| Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1 | /* | 
 | 2 |  * linux/ipc/shm.c | 
 | 3 |  * Copyright (C) 1992, 1993 Krishna Balasubramanian | 
 | 4 |  *	 Many improvements/fixes by Bruno Haible. | 
 | 5 |  * Replaced `struct shm_desc' by `struct vm_area_struct', July 1994. | 
 | 6 |  * Fixed the shm swap deallocation (shm_unuse()), August 1998 Andrea Arcangeli. | 
 | 7 |  * | 
 | 8 |  * /proc/sysvipc/shm support (c) 1999 Dragos Acostachioaie <dragos@iname.com> | 
 | 9 |  * BIGMEM support, Andrea Arcangeli <andrea@suse.de> | 
 | 10 |  * SMP thread shm, Jean-Luc Boyard <jean-luc.boyard@siemens.fr> | 
 | 11 |  * HIGHMEM support, Ingo Molnar <mingo@redhat.com> | 
 | 12 |  * Make shmmax, shmall, shmmni sysctl'able, Christoph Rohland <cr@sap.com> | 
 | 13 |  * Shared /dev/zero support, Kanoj Sarcar <kanoj@sgi.com> | 
 | 14 |  * Move the mm functionality over to mm/shmem.c, Christoph Rohland <cr@sap.com> | 
 | 15 |  * | 
 | 16 |  */ | 
 | 17 |  | 
 | 18 | #include <linux/config.h> | 
 | 19 | #include <linux/slab.h> | 
 | 20 | #include <linux/mm.h> | 
 | 21 | #include <linux/hugetlb.h> | 
 | 22 | #include <linux/shm.h> | 
 | 23 | #include <linux/init.h> | 
 | 24 | #include <linux/file.h> | 
 | 25 | #include <linux/mman.h> | 
 | 26 | #include <linux/proc_fs.h> | 
 | 27 | #include <linux/shmem_fs.h> | 
 | 28 | #include <linux/security.h> | 
 | 29 | #include <linux/syscalls.h> | 
 | 30 | #include <linux/audit.h> | 
| Stephen Rothwell | 7d87e14 | 2005-05-01 08:59:12 -0700 | [diff] [blame] | 31 | #include <linux/ptrace.h> | 
 | 32 |  | 
| Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 33 | #include <asm/uaccess.h> | 
 | 34 |  | 
 | 35 | #include "util.h" | 
 | 36 |  | 
 | 37 | #define shm_flags	shm_perm.mode | 
 | 38 |  | 
 | 39 | static struct file_operations shm_file_operations; | 
 | 40 | static struct vm_operations_struct shm_vm_ops; | 
 | 41 |  | 
 | 42 | static struct ipc_ids shm_ids; | 
 | 43 |  | 
 | 44 | #define shm_lock(id)	((struct shmid_kernel*)ipc_lock(&shm_ids,id)) | 
 | 45 | #define shm_unlock(shp)	ipc_unlock(&(shp)->shm_perm) | 
 | 46 | #define shm_get(id)	((struct shmid_kernel*)ipc_get(&shm_ids,id)) | 
 | 47 | #define shm_buildid(id, seq) \ | 
 | 48 | 	ipc_buildid(&shm_ids, id, seq) | 
 | 49 |  | 
 | 50 | static int newseg (key_t key, int shmflg, size_t size); | 
 | 51 | static void shm_open (struct vm_area_struct *shmd); | 
 | 52 | static void shm_close (struct vm_area_struct *shmd); | 
 | 53 | #ifdef CONFIG_PROC_FS | 
 | 54 | static int sysvipc_shm_read_proc(char *buffer, char **start, off_t offset, int length, int *eof, void *data); | 
 | 55 | #endif | 
 | 56 |  | 
 | 57 | size_t	shm_ctlmax = SHMMAX; | 
 | 58 | size_t 	shm_ctlall = SHMALL; | 
 | 59 | int 	shm_ctlmni = SHMMNI; | 
 | 60 |  | 
 | 61 | static int shm_tot; /* total number of shared memory pages */ | 
 | 62 |  | 
 | 63 | void __init shm_init (void) | 
 | 64 | { | 
 | 65 | 	ipc_init_ids(&shm_ids, 1); | 
 | 66 | #ifdef CONFIG_PROC_FS | 
 | 67 | 	create_proc_read_entry("sysvipc/shm", 0, NULL, sysvipc_shm_read_proc, NULL); | 
 | 68 | #endif | 
 | 69 | } | 
 | 70 |  | 
 | 71 | static inline int shm_checkid(struct shmid_kernel *s, int id) | 
 | 72 | { | 
 | 73 | 	if (ipc_checkid(&shm_ids,&s->shm_perm,id)) | 
 | 74 | 		return -EIDRM; | 
 | 75 | 	return 0; | 
 | 76 | } | 
 | 77 |  | 
 | 78 | static inline struct shmid_kernel *shm_rmid(int id) | 
 | 79 | { | 
 | 80 | 	return (struct shmid_kernel *)ipc_rmid(&shm_ids,id); | 
 | 81 | } | 
 | 82 |  | 
 | 83 | static inline int shm_addid(struct shmid_kernel *shp) | 
 | 84 | { | 
 | 85 | 	return ipc_addid(&shm_ids, &shp->shm_perm, shm_ctlmni); | 
 | 86 | } | 
 | 87 |  | 
 | 88 |  | 
 | 89 |  | 
 | 90 | static inline void shm_inc (int id) { | 
 | 91 | 	struct shmid_kernel *shp; | 
 | 92 |  | 
 | 93 | 	if(!(shp = shm_lock(id))) | 
 | 94 | 		BUG(); | 
 | 95 | 	shp->shm_atim = get_seconds(); | 
 | 96 | 	shp->shm_lprid = current->tgid; | 
 | 97 | 	shp->shm_nattch++; | 
 | 98 | 	shm_unlock(shp); | 
 | 99 | } | 
 | 100 |  | 
 | 101 | /* This is called by fork, once for every shm attach. */ | 
 | 102 | static void shm_open (struct vm_area_struct *shmd) | 
 | 103 | { | 
