Online Linux Driver Verification Service (alpha)

1 #include <linux/spinlock.h> 2 void ldv_spin_lock_irqsave(spinlock_t *lock, unsigned long flags); 3 void ldv_spin_lock_nested(spinlock_t *lock, int subclass); 4 void ldv_spin_lock_nest_lock(spinlock_t *lock, void *map); 5 void ldv_spin_lock_irqsave_nested(spinlock_t *lock, int subclass); 6 int ldv_spin_trylock_irqsave(spinlock_t *lock, unsigned long flags); 7 void ldv_spin_lock(spinlock_t *lock); 8 void ldv_spin_lock_bh(spinlock_t *lock); 9 void ldv_spin_lock_irq(spinlock_t *lock); 10 int ldv_spin_trylock(spinlock_t *lock); 11 int ldv_spin_trylock_bh(spinlock_t *lock); 12 int ldv_spin_trylock_irq(spinlock_t *lock); 13 void ldv_spin_unlock(spinlock_t *lock); 14 void ldv_spin_unlock_bh(spinlock_t *lock); 15 void ldv_spin_unlock_irq(spinlock_t *lock); 16 void ldv_spin_unlock_irqrestore(spinlock_t *lock, unsigned long flags); 17 void ldv_spin_unlock_wait(spinlock_t *lock); 18 int ldv_spin_is_locked(spinlock_t *lock); 19 int ldv_spin_is_contended(spinlock_t *lock); 20 int ldv_spin_can_lock(spinlock_t *lock); 21 int ldv_atomic_dec_and_lock(spinlock_t *lock, atomic_t *atomic); 22 #define ldv_atomic_dec_and_lock_macro(atomic,lock) ldv_atomic_dec_and_lock(lock,atomic) 23 // Michal Marschall 24 // Numer indeksu: 291693 25 26 #include "circ_dma_buf.h" 27 #include <linux/kernel.h> 28 #include <linux/pci.h> 29 30 const int CDB_SIZE = 4096; 31 32 struct circ_dma_buf *cdb_alloc(struct device *device) { 33 struct circ_dma_buf *buffer = kmalloc(sizeof(struct circ_dma_buf), GFP_KERNEL); 34 if(buffer == NULL) 35 return NULL; 36 buffer->data = dma_alloc_coherent(device, CDB_SIZE, &buffer->handle, GFP_KERNEL); 37 if(buffer->data == NULL) { 38 kfree(buffer); // avoid memory leak 39 return NULL; 40 } 41 buffer->start = 0; 42 buffer->fill = 0; 43 buffer->device = device; 44 spin_lock_init(&buffer->lock); 45 return buffer; 46 } 47 48 void cdb_free(struct circ_dma_buf *buffer) { 49 dma_free_coherent(buffer->device, CDB_SIZE, buffer->data, buffer->handle); 50 kfree(buffer); 51 } 52 53 int cdb_inc_begin(struct circ_dma_buf *buffer, int value) { 54 if(value > buffer->fill) 55 value = buffer->fill; 56 buffer->fill -= value; 57 buffer->start = (buffer->start + value) % CDB_SIZE; 58 return value; 59 } 60 61 int cdb_inc_end(struct circ_dma_buf *buffer, int value) { 62 if(value > CDB_SIZE - buffer->fill) 63 value = CDB_SIZE - buffer->fill; 64 buffer->fill += value; 65 return value; 66 } 67 68 int cdb_copy_from(struct circ_dma_buf *source, void *dest, int count, int (*copy_func)(void *, void *, int)) { 69 int first, copied1, copied2; 70 unsigned long irq_flags; 71 72 ldv_spin_lock_irqsave(&source->lock, irq_flags); 73 if(count > source->fill) 74 count = source->fill; 75 76 // copy first part: 77 if(count > CDB_SIZE - source->start) 78 first = CDB_SIZE - source->start; 79 else 80 first = count; 81 ldv_spin_unlock_irqrestore(&source->lock, irq_flags); 82 copied1 = copy_func(source->data + source->start, dest, first); 83 ldv_spin_lock_irqsave(&source->lock, irq_flags); 84 cdb_inc_begin(source, copied1); 85 ldv_spin_unlock_irqrestore(&source->lock, irq_flags); 86 if(first == count || copied1 < first) // copied