 550 |
|
|
|
| 551 |
|
|
return 0;
|
| 552 |
|
|
}
|
| 553 |
|
|
|
| 554 |
|
|
/*
|
| 555 |
|
|
* Fill the binprm structure from the inode.
|
| 556 |
|
|
* Check permissions, then read the first 512 bytes
|
| 557 |
|
|
*/
|
| 558 |
|
|
int prepare_binprm(struct linux_binprm *bprm)
|
| 559 |
|
|
{
|
| 560 |
|
|
int mode;
|
| 561 |
|
|
int retval,id_change;
|
| 562 |
|
|
|
| 563 |
|
|
mode = bprm->inode->i_mode;
|
| 564 |
|
|
if (!S_ISREG(mode)) /* must be regular file */
|
| 565 |
|
|
return -EACCES;
|
| 566 |
|
|
if (!(mode & 0111)) /* with at least _one_ execute bit set */
|
| 567 |
|
|
return -EACCES;
|
| 568 |
|
|
if (IS_NOEXEC(bprm->inode)) /* FS mustn't be mounted noexec */
|
| 569 |
|
|
return -EACCES;
|
| 570 |
|
|
if (!bprm->inode->i_sb)
|
| 571 |
|
|
return -EACCES;
|
| 572 |
|
|
if ((retval = permission(bprm->inode, MAY_EXEC)) != 0)
|
| 573 |
|
|
return retval;
|
| 574 |
|
|
/* better not execute files which are being written to */
|
| 575 |
|
|
if (bprm->inode->i_writecount > 0)
|
| 576 |
|
|
return -ETXTBSY;
|
| 577 |
|
|
|
| 578 |
|
|
bprm->e_uid = current->euid;
|
| 579 |
|
|
bprm->e_gid = current->egid;
|
| 580 |
|
|
id_change = 0;
|
| 581 |
|
|
|
| 582 |
|
|
/* Set-uid? */
|
| 583 |
|
|
if (mode & S_ISUID) {
|
| 584 |
|
|
bprm->e_uid = bprm->inode->i_uid;
|
| 585 |
|
|
if (bprm->e_uid != current->euid)
|
| 586 |
|
|
id_change = 1;
|
| 587 |
|
|
}
|
| 588 |
|
|
|
| 589 |
|
|
/* Set-gid? */
|
| 590 |
|
|
/*
|
| 591 |
|
|
* If setgid is set but no group execute bit then this
|
| 592 |
|
|
* is a candidate for mandatory locking, not a setgid
|
| 593 |
|
|
* executable.
|
| 594 |
|
|
*/
|
| 595 |
|
|
if ((mode & (S_ISGID | S_IXGRP)) == (S_ISGID | S_IXGRP)) {
|
| 596 |
|
|
bprm->e_gid = bprm->inode->i_gid;
|
| 597 |
|
|
if (!in_group_p(bprm->e_gid))
|
| 598 |
|
|
id_change = 1;
|
| 599 |
|
|
}
|
| 600 |
|
|
|
| 601 |
|
|
if (id_change) {
|
| 602 |
|
|
/* We can't suid-execute if we're sharing parts of the executable */
|
| 603 |
|
|
/* or if we're being traced (or if suid execs are not allowed) */
|
| 604 |
|
|
/* (current->mm->count > 1 is ok, as we'll get a new mm anyway) */
|
| 605 |
|
|
if (IS_NOSUID(bprm->inode)
|
| 606 |
|
|
|| (current->flags & PF_PTRACED)
|
| 607 |
|
|
|| (current->fs->count > 1)
|
| 608 |
|
|
|| (current->sig->count > 1)
|
| 609 |
|
|
|| (current->files->count > 1)) {
|
| 610 |
|
|
if (!suser())
|
| 611 |
|
|
return -EPERM;
|
| 612 |
|
|
}
|
| 613 |
|
|
}
|
| 614 |
|
|
|
| 615 |
|
|
memset(bprm->buf,0,sizeof(bprm->buf));
|
| 616 |
|
|
return read_exec(bprm->inode,0,bprm->buf,128,1);
|
| 617 |
|
|
}
|
| 618 |
|
|
|
| 619 |
|
|
#ifndef NO_MM
|
| 620 |
|
|
void remove_arg_zero(struct linux_binprm *bprm)
|
| 621 |
|
|
{
|
| 622 |
|
|
if (bprm->argc) {
|
| 623 |
|
|
unsigned long offset;
|
| 624 |
|
|
char * page;
|
| 625 |
|
|
offset = bprm->p % PAGE_SIZE;
|
| 626 |
|
|
page = (char*)bprm->page[bprm->p/PAGE_SIZE];
|
| 627 |
|
|
while(bprm->p++,*(page+offset++))
|
| 628 |
|
|