 | 104 | 	shm_inc (shmd->vm_file->f_dentry->d_inode->i_ino); | 
 | 105 | } | 
 | 106 |  | 
 | 107 | /* | 
 | 108 |  * shm_destroy - free the struct shmid_kernel | 
 | 109 |  * | 
 | 110 |  * @shp: struct to free | 
 | 111 |  * | 
 | 112 |  * It has to be called with shp and shm_ids.sem locked, | 
 | 113 |  * but returns with shp unlocked and freed. | 
 | 114 |  */ | 
 | 115 | static void shm_destroy (struct shmid_kernel *shp) | 
 | 116 | { | 
 | 117 | 	shm_tot -= (shp->shm_segsz + PAGE_SIZE - 1) >> PAGE_SHIFT; | 
 | 118 | 	shm_rmid (shp->id); | 
 | 119 | 	shm_unlock(shp); | 
 | 120 | 	if (!is_file_hugepages(shp->shm_file)) | 
 | 121 | 		shmem_lock(shp->shm_file, 0, shp->mlock_user); | 
 | 122 | 	else | 
 | 123 | 		user_shm_unlock(shp->shm_file->f_dentry->d_inode->i_size, | 
 | 124 | 						shp->mlock_user); | 
 | 125 | 	fput (shp->shm_file); | 
 | 126 | 	security_shm_free(shp); | 
 | 127 | 	ipc_rcu_putref(shp); | 
 | 128 | } | 
 | 129 |  | 
 | 130 | /* | 
 | 131 |  * remove the attach descriptor shmd. | 
 | 132 |  * free memory for segment if it is marked destroyed. | 
 | 133 |  * The descriptor has already been removed from the current->mm->mmap list | 
 | 134 |  * and will later be kfree()d. | 
 | 135 |  */ | 
 | 136 | static void shm_close (struct vm_area_struct *shmd) | 
 | 137 | { | 
 | 138 | 	struct file * file = shmd->vm_file; | 
 | 139 | 	int id = file->f_dentry->d_inode->i_ino; | 
 | 140 | 	struct shmid_kernel *shp; | 
 | 141 |  | 
 | 142 | 	down (&shm_ids.sem); | 
 | 143 | 	/* remove from the list of attaches of the shm segment */ | 
 | 144 | 	if(!(shp = shm_lock(id))) | 
 | 145 | 		BUG(); | 
 | 146 | 	shp->shm_lprid = current->tgid; | 
 | 147 | 	shp->shm_dtim = get_seconds(); | 
 | 148 | 	shp->shm_nattch--; | 
 | 149 | 	if(shp->shm_nattch == 0 && | 
 | 150 | 	   shp->shm_flags & SHM_DEST) | 
 | 151 | 		shm_destroy (shp); | 
 | 152 | 	else | 
 | 153 | 		shm_unlock(shp); | 
 | 154 | 	up (&shm_ids.sem); | 
 | 155 | } | 
 | 156 |  | 
 | 157 | static int shm_mmap(struct file * file, struct vm_area_struct * vma) | 
 | 158 | { | 
 | 159 | 	file_accessed(file); | 
 | 160 | 	vma->vm_ops = &shm_vm_ops; | 
 | 161 | 	shm_inc(file->f_dentry->d_inode->i_ino); | 
 | 162 | 	return 0; | 
 | 163 | } | 
 | 164 |  | 
 | 165 | static struct file_operations shm_file_operations = { | 
 | 166 | 	.mmap	= shm_mmap | 
 | 167 | }; | 
 | 168 |  | 
 | 169 | static struct vm_operations_struct shm_vm_ops = { | 
 | 170 | 	.open	= shm_open,	/* callback for a new vm-area open */ | 
 | 171 | 	.close	= shm_close,	/* callback for when the vm-area is released */ | 
 | 172 | 	.nopage	= shmem_nopage, | 
| Andrew Morton | 6ade43f | 2005-08-01 21:11:45 -0700 | [diff] [blame] | 173 | #if defined(CONFIG_NUMA) && defined(CONFIG_SHMEM) | 
| Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 174 | 	.set_policy = shmem_set_policy, | 
 | 175 | 	.get_policy = shmem_get_policy, | 
 | 176 | #endif | 
 | 177 | }; | 
 | 178 |  | 
 | 179 | static int newseg (key_t key, int shmflg, size_t size) | 
 | 180 | { | 
 | 181 | 	int error; | 
 | 182 | 	struct shmid_kernel *shp; | 
 | 183 | 	int numpages = (size + PAGE_SIZE -1) >> PAGE_SHIFT; | 
 | 184 | 	struct file * file; | 
 | 185 | 	char name[13]; | 
 | 186 | 	int id; | 
 | 187 |  | 
 | 188 | 	if (size < SHMMIN || size > shm_ctlmax) | 
 | 189 | 		return -EINVAL; | 
 | 190 |  | 
 | 191 | 	if (shm_tot + numpages >= shm_ctlall) | 
 | 192 | 		return -ENOSPC; | 
 | 193 |  | 
 | 194 | 	shp = ipc_rcu_alloc(sizeof(*shp)); | 
 | 195 | 	if (!shp) | 
 | 196 | 		return -ENOMEM; | 
 | 197 |  | 
 | 198 | 	shp->shm_perm.key = key; | 
 | 199 | 	shp->shm_flags = (shmflg & S_IRWXUGO); | 
 | 200 | 	shp->mlock_user = NULL; | 
 | 201 |  | 
 | 202 | 	shp->shm_perm.security = NULL; | 
 | 203 | 	error = security_shm_alloc(shp); | 
 | 204 | 	if (error) { | 
 | 205 | 		ipc_rcu_putref(shp); | 
 | 206 | 		return error; | 
 | 207 | 	} | 
 | 208 |  | 
 | 209 | 	if (shmflg & SHM_HUGETLB) { | 
 | 210 | 		/* hugetlb_zero_setup takes care of mlock user accounting */ | 
 | 211 | 		file = hugetlb_zero_setup(size); | 
 | 212 | 		shp->mlock_user = current->user; | 
 | 213 | 	} else { | 
 | 214 | 		sprintf (name, "SYSV%08x", key); | 