everything or error 87 return copied1; 88 89 // copy second part: 90 copied2 = copy_func(source->data, dest + first, count - first); 91 cdb_inc_begin(source, copied2); 92 93 return copied1 + copied2; 94 } 95 96 int cdb_copy_to(void *source, struct circ_dma_buf *dest, int count, int (*copy_func)(void *, void *, int)) { 97 int end, first, copied1, copied2; 98 unsigned long irq_flags; 99 100 ldv_spin_lock_irqsave(&dest->lock, irq_flags); 101 if(count > CDB_SIZE - dest->fill) 102 count = CDB_SIZE - dest->fill; 103 104 // copy first part: 105 end = (dest->start + dest->fill) % CDB_SIZE; 106 if(end + count > CDB_SIZE) 107 first = CDB_SIZE - end; 108 else 109 first = count; 110 ldv_spin_unlock_irqrestore(&dest->lock, irq_flags); 111 copied1 = copy_func(source, dest->data + end, first); 112 ldv_spin_lock_irqsave(&dest->lock, irq_flags); 113 cdb_inc_end(dest, copied1); 114 ldv_spin_unlock_irqrestore(&dest->lock, irq_flags); 115 if(first == count || copied1 < first) // copied everything or error 116 return copied1; 117 118 // copy second part: 119 copied2 = copy_func(source + first, dest->data, count - first); 120 cdb_inc_end(dest, copied2); 121 122 return copied1 + copied2; 123 } 124 /* LDV_COMMENT_BEGIN_MODEL */ 125 #include <linux/kernel.h> 126 #include <linux/spinlock.h> 127 128 /* 129 CONFIG_DEBUG_SPINLOCK should be true 130 make menuconfig 131 Kernel hacking->Kernel debugging->Spinlock and rw-lock debugging: basic checks 132 */ 133 134 /* the function works only without aspectator */ 135 long __builtin_expect(long val, long res) { 136 return val; 137 } 138 139 #include "engine-blast.h" 140 141 142 /* Need this because rerouter is buggy!.. */ 143 extern int ldv_lock_TEMPLATE; 144 /* Now the actual variable goes... */ 145 int ldv_lock_TEMPLATE = 1; 146 147 #define __ldv_spin_lock() \ 148 do {\ 149 /* LDV_COMMENT_ASSERT Lock should be in a free state*/\ 150 ldv_assert(ldv_lock_TEMPLATE==1);\ 151 /* LDV_COMMENT_CHANGE_STATE Goto locked state*/\ 152 ldv_lock_TEMPLATE=2;\ 153 } while(0) 154 155 #define __ldv_spin_unlock() \ 156 do {\ 157 /* LDV_COMMENT_ASSERT Lock should be in a locked state*/\ 158 ldv_assert(ldv_lock_TEMPLATE!=1);\ 159 /* LDV_COMMENT_CHANGE_STATE Goto free state*/\ 160 ldv_lock_TEMPLATE=1;\ 161 } while(0) 162 163 #define __ldv_spin_trylock() \ 164 do {\ 165 int is_lock_held_by_another_thread;\ 166 /* LDV_COMMENT_OTHER Construct an arbitrary flag*/\ 167 is_lock_held_by_another_thread = ldv_undef_int();\ 168 /* LDV_COMMENT_OTHER If lock is free choose arbitrary action*/\ 169 if(ldv_lock_TEMPLATE==1 && is_lock_held_by_another_thread)\ 170 {\ 171 /* LDV_COMMENT_CHANGE_STATE Goto locked state*/\ 172 ldv_lock_TEMPLATE=2;\ 173 /* LDV_COMMENT_RETURN The lock is acquired*/\ 174 return 1;\ 175 }\ 176 else\ 177 {\ 178 /* LDV_COMMENT_RETURN The lock is not acquired*/\ 179 return 0;\ 180 }\ 181 } while(0) 182 183 #define __ldv_spin_checklocked(free,busy) \ 184 do {\ 185 int is_lock_held_by_another_thread;\ 186 /* LDV_COMMENT_OTHER Construct an arbitrary flag*/\ 187 is_lock_held_by_another_thread = ldv_undef_int();\ 188 /* LDV_COMMENT_OTHER If lock is free choose arbitrary action*/\ 189 if(ldv_lock_TEMPLATE==1 && is_lock_held_by_another_thread)\ 190 {\ 191 /* LDV_COMMENT_RETURN The lock is free*/\ 192 return free;\ 193 }\ 194 else\ 195 {\ 196 /* LDV_COMMENT_RETURN The lock is not free*/\ 197 return busy;\ 198 }\ 199 } while(0) 200 201 /* LDV_COMMENT_MODEL_FUNCTION_DEFINITION(name='ldv_spin_lock_irqsave(?!