if(offset==PAGE_SIZE){
|
| 629 |
|
|
offset=0;
|
| 630 |
|
|
page = (char*)bprm->page[bprm->p/PAGE_SIZE];
|
| 631 |
|
|
}
|
| 632 |
|
|
bprm->argc--;
|
| 633 |
|
|
}
|
| 634 |
|
|
}
|
| 635 |
|
|
#endif /* !NO_MM */
|
| 636 |
|
|
|
| 637 |
|
|
/*
|
| 638 |
|
|
* cycle the list of binary formats handler, until one recognizes the image
|
| 639 |
|
|
*/
|
| 640 |
|
|
int search_binary_handler(struct linux_binprm *bprm,struct pt_regs *regs)
|
| 641 |
|
|
{
|
| 642 |
|
|
int try,retval=0;
|
| 643 |
|
|
struct linux_binfmt *fmt;
|
| 644 |
|
|
#ifdef __alpha__
|
| 645 |
|
|
/* handle /sbin/loader.. */
|
| 646 |
|
|
{
|
| 647 |
|
|
struct exec * eh = (struct exec *) bprm->buf;
|
| 648 |
|
|
|
| 649 |
|
|
if (!bprm->loader && eh->fh.f_magic == 0x183 &&
|
| 650 |
|
|
(eh->fh.f_flags & 0x3000) == 0x3000)
|
| 651 |
|
|
{
|
| 652 |
|
|
char * dynloader[] = { "/sbin/loader" };
|
| 653 |
|
|
iput(bprm->inode);
|
| 654 |
|
|
bprm->dont_iput = 1;
|
| 655 |
|
|
remove_arg_zero(bprm);
|
| 656 |
|
|
bprm->p = copy_strings(1, dynloader, bprm->page, bprm->p, 2);
|
| 657 |
|
|
bprm->argc++;
|
| 658 |
|
|
bprm->loader = bprm->p;
|
| 659 |
|
|
retval = open_namei(dynloader[0], 0, 0, &bprm->inode, NULL);
|
| 660 |
|
|
if (retval)
|
| 661 |
|
|
return retval;
|
| 662 |
|
|
bprm->dont_iput = 0;
|
| 663 |
|
|
retval = prepare_binprm(bprm);
|
| 664 |
|
|
if (retval<0)
|
| 665 |
|
|
return retval;
|
| 666 |
|
|
/* should call search_binary_handler recursively here,
|
| 667 |
|
|
but it does not matter */
|
| 668 |
|
|
}
|
| 669 |
|
|
}
|
| 670 |
|
|
#endif
|
| 671 |
|
|
for (try=0; try<2; try++) {
|
| 672 |
|
|
for (fmt = formats ; fmt ; fmt = fmt->next) {
|
| 673 |
|
|
int (*fn)(struct linux_binprm *, struct pt_regs *) = fmt->load_binary;
|
| 674 |
|
|
if (!fn)
|
| 675 |
|
|
continue;
|
| 676 |
|
|
retval = fn(bprm, regs);
|
| 677 |
|
|
if (retval >= 0) {
|
| 678 |
|
|
if(!bprm->dont_iput)
|
| 679 |
|
|
iput(bprm->inode);
|
| 680 |
|
|
bprm->dont_iput=1;
|
| 681 |
|
|
current->did_exec = 1;
|
| 682 |
|
|
return retval;
|
| 683 |
|
|
}
|
| 684 |
|
|
if (retval != -ENOEXEC)
|
| 685 |
|
|
break;
|
| 686 |
|
|
if (bprm->dont_iput) /* We don't have the inode anymore*/
|
| 687 |
|
|
return retval;
|
| 688 |
|
|
}
|
| 689 |
|
|
if (retval != -ENOEXEC) {
|
| 690 |
|
|
break;
|
| 691 |
|
|
#ifdef CONFIG_KERNELD
|
| 692 |
|
|
}else{
|
| 693 |
|
|
#define printable(c) (((c)=='\t') || ((c)=='\n') || (0x20<=(c) && (c)<=0x7e))
|
| 694 |
|
|
char modname[20];
|
| 695 |
|
|
if (printable(bprm->buf[0]) &&
|
| 696 |
|
|
printable(bprm->buf[1]) &&
|
| 697 |
|
|
printable(bprm->buf[2]) &&
|
| 698 |
|
|
printable(bprm->buf[3]))
|
| 699 |
|
|
break; /* -ENOEXEC */
|
| 700 |
|
|
sprintf(modname, "binfmt-%hd", *(short*)(&bprm->buf));
|
| 701 |
|
|
request_module(modname);
|
| 702 |
|
|
#endif
|
| 703 |
|
|
}
|
| 704 |
|
|
}
|
| 705 |
|
|
return retval;
|
| 706 |
|
|
}
|
| 707 |
|
|
|
| 708 |
|
|
|
| 709 |
|
|
/*
|
| 710 |
|
|
* sys_execve() executes a new program.