 | 215 | 		file = shmem_file_setup(name, size, VM_ACCOUNT); | 
 | 216 | 	} | 
 | 217 | 	error = PTR_ERR(file); | 
 | 218 | 	if (IS_ERR(file)) | 
 | 219 | 		goto no_file; | 
 | 220 |  | 
 | 221 | 	error = -ENOSPC; | 
 | 222 | 	id = shm_addid(shp); | 
 | 223 | 	if(id == -1)  | 
 | 224 | 		goto no_id; | 
 | 225 |  | 
 | 226 | 	shp->shm_cprid = current->tgid; | 
 | 227 | 	shp->shm_lprid = 0; | 
 | 228 | 	shp->shm_atim = shp->shm_dtim = 0; | 
 | 229 | 	shp->shm_ctim = get_seconds(); | 
 | 230 | 	shp->shm_segsz = size; | 
 | 231 | 	shp->shm_nattch = 0; | 
 | 232 | 	shp->id = shm_buildid(id,shp->shm_perm.seq); | 
 | 233 | 	shp->shm_file = file; | 
 | 234 | 	file->f_dentry->d_inode->i_ino = shp->id; | 
 | 235 | 	if (shmflg & SHM_HUGETLB) | 
 | 236 | 		set_file_hugepages(file); | 
 | 237 | 	else | 
 | 238 | 		file->f_op = &shm_file_operations; | 
 | 239 | 	shm_tot += numpages; | 
 | 240 | 	shm_unlock(shp); | 
 | 241 | 	return shp->id; | 
 | 242 |  | 
 | 243 | no_id: | 
 | 244 | 	fput(file); | 
 | 245 | no_file: | 
 | 246 | 	security_shm_free(shp); | 
 | 247 | 	ipc_rcu_putref(shp); | 
 | 248 | 	return error; | 
 | 249 | } | 
 | 250 |  | 
 | 251 | asmlinkage long sys_shmget (key_t key, size_t size, int shmflg) | 
 | 252 | { | 
 | 253 | 	struct shmid_kernel *shp; | 
 | 254 | 	int err, id = 0; | 
 | 255 |  | 
 | 256 | 	down(&shm_ids.sem); | 
 | 257 | 	if (key == IPC_PRIVATE) { | 
 | 258 | 		err = newseg(key, shmflg, size); | 
 | 259 | 	} else if ((id = ipc_findkey(&shm_ids, key)) == -1) { | 
 | 260 | 		if (!(shmflg & IPC_CREAT)) | 
 | 261 | 			err = -ENOENT; | 
 | 262 | 		else | 
 | 263 | 			err = newseg(key, shmflg, size); | 
 | 264 | 	} else if ((shmflg & IPC_CREAT) && (shmflg & IPC_EXCL)) { | 
 | 265 | 		err = -EEXIST; | 
 | 266 | 	} else { | 
 | 267 | 		shp = shm_lock(id); | 
 | 268 | 		if(shp==NULL) | 
 | 269 | 			BUG(); | 
 | 270 | 		if (shp->shm_segsz < size) | 
 | 271 | 			err = -EINVAL; | 
 | 272 | 		else if (ipcperms(&shp->shm_perm, shmflg)) | 
 | 273 | 			err = -EACCES; | 
 | 274 | 		else { | 
 | 275 | 			int shmid = shm_buildid(id, shp->shm_perm.seq); | 
 | 276 | 			err = security_shm_associate(shp, shmflg); | 
 | 277 | 			if (!err) | 
 | 278 | 				err = shmid; | 
 | 279 | 		} | 
 | 280 | 		shm_unlock(shp); | 
 | 281 | 	} | 
 | 282 | 	up(&shm_ids.sem); | 
 | 283 |  | 
 | 284 | 	return err; | 
 | 285 | } | 
 | 286 |  | 
 | 287 | static inline unsigned long copy_shmid_to_user(void __user *buf, struct shmid64_ds *in, int version) | 
 | 288 | { | 
 | 289 | 	switch(version) { | 
 | 290 | 	case IPC_64: | 
 | 291 | 		return copy_to_user(buf, in, sizeof(*in)); | 
 | 292 | 	case IPC_OLD: | 
 | 293 | 	    { | 
 | 294 | 		struct shmid_ds out; | 
 | 295 |  | 
 | 296 | 		ipc64_perm_to_ipc_perm(&in->shm_perm, &out.shm_perm); | 
 | 297 | 		out.shm_segsz	= in->shm_segsz; | 
 | 298 | 		out.shm_atime	= in->shm_atime; | 
 | 299 | 		out.shm_dtime	= in->shm_dtime; | 
 | 300 | 		out.shm_ctime	= in->shm_ctime; | 
 | 301 | 		out.shm_cpid	= in->shm_cpid; | 
 | 302 | 		out.shm_lpid	= in->shm_lpid; | 
 | 303 | 		out.shm_nattch	= in->shm_nattch; | 
 | 304 |  | 
 | 305 | 		return copy_to_user(buf, &out, sizeof(out)); | 
 | 306 | 	    } | 
 | 307 | 	default: | 
 | 308 | 		return -EINVAL; | 
 | 309 | 	} | 
 | 310 | } | 
 | 311 |  | 
 | 312 | struct shm_setbuf { | 
 | 313 | 	uid_t	uid; | 
 | 314 | 	gid_t	gid; | 
 | 315 | 	mode_t	mode; | 
 | 316 | };	 | 
 | 317 |  | 
 | 318 | static inline unsigned long copy_shmid_from_user(struct shm_setbuf *out, void __user *buf, int version) | 
 | 319 | { | 
 | 320 | 	switch(version) { | 
 | 321 | 	case IPC_64: | 
 | 322 | 	    { | 
 | 323 | 		struct shmid64_ds tbuf; | 
 | 324 |  | 
 | 325 | 		if (copy_from_user(&tbuf, buf, sizeof(tbuf))) | 
 | 326 | 			return -EFAULT; | 
 | 327 |  | 
 | 328 | 		out->uid	= tbuf.shm_perm.uid; | 
 | 329 | 		out->gid	= tbuf.shm_perm.gid; | 
 | 330 | 		out->mode	= tbuf.shm_flags; | 
 | 331 |  | 
 | 332 | 		return 0; | 
 | 333 | 	    } | 
 | 334 | 	case IPC_OLD: | 
 | 335 | 	    { | 
 | 336 | 		struct shmid_ds tbuf_old; | 
 | 337 |  | 
 | 338 | 		if (copy_from_user(&tbuf_old, buf, sizeof(tbuf_old))) | 
 | 339 | 			return -EFAULT; | 
 | 340 |  | 