_nested)') Acquires the lock and checks for double spin lock*/ 202 void ldv_spin_lock_irqsave_TEMPLATE(spinlock_t *lock, unsigned long flags) { 203 __ldv_spin_lock(); 204 } 205 206 /* LDV_COMMENT_MODEL_FUNCTION_DEFINITION(name='ldv_spin_lock_nested') Acquires the lock and checks for double spin lock*/ 207 void ldv_spin_lock_nested_TEMPLATE(spinlock_t *lock, int subclass) { 208 __ldv_spin_lock(); 209 } 210 211 /* LDV_COMMENT_MODEL_FUNCTION_DEFINITION(name='ldv_spin_lock_nest_lock') Acquires the lock and checks for double spin lock*/ 212 void ldv_spin_lock_nest_lock_TEMPLATE(spinlock_t *lock, void *map) { 213 __ldv_spin_lock(); 214 } 215 216 /* LDV_COMMENT_MODEL_FUNCTION_DEFINITION(name='ldv_spin_lock_irqsave_nested') Acquires the lock and checks for double spin lock*/ 217 void ldv_spin_lock_irqsave_nested_TEMPLATE(spinlock_t *lock, unsigned long flags, int subclass) { 218 __ldv_spin_lock(); 219 } 220 221 /* LDV_COMMENT_MODEL_FUNCTION_DEFINITION(name='ldv_spin_trylock_irqsave(?!_nested)') Tryies to acquire the lock and returns one if successful*/ 222 int ldv_spin_trylock_irqsave_TEMPLATE(spinlock_t *lock, unsigned long flags) { 223 __ldv_spin_trylock(); 224 } 225 226 /* LDV_COMMENT_MODEL_FUNCTION_DEFINITION(name='ldv_spin_lock(?!_bh|_irq|_irqsave|_nested|_irqsave_nested|_nest_lock)') Acquires the lock and checks for double spin lock*/ 227 void ldv_spin_lock_TEMPLATE(spinlock_t *lock) { 228 __ldv_spin_lock(); 229 } 230 231 /* LDV_COMMENT_MODEL_FUNCTION_DEFINITION(name='ldv_spin_lock_bh') Acquires the lock and checks for double spin lock*/ 232 void ldv_spin_lock_bh_TEMPLATE(spinlock_t *lock) { 233 __ldv_spin_lock(); 234 } 235 236 /* LDV_COMMENT_MODEL_FUNCTION_DEFINITION(name='ldv_spin_lock_irq(?!save|save_nested)') Acquires the lock and checks for double spin lock*/ 237 void ldv_spin_lock_irq_TEMPLATE(spinlock_t *lock) { 238 __ldv_spin_lock(); 239 } 240 241 /* LDV_COMMENT_MODEL_FUNCTION_DEFINITION(name='ldv_spin_trylock(?!_bh|_irq|_irqsave|_irqsave_nested)') Tryies to acquire the lock and returns one if successful*/ 242 int ldv_spin_trylock_TEMPLATE(spinlock_t *lock) { 243 __ldv_spin_trylock(); 244 } 245 246 /* LDV_COMMENT_MODEL_FUNCTION_DEFINITION(name='ldv_spin_trylock_bh') Tryies to acquire the lock and returns one if successful*/ 247 int ldv_spin_trylock_bh_TEMPLATE(spinlock_t *lock) { 248 __ldv_spin_trylock(); 249 } 250 251 /* LDV_COMMENT_MODEL_FUNCTION_DEFINITION(name='ldv_spin_trylock_irq(?!save|save_nested)') Tryies to acquire the lock and returns one if successful*/ 252 int ldv_spin_trylock_irq_TEMPLATE(spinlock_t *lock) { 253 __ldv_spin_trylock(); 254 } 255 256 /* LDV_COMMENT_MODEL_FUNCTION_DEFINITION(name='ldv_spin_unlock(?!