|
| 711 |
|
|
*/
|
| 712 |
|
|
int do_execve(char * filename, char ** argv, char ** envp, struct pt_regs * regs)
|
| 713 |
|
|
{
|
| 714 |
|
|
struct linux_binprm bprm;
|
| 715 |
|
|
int retval;
|
| 716 |
|
|
#ifndef NO_MM
|
| 717 |
|
|
int i;
|
| 718 |
|
|
#endif /* !NO_MM */
|
| 719 |
|
|
|
| 720 |
|
|
#ifndef NO_MM
|
| 721 |
|
|
bprm.p = PAGE_SIZE*MAX_ARG_PAGES-sizeof(void *);
|
| 722 |
|
|
for (i=0 ; i<MAX_ARG_PAGES ; i++) /* clear page-table */
|
| 723 |
|
|
bprm.page[i] = 0;
|
| 724 |
|
|
#endif /* !NO_MM */
|
| 725 |
|
|
retval = open_namei(filename, 0, 0, &bprm.inode, NULL);
|
| 726 |
|
|
if (retval)
|
| 727 |
|
|
return retval;
|
| 728 |
|
|
bprm.filename = filename;
|
| 729 |
|
|
bprm.sh_bang = 0;
|
| 730 |
|
|
bprm.loader = 0;
|
| 731 |
|
|
bprm.exec = 0;
|
| 732 |
|
|
bprm.dont_iput = 0;
|
| 733 |
|
|
if ((bprm.argc = count(argv)) < 0)
|
| 734 |
|
|
return bprm.argc;
|
| 735 |
|
|
if ((bprm.envc = count(envp)) < 0)
|
| 736 |
|
|
return bprm.envc;
|
| 737 |
|
|
|
| 738 |
|
|
retval = prepare_binprm(&bprm);
|
| 739 |
|
|
|
| 740 |
|
|
#ifndef NO_MM
|
| 741 |
|
|
if(retval>=0) {
|
| 742 |
|
|
bprm.p = copy_strings(1, &bprm.filename, bprm.page, bprm.p, 2);
|
| 743 |
|
|
bprm.exec = bprm.p;
|
| 744 |
|
|
bprm.p = copy_strings(bprm.envc,envp,bprm.page,bprm.p,0);
|
| 745 |
|
|
bprm.p = copy_strings(bprm.argc,argv,bprm.page,bprm.p,0);
|
| 746 |
|
|
if (!bprm.p)
|
| 747 |
|
|
retval = -E2BIG;
|
| 748 |
|
|
}
|
| 749 |
|
|
#else /* NO_MM */
|
| 750 |
|
|
bprm.envp = envp;
|
| 751 |
|
|
bprm.argv = argv;
|
| 752 |
|
|
#endif /* NO_MM */
|
| 753 |
|
|
|
| 754 |
|
|
if(retval>=0)
|
| 755 |
|
|
retval = search_binary_handler(&bprm,regs);
|
| 756 |
|
|
if(retval>=0) {
|
| 757 |
|
|
#ifdef NO_MM
|
| 758 |
|
|
/* Wake up parent that vforked me */
|
| 759 |
|
|
wake_up(¤t->p_opptr->mm->vforkwait);
|
| 760 |
|
|
#endif /* NO_MM */
|
| 761 |
|
|
/* execve success */
|
| 762 |
|
|
return retval;
|
| 763 |
|
|
}
|
| 764 |
|
|
|
| 765 |
|
|
/* Something went wrong, return the inode and free the argument pages*/
|
| 766 |
|
|
if(!bprm.dont_iput)
|
| 767 |
|
|
iput(bprm.inode);
|
| 768 |
|
|
#ifndef NO_MM
|
| 769 |
|
|
for (i=0 ; i<MAX_ARG_PAGES ; i++)
|
| 770 |
|
|
free_page(bprm.page[i]);
|
| 771 |
|
|
#endif /* !NO_MM */
|
| 772 |
|
|
return(retval);
|
| 773 |
|
|
}
|