 | 341 | 		out->uid	= tbuf_old.shm_perm.uid; | 
 | 342 | 		out->gid	= tbuf_old.shm_perm.gid; | 
 | 343 | 		out->mode	= tbuf_old.shm_flags; | 
 | 344 |  | 
 | 345 | 		return 0; | 
 | 346 | 	    } | 
 | 347 | 	default: | 
 | 348 | 		return -EINVAL; | 
 | 349 | 	} | 
 | 350 | } | 
 | 351 |  | 
 | 352 | static inline unsigned long copy_shminfo_to_user(void __user *buf, struct shminfo64 *in, int version) | 
 | 353 | { | 
 | 354 | 	switch(version) { | 
 | 355 | 	case IPC_64: | 
 | 356 | 		return copy_to_user(buf, in, sizeof(*in)); | 
 | 357 | 	case IPC_OLD: | 
 | 358 | 	    { | 
 | 359 | 		struct shminfo out; | 
 | 360 |  | 
 | 361 | 		if(in->shmmax > INT_MAX) | 
 | 362 | 			out.shmmax = INT_MAX; | 
 | 363 | 		else | 
 | 364 | 			out.shmmax = (int)in->shmmax; | 
 | 365 |  | 
 | 366 | 		out.shmmin	= in->shmmin; | 
 | 367 | 		out.shmmni	= in->shmmni; | 
 | 368 | 		out.shmseg	= in->shmseg; | 
 | 369 | 		out.shmall	= in->shmall;  | 
 | 370 |  | 
 | 371 | 		return copy_to_user(buf, &out, sizeof(out)); | 
 | 372 | 	    } | 
 | 373 | 	default: | 
 | 374 | 		return -EINVAL; | 
 | 375 | 	} | 
 | 376 | } | 
 | 377 |  | 
 | 378 | static void shm_get_stat(unsigned long *rss, unsigned long *swp)  | 
 | 379 | { | 
 | 380 | 	int i; | 
 | 381 |  | 
 | 382 | 	*rss = 0; | 
 | 383 | 	*swp = 0; | 
 | 384 |  | 
 | 385 | 	for (i = 0; i <= shm_ids.max_id; i++) { | 
 | 386 | 		struct shmid_kernel *shp; | 
 | 387 | 		struct inode *inode; | 
 | 388 |  | 
 | 389 | 		shp = shm_get(i); | 
 | 390 | 		if(!shp) | 
 | 391 | 			continue; | 
 | 392 |  | 
 | 393 | 		inode = shp->shm_file->f_dentry->d_inode; | 
 | 394 |  | 
 | 395 | 		if (is_file_hugepages(shp->shm_file)) { | 
 | 396 | 			struct address_space *mapping = inode->i_mapping; | 
 | 397 | 			*rss += (HPAGE_SIZE/PAGE_SIZE)*mapping->nrpages; | 
 | 398 | 		} else { | 
 | 399 | 			struct shmem_inode_info *info = SHMEM_I(inode); | 
 | 400 | 			spin_lock(&info->lock); | 
 | 401 | 			*rss += inode->i_mapping->nrpages; | 
 | 402 | 			*swp += info->swapped; | 
 | 403 | 			spin_unlock(&info->lock); | 
 | 404 | 		} | 
 | 405 | 	} | 
 | 406 | } | 
 | 407 |  | 
 | 408 | asmlinkage long sys_shmctl (int shmid, int cmd, struct shmid_ds __user *buf) | 
 | 409 | { | 
 | 410 | 	struct shm_setbuf setbuf; | 
 | 411 | 	struct shmid_kernel *shp; | 
 | 412 | 	int err, version; | 
 | 413 |  | 
 | 414 | 	if (cmd < 0 || shmid < 0) { | 
 | 415 | 		err = -EINVAL; | 
 | 416 | 		goto out; | 
 | 417 | 	} | 
 | 418 |  | 
 | 419 | 	version = ipc_parse_version(&cmd); | 
 | 420 |  | 
 | 421 | 	switch (cmd) { /* replace with proc interface ? */ | 
 | 422 | 	case IPC_INFO: | 
 | 423 | 	{ | 
 | 424 | 		struct shminfo64 shminfo; | 
 | 425 |  | 
 | 426 | 		err = security_shm_shmctl(NULL, cmd); | 
 | 427 | 		if (err) | 
 | 428 | 			return err; | 
 | 429 |  | 
 | 430 | 		memset(&shminfo,0,sizeof(shminfo)); | 
 | 431 | 		shminfo.shmmni = shminfo.shmseg = shm_ctlmni; | 
 | 432 | 		shminfo.shmmax = shm_ctlmax; | 
 | 433 | 		shminfo.shmall = shm_ctlall; | 
 | 434 |  | 
 | 435 | 		shminfo.shmmin = SHMMIN; | 
 | 436 | 		if(copy_shminfo_to_user (buf, &shminfo, version)) | 
 | 437 | 			return -EFAULT; | 
 | 438 | 		/* reading a integer is always atomic */ | 
 | 439 | 		err= shm_ids.max_id; | 
 | 440 | 		if(err<0) | 
 | 441 | 			err = 0; | 
 | 442 | 		goto out; | 
 | 443 | 	} | 
 | 444 | 	case SHM_INFO: | 
 | 445 | 	{ | 
 | 446 | 		struct shm_info shm_info; | 
 | 447 |  | 
 | 448 | 		err = security_shm_shmctl(NULL, cmd); | 
 | 449 | 		if (err) | 
 | 450 | 			return err; | 
 | 451 |  | 
 | 452 | 		memset(&shm_info,0,sizeof(shm_info)); | 
 | 453 | 		down(&shm_ids.sem); | 
 | 454 | 		shm_info.used_ids = shm_ids.in_use; | 
 | 455 | 		shm_get_stat (&shm_info.shm_rss, &shm_info.shm_swp); | 
 | 456 | 		shm_info.shm_tot = shm_tot; | 
 | 457 | 		shm_info.swap_attempts = 0; | 
 | 458 | 		shm_info.swap_successes = 0; | 
 | 459 | 		err = shm_ids.max_id; | 
 | 460 | 		up(&shm_ids.sem); | 
 | 461 | 		if(copy_to_user (buf, &shm_info, sizeof(shm_info))) { | 
 | 462 | 			err = -EFAULT; | 
 | 463 | 			goto out; | 
 | 464 | 		} | 
 | 465 |  | 
 | 466 | 		err = err < 0 ? 0 : err; | 
 | 467 | 		goto out; | 
 | 468 | 	} | 
 | 469 | 	case SHM_STAT: | 