_bh|_irq|_irqrestore)') Releases the lock and checks that lock was acquired before*/ 257 void ldv_spin_unlock_TEMPLATE(spinlock_t *lock) { 258 __ldv_spin_unlock(); 259 } 260 261 /* LDV_COMMENT_MODEL_FUNCTION_DEFINITION(name='ldv_spin_unlock_bh') Releases the lock and checks that lock was acquired before*/ 262 void ldv_spin_unlock_bh_TEMPLATE(spinlock_t *lock) { 263 __ldv_spin_unlock(); 264 } 265 266 /* LDV_COMMENT_MODEL_FUNCTION_DEFINITION(name='ldv_spin_unlock_irq(?!restore)') Releases the lock and checks that lock was acquired before*/ 267 void ldv_spin_unlock_irq_TEMPLATE(spinlock_t *lock) { 268 __ldv_spin_unlock(); 269 } 270 271 /* LDV_COMMENT_MODEL_FUNCTION_DEFINITION(name='ldv_spin_unlock_irqrestore') Releases the lock and checks that lock was acquired before*/ 272 void ldv_spin_unlock_irqrestore_TEMPLATE(spinlock_t *lock, unsigned long flags) { 273 __ldv_spin_unlock(); 274 } 275 276 /* LDV_COMMENT_MODEL_FUNCTION_DEFINITION(name='ldv_spin_unlock_wait') If the some process is holding the lock then it waits until it will be released*/ 277 void ldv_spin_unlock_wait_TEMPLATE(spinlock_t *lock) { 278 /* LDV_COMMENT_ASSERT The spinlock must not be locked by the current process*/ 279 ldv_assert(ldv_lock_TEMPLATE==1); 280 } 281 282 /* LDV_COMMENT_MODEL_FUNCTION_DEFINITION(name='ldv_spin_is_locked') Checks whether the lock is free or not*/ 283 int ldv_spin_is_locked_TEMPLATE(spinlock_t *lock) { 284 __ldv_spin_checklocked(0,1); 285 } 286 287 /* LDV_COMMENT_MODEL_FUNCTION_DEFINITION(name='ldv_spin_is_contended')*/ 288 int ldv_spin_is_contended_TEMPLATE(spinlock_t *lock) { 289 int is_lock_contended; 290 /* LDV_COMMENT_OTHER Construct an arbitrary flag*/ 291 is_lock_contended = ldv_undef_int(); 292 /* LDV_COMMENT_OTHER Choose arbitrary action*/ 293 if(is_lock_contended) 294 { 295 /* LDV_COMMENT_RETURN The lock is contended*/ 296 return 1; 297 } 298 else 299 { 300 /* LDV_COMMENT_RETURN The lock is not contended*/ 301 return 0; 302 } 303 } 304 305 /* LDV_COMMENT_MODEL_FUNCTION_DEFINITION(name='ldv_spin_can_lock') Checks whether the lock is free or not*/ 306 int ldv_spin_can_lock_TEMPLATE(spinlock_t *lock) { 307 __ldv_spin_checklocked(1,0); 308 } 309 310 /* LDV_COMMENT_MODEL_FUNCTION_DEFINITION(name='ldv_atomic_dec_and_lock') Locks on reaching reference count zero*/ 311 int ldv_atomic_dec_and_lock_TEMPLATE(spinlock_t *lock, atomic_t *atomic) { 312 int is_atomic_counter_is_one; 313 /* LDV_COMMENT_OTHER Construct an arbitrary flag*/ 314 is_atomic_counter_is_one = ldv_undef_int(); 315 /* LDV_COMMENT_OTHER Choose arbitrary action*/ 316 if(is_atomic_counter_is_one) { 317 /* LDV_COMMENT_RETURN Set counter to zero*/ 318 atomic_dec(atomic); 319 /* LDV_COMMENT_RETURN Acquire the lock and return true*/ 320 __ldv_spin_lock(); 321 return 1; 322 } else { 323 /* LDV_COMMENT_RETURN Return false*/ 324 return 0; 325 } 326 } 327 328 /* LDV_COMMENT_MODEL_FUNCTION_DEFINITION(name='ldv_check_final_state') Check that the spinlock is unlocked at the end*/ 329 void ldv_check_final_state_TEMPLATE(void) 330 { 331 /* LDV_COMMENT_ASSERT The spinlock must be unlocked at the end*/ 332 ldv_assert(ldv_lock_TEMPLATE == 1); 333 } 334 335 /* LDV_COMMENT_MODEL_FUNCTION_DEFINITION(name='ldv_initialize') Initialize spinlock variables*/ 336 void ldv_initialize_TEMPLATE(void) 337 { 338 /* LDV_COMMENT_ASSERT Initialize spinlock with initial model value*/ 339 ldv_lock