 | 470 | 	case IPC_STAT: | 
 | 471 | 	{ | 
 | 472 | 		struct shmid64_ds tbuf; | 
 | 473 | 		int result; | 
 | 474 | 		memset(&tbuf, 0, sizeof(tbuf)); | 
 | 475 | 		shp = shm_lock(shmid); | 
 | 476 | 		if(shp==NULL) { | 
 | 477 | 			err = -EINVAL; | 
 | 478 | 			goto out; | 
 | 479 | 		} else if(cmd==SHM_STAT) { | 
 | 480 | 			err = -EINVAL; | 
 | 481 | 			if (shmid > shm_ids.max_id) | 
 | 482 | 				goto out_unlock; | 
 | 483 | 			result = shm_buildid(shmid, shp->shm_perm.seq); | 
 | 484 | 		} else { | 
 | 485 | 			err = shm_checkid(shp,shmid); | 
 | 486 | 			if(err) | 
 | 487 | 				goto out_unlock; | 
 | 488 | 			result = 0; | 
 | 489 | 		} | 
 | 490 | 		err=-EACCES; | 
 | 491 | 		if (ipcperms (&shp->shm_perm, S_IRUGO)) | 
 | 492 | 			goto out_unlock; | 
 | 493 | 		err = security_shm_shmctl(shp, cmd); | 
 | 494 | 		if (err) | 
 | 495 | 			goto out_unlock; | 
 | 496 | 		kernel_to_ipc64_perm(&shp->shm_perm, &tbuf.shm_perm); | 
 | 497 | 		tbuf.shm_segsz	= shp->shm_segsz; | 
 | 498 | 		tbuf.shm_atime	= shp->shm_atim; | 
 | 499 | 		tbuf.shm_dtime	= shp->shm_dtim; | 
 | 500 | 		tbuf.shm_ctime	= shp->shm_ctim; | 
 | 501 | 		tbuf.shm_cpid	= shp->shm_cprid; | 
 | 502 | 		tbuf.shm_lpid	= shp->shm_lprid; | 
 | 503 | 		if (!is_file_hugepages(shp->shm_file)) | 
 | 504 | 			tbuf.shm_nattch	= shp->shm_nattch; | 
 | 505 | 		else | 
 | 506 | 			tbuf.shm_nattch = file_count(shp->shm_file) - 1; | 
 | 507 | 		shm_unlock(shp); | 
 | 508 | 		if(copy_shmid_to_user (buf, &tbuf, version)) | 
 | 509 | 			err = -EFAULT; | 
 | 510 | 		else | 
 | 511 | 			err = result; | 
 | 512 | 		goto out; | 
 | 513 | 	} | 
 | 514 | 	case SHM_LOCK: | 
 | 515 | 	case SHM_UNLOCK: | 
 | 516 | 	{ | 
 | 517 | 		shp = shm_lock(shmid); | 
 | 518 | 		if(shp==NULL) { | 
 | 519 | 			err = -EINVAL; | 
 | 520 | 			goto out; | 
 | 521 | 		} | 
 | 522 | 		err = shm_checkid(shp,shmid); | 
 | 523 | 		if(err) | 
 | 524 | 			goto out_unlock; | 
 | 525 |  | 
 | 526 | 		if (!capable(CAP_IPC_LOCK)) { | 
 | 527 | 			err = -EPERM; | 
 | 528 | 			if (current->euid != shp->shm_perm.uid && | 
 | 529 | 			    current->euid != shp->shm_perm.cuid) | 
 | 530 | 				goto out_unlock; | 
 | 531 | 			if (cmd == SHM_LOCK && | 
 | 532 | 			    !current->signal->rlim[RLIMIT_MEMLOCK].rlim_cur) | 
 | 533 | 				goto out_unlock; | 
 | 534 | 		} | 
 | 535 |  | 
 | 536 | 		err = security_shm_shmctl(shp, cmd); | 
 | 537 | 		if (err) | 
 | 538 | 			goto out_unlock; | 
 | 539 | 		 | 
 | 540 | 		if(cmd==SHM_LOCK) { | 
 | 541 | 			struct user_struct * user = current->user; | 
 | 542 | 			if (!is_file_hugepages(shp->shm_file)) { | 
 | 543 | 				err = shmem_lock(shp->shm_file, 1, user); | 
 | 544 | 				if (!err) { | 
 | 545 | 					shp->shm_flags |= SHM_LOCKED; | 
 | 546 | 					shp->mlock_user = user; | 
 | 547 | 				} | 
 | 548 | 			} | 
 | 549 | 		} else if (!is_file_hugepages(shp->shm_file)) { | 
 | 550 | 			shmem_lock(shp->shm_file, 0, shp->mlock_user); | 
 | 551 | 			shp->shm_flags &= ~SHM_LOCKED; | 
 | 552 | 			shp->mlock_user = NULL; | 
 | 553 | 		} | 
 | 554 | 		shm_unlock(shp); | 
 | 555 | 		goto out; | 
 | 556 | 	} | 
 | 557 | 	case IPC_RMID: | 
 | 558 | 	{ | 
 | 559 | 		/* | 
 | 560 | 		 *	We cannot simply remove the file. The SVID states | 
 | 561 | 		 *	that the block remains until the last person | 
 | 562 | 		 *	detaches from it, then is deleted. A shmat() on | 
 | 563 | 		 *	an RMID segment is legal in older Linux and if  | 
 | 564 | 		 *	we change it apps break... | 
 | 565 | 		 * | 
 | 566 | 		 *	Instead we set a destroyed flag, and then blow | 
 | 567 | 		 *	the name away when the usage hits zero. | 
 | 568 | 		 */ | 
 | 569 | 		down(&shm_ids.sem); | 
 | 570 | 		shp = shm_lock(shmid); | 
 | 571 | 		err = -EINVAL; | 
 | 572 | 		if (shp == NULL)  | 
 | 573 | 			goto out_up; | 
 | 574 | 		err = shm_checkid(shp, shmid); | 
 | 575 | 		if(err) | 
 | 576 | 			goto out_unlock_up; | 
 | 577 |  | 
 | 578 | 		if (current->euid != shp->shm_perm.uid && | 
 | 579 | 		    current->euid != shp->shm_perm.cuid &&  | 
 | 580 | 		    !capable(CAP_SYS_ADMIN)) { | 
 | 581 | 			err=-EPERM; | 
 | 582 | 			goto out_unlock_up; | 
 | 583 | 		} | 
 | 584 |  | 
 | 585 | 		err = security_shm_shmctl(shp, cmd); | 
 | 586 | 		if (err) | 
 | 587 | 			goto out_unlock_up; | 
 | 588 |  | 
 | 589 | 		if (shp->shm_nattch){ | 
 | 590 | 			shp->shm_flags |= SHM_DEST; | 
 | 591 | 			/* Do not find it any more */ | 
 | 592 | 			shp->shm_perm.key = IPC_PRIVATE; | 
 | 593 | 			shm_unlock(shp); | 
 | 594 | 		} else | 
 | 595 | 			shm_destroy (shp); | 
 | 596 | 		up(&shm_ids.sem); | 
 | 597 | 		goto out; | 
 | 598 | 	} | 
 | 599 |  | 
 | 600 | 	case IPC_SET: | 
 | 601 | 	{ | 
 | 602 | 		if (copy_shmid_from_user (&setbuf, buf, version)) { | 
 | 603 | 			err = -EFAULT; | 
 | 604 | 			goto out; | 
 | 605 | 		} | 
 | 606 | 		if ((err = audit_ipc_perms(0, setbuf.uid, setbuf.gid, setbuf.mode))) | 
 | 607 | 			return err; | 
 | 608 | 		down(&shm_ids.sem); | 
 | 609 | 		shp = shm_lock(shmid); | 
 | 610 | 		err=-EINVAL; | 
 | 611 | 		if(shp==NULL) | 
 | 612 | 			goto out_up; | 
 | 613 | 		err = shm_checkid(shp,shmid); | 
 | 614 | 		if(err) | 
 | 615 | 			goto out_unlock_up; | 
 | 616 | 		err=-EPERM; | 
 | 617 | 		if (current->euid != shp->shm_perm.uid && | 
 | 618 | 		    current->euid != shp->shm_perm.cuid &&  | 
 | 619 | 		    !capable(CAP_SYS_ADMIN)) { | 
 | 620 | 			goto out_unlock_up; | 
 | 621 | 		} | 
 | 622 |  | 
 | 623 | 		err = security_shm_shmctl(shp, cmd); | 
 | 624 | 		if (err) | 
 | 625 | 			goto out_unlock_up; | 
 | 626 | 		 | 
 | 627 | 		shp->shm_perm.uid = setbuf.uid; | 
 | 628 | 		shp->shm_perm.gid = setbuf.gid; | 
 | 629 | 		shp->shm_flags = (shp->shm_flags & ~S_IRWXUGO) | 
 | 630 | 			| (setbuf.mode & S_IRWXUGO); | 
 | 631 | 		shp->shm_ctim = get_seconds(); | 
 | 632 | 		break; | 
 | 633 | 	} | 
 | 634 |  | 
 | 635 | 	default: | 
 | 636 | 		err = -EINVAL; | 
 | 637 | 		goto out; | 
 | 638 | 	} | 
 | 639 |  | 
 | 640 | 	err = 0; | 
 | 641 | out_unlock_up: | 
 | 642 | 	shm_unlock(shp); | 
 | 643 | out_up: | 
 | 644 | 	up(&shm_ids.sem); | 
 | 645 | 	goto out; | 
 | 646 | out_unlock: | 
 | 647 | 	shm_unlock(shp); | 
 | 648 | out: | 
 | 649 | 	return err; | 
 | 650 | } | 
 | 651 |  | 
 | 652 | /* | 
 | 653 |  * Fix shmaddr, allocate descriptor, map shm, add attach descriptor to lists. | 
 | 654 |  * | 
 | 655 |  * NOTE! Despite the name, this is NOT a direct system call entrypoint. The | 
 | 656 |  * "raddr" thing points to kernel space, and there has to be a wrapper around | 
 | 657 |  * this. | 
 | 658 |  */ | 
 | 659 | long do_shmat(int shmid, char __user *shmaddr, int shmflg, ulong *raddr) | 
 | 660 | { | 
 | 661 | 	struct shmid_kernel *shp; | 
 | 662 | 	unsigned long addr; | 
 | 663 | 	unsigned long size; | 
 | 664 | 	struct file * file; | 
 | 665 | 	int    err; | 
 | 666 | 	unsigned long flags; | 
 | 667 | 	unsigned long prot; | 
 | 668 | 	unsigned long o_flags; | 
 | 669 | 	int acc_mode; | 
 | 670 | 	void *user_addr; | 
 | 671 |  | 
 | 672 | 	if (shmid < 0) { | 
 | 673 | 		err = -EINVAL; | 
 | 674 | 		goto out; | 
 | 675 | 	} else if ((addr = (ulong)shmaddr)) { | 
 | 676 | 		if (addr & (SHMLBA-1)) { | 
 | 677 | 			if (shmflg & SHM_RND) | 
 | 678 | 				addr &= ~(SHMLBA-1);	   /* round down */ | 
 | 679 | 			else | 
 | 680 | #ifndef __ARCH_FORCE_SHMLBA | 
 | 681 | 				if (addr & ~PAGE_MASK) | 
 | 682 | #endif | 
 | 683 | 					return -EINVAL; | 
 | 684 | 		} | 
 | 685 | 		flags = MAP_SHARED | MAP_FIXED; | 
 | 686 | 	} else { | 
 | 687 | 		if ((shmflg & SHM_REMAP)) | 
 | 688 | 			return -EINVAL; | 
 | 689 |  | 
 | 690 | 		flags = MAP_SHARED; | 
 | 691 | 	} | 
 | 692 |  | 
 | 693 | 	if (shmflg & SHM_RDONLY) { | 
 | 694 | 		prot = PROT_READ; | 
 | 695 | 		o_flags = O_RDONLY; | 
 | 696 | 		acc_mode = S_IRUGO; | 
 | 697 | 	} else { | 
 | 698 | 		prot = PROT_READ | PROT_WRITE; | 
 | 699 | 		o_flags = O_RDWR; | 
 | 700 | 		acc_mode = S_IRUGO | S_IWUGO; | 
 | 701 | 	} | 
 | 702 | 	if (shmflg & SHM_EXEC) { | 
 | 703 | 		prot |= PROT_EXEC; | 
 | 704 | 		acc_mode |= S_IXUGO; | 
 | 705 | 	} | 
 | 706 |  | 
 | 707 | 	/* | 
 | 708 | 	 * We cannot rely on the fs check since SYSV IPC does have an | 
 | 709 | 	 * additional creator id... | 
 | 710 | 	 */ | 
 | 711 | 	shp = shm_lock(shmid); | 
 | 712 | 	if(shp == NULL) { | 
 | 713 | 		err = -EINVAL; | 
 | 714 | 		goto out; | 
 | 715 | 	} | 
 | 716 | 	err = shm_checkid(shp,shmid); | 
 | 717 | 	if (err) { | 
 | 718 | 		shm_unlock(shp); | 
 | 719 | 		goto out; | 
 | 720 | 	} | 
 | 721 | 	if (ipcperms(&shp->shm_perm, acc_mode)) { | 
 | 722 | 		shm_unlock(shp); | 
 | 723 | 		err = -EACCES; | 
 | 724 | 		goto out; | 
 | 725 | 	} | 
 | 726 |  | 
 | 727 | 	err = security_shm_shmat(shp, shmaddr, shmflg); | 
 | 728 | 	if (err) { | 
 | 729 | 		shm_unlock(shp); | 
 | 730 | 		return err; | 
 | 731 | 	} | 
 | 732 | 		 | 
 | 733 | 	file = shp->shm_file; | 
 | 734 | 	size = i_size_read(file->f_dentry->d_inode); | 
 | 735 | 	shp->shm_nattch++; | 
 | 736 | 	shm_unlock(shp); | 
 | 737 |  | 
 | 738 | 	down_write(¤t->mm->mmap_sem); | 
 | 739 | 	if (addr && !(shmflg & SHM_REMAP)) { | 
 | 740 | 		user_addr = ERR_PTR(-EINVAL); | 
 | 741 | 		if (find_vma_intersection(current->mm, addr, addr + size)) | 
 | 742 | 			goto invalid; | 
 | 743 | 		/* | 
 | 744 | 		 * If shm segment goes below stack, make sure there is some | 
 | 745 | 		 * space left for the stack to grow (at least 4 pages). | 
 | 746 | 		 */ | 
 | 747 | 		if (addr < current->mm->start_stack && | 
 | 748 | 		    addr > current->mm->start_stack - size - PAGE_SIZE * 5) | 
 | 749 | 			goto invalid; | 
 | 750 | 	} | 
 | 751 | 		 | 
 | 752 | 	user_addr = (void*) do_mmap (file, addr, size, prot, flags, 0); | 
 | 753 |  | 
 | 754 | invalid: | 
 | 755 | 	up_write(¤t->mm->mmap_sem); | 
 | 756 |  | 
 | 757 | 	down (&shm_ids.sem); | 
 | 758 | 	if(!(shp = shm_lock(shmid))) | 
 | 759 | 		BUG(); | 
 | 760 | 	shp->shm_nattch--; | 
 | 761 | 	if(shp->shm_nattch == 0 && | 
 | 762 | 	   shp->shm_flags & SHM_DEST) | 
 | 763 | 		shm_destroy (shp); | 
 | 764 | 	else | 
 | 765 | 		shm_unlock(shp); | 
 | 766 | 	up (&shm_ids.sem); | 
 | 767 |  | 
 | 768 | 	*raddr = (unsigned long) user_addr; | 
 | 769 | 	err = 0; | 
 | 770 | 	if (IS_ERR(user_addr)) | 
 | 771 | 		err = PTR_ERR(user_addr); | 
 | 772 | out: | 
 | 773 | 	return err; | 
 | 774 | } | 
 | 775 |  | 
| Stephen Rothwell | 7d87e14 | 2005-05-01 08:59:12 -0700 | [diff] [blame] | 776 | asmlinkage long sys_shmat(int shmid, char __user *shmaddr, int shmflg) | 
 | 777 | { | 
 | 778 | 	unsigned long ret; | 
 | 779 | 	long err; | 
 | 780 |  | 
 | 781 | 	err = do_shmat(shmid, shmaddr, shmflg, &ret); | 
 | 782 | 	if (err) | 
 | 783 | 		return err; | 
 | 784 | 	force_successful_syscall_return(); | 
 | 785 | 	return (long)ret; | 
 | 786 | } | 
 | 787 |  | 
| Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 788 | /* | 
 | 789 |  * detach and kill segment if marked destroyed. | 
 | 790 |  * The work is done in shm_close. | 
 | 791 |  */ | 
 | 792 | asmlinkage long sys_shmdt(char __user *shmaddr) | 
 | 793 | { | 
 | 794 | 	struct mm_struct *mm = current->mm; | 
 | 795 | 	struct vm_area_struct *vma, *next; | 
 | 796 | 	unsigned long addr = (unsigned long)shmaddr; | 
 | 797 | 	loff_t size = 0; | 
 | 798 | 	int retval = -EINVAL; | 
 | 799 |  | 
 | 800 | 	down_write(&mm->mmap_sem); | 
 | 801 |  | 
 | 802 | 	/* | 
 | 803 | 	 * This function tries to be smart and unmap shm segments that | 
 | 804 | 	 * were modified by partial mlock or munmap calls: | 
 | 805 | 	 * - It first determines the size of the shm segment that should be | 
 | 806 | 	 *   unmapped: It searches for a vma that is backed by shm and that | 
 | 807 | 	 *   started at address shmaddr. It records it's size and then unmaps | 
 | 808 | 	 *   it. | 
 | 809 | 	 * - Then it unmaps all shm vmas that started at shmaddr and that | 
 | 810 | 	 *   are within the initially determined size. | 
 | 811 | 	 * Errors from do_munmap are ignored: the function only fails if | 
 | 812 | 	 * it's called with invalid parameters or if it's called to unmap | 
 | 813 | 	 * a part of a vma. Both calls in this function are for full vmas, | 
 | 814 | 	 * the parameters are directly copied from the vma itself and always | 
 | 815 | 	 * valid - therefore do_munmap cannot fail. (famous last words?) | 
 | 816 | 	 */ | 
 | 817 | 	/* | 
 | 818 | 	 * If it had been mremap()'d, the starting address would not | 
 | 819 | 	 * match the usual checks anyway. So assume all vma's are | 
 | 820 | 	 * above the starting address given. | 
 | 821 | 	 */ | 
 | 822 | 	vma = find_vma(mm, addr); | 
 | 823 |  | 
 | 824 | 	while (vma) { | 
 | 825 | 		next = vma->vm_next; | 
 | 826 |  | 
 | 827 | 		/* | 
 | 828 | 		 * Check if the starting address would match, i.e. it's | 
 | 829 | 		 * a fragment created by mprotect() and/or munmap(), or it | 
 | 830 | 		 * otherwise it starts at this address with no hassles. | 
 | 831 | 		 */ | 
 | 832 | 		if ((vma->vm_ops == &shm_vm_ops || is_vm_hugetlb_page(vma)) && | 
 | 833 | 			(vma->vm_start - addr)/PAGE_SIZE == vma->vm_pgoff) { | 
 | 834 |  | 
 | 835 |  | 
 | 836 | 			size = vma->vm_file->f_dentry->d_inode->i_size; | 
 | 837 | 			do_munmap(mm, vma->vm_start, vma->vm_end - vma->vm_start); | 
 | 838 | 			/* | 
 | 839 | 			 * We discovered the size of the shm segment, so | 
 | 840 | 			 * break out of here and fall through to the next | 
 | 841 | 			 * loop that uses the size information to stop | 
 | 842 | 			 * searching for matching vma's. | 
 | 843 | 			 */ | 
 | 844 | 			retval = 0; | 
 | 845 | 			vma = next; | 
 | 846 | 			break; | 
 | 847 | 		} | 
 | 848 | 		vma = next; | 
 | 849 | 	} | 
 | 850 |  | 
 | 851 | 	/* | 
 | 852 | 	 * We need look no further than the maximum address a fragment | 
 | 853 | 	 * could possibly have landed at. Also cast things to loff_t to | 
 | 854 | 	 * prevent overflows and make comparisions vs. equal-width types. | 
 | 855 | 	 */ | 
 | 856 | 	while (vma && (loff_t)(vma->vm_end - addr) <= size) { | 
 | 857 | 		next = vma->vm_next; | 
 | 858 |  | 
 | 859 | 		/* finding a matching vma now does not alter retval */ | 
 | 860 | 		if ((vma->vm_ops == &shm_vm_ops || is_vm_hugetlb_page(vma)) && | 
 | 861 | 			(vma->vm_start - addr)/PAGE_SIZE == vma->vm_pgoff) | 
 | 862 |  | 
 | 863 | 			do_munmap(mm, vma->vm_start, vma->vm_end - vma->vm_start); | 
 | 864 | 		vma = next; | 
 | 865 | 	} | 
 | 866 |  | 
 | 867 | 	up_write(&mm->mmap_sem); | 
 | 868 | 	return retval; | 
 | 869 | } | 
 | 870 |  | 
 | 871 | #ifdef CONFIG_PROC_FS | 
 | 872 | static int sysvipc_shm_read_proc(char *buffer, char **start, off_t offset, int length, int *eof, void *data) | 
 | 873 | { | 
 | 874 | 	off_t pos = 0; | 
 | 875 | 	off_t begin = 0; | 
 | 876 | 	int i, len = 0; | 
 | 877 |  | 
 | 878 | 	down(&shm_ids.sem); | 
 | 879 | 	len += sprintf(buffer, "       key      shmid perms       size  cpid  lpid nattch   uid   gid  cuid  cgid      atime      dtime      ctime\n"); | 
 | 880 |  | 
 | 881 | 	for(i = 0; i <= shm_ids.max_id; i++) { | 
 | 882 | 		struct shmid_kernel* shp; | 
 | 883 |  | 
 | 884 | 		shp = shm_lock(i); | 
 | 885 | 		if(shp!=NULL) { | 
 | 886 | #define SMALL_STRING "%10d %10d  %4o %10u %5u %5u  %5d %5u %5u %5u %5u %10lu %10lu %10lu\n" | 
 | 887 | #define BIG_STRING   "%10d %10d  %4o %21u %5u %5u  %5d %5u %5u %5u %5u %10lu %10lu %10lu\n" | 
 | 888 | 			char *format; | 
 | 889 |  | 
 | 890 | 			if (sizeof(size_t) <= sizeof(int)) | 
 | 891 | 				format = SMALL_STRING; | 
 | 892 | 			else | 
 | 893 | 				format = BIG_STRING; | 
 | 894 | 			len += sprintf(buffer + len, format, | 
 | 895 | 				shp->shm_perm.key, | 
 | 896 | 				shm_buildid(i, shp->shm_perm.seq), | 
 | 897 | 				shp->shm_flags, | 
 | 898 | 				shp->shm_segsz, | 
 | 899 | 				shp->shm_cprid, | 
 | 900 | 				shp->shm_lprid, | 
 | 901 | 				is_file_hugepages(shp->shm_file) ? (file_count(shp->shm_file) - 1) : shp->shm_nattch, | 
 | 902 | 				shp->shm_perm.uid, | 
 | 903 | 				shp->shm_perm.gid, | 
 | 904 | 				shp->shm_perm.cuid, | 
 | 905 | 				shp->shm_perm.cgid, | 
 | 906 | 				shp->shm_atim, | 
 | 907 | 				shp->shm_dtim, | 
 | 908 | 				shp->shm_ctim); | 
 | 909 | 			shm_unlock(shp); | 
 | 910 |  | 
 | 911 | 			pos += len; | 
 | 912 | 			if(pos < offset) { | 
 | 913 | 				len = 0; | 
 | 914 | 				begin = pos; | 
 | 915 | 			} | 
 | 916 | 			if(pos > offset + length) | 
 | 917 | 				goto done; | 
 | 918 | 		} | 
 | 919 | 	} | 
 | 920 | 	*eof = 1; | 
 | 921 | done: | 
 | 922 | 	up(&shm_ids.sem); | 
 | 923 | 	*start = buffer + (offset - begin); | 
 | 924 | 	len -= (offset - begin); | 
 | 925 | 	if(len > length) | 
 | 926 | 		len = length; | 
 | 927 | 	if(len < 0) | 
 | 928 | 		len = 0; | 
 | 929 | 	return len; | 
 | 930 | } | 
 | 931 | #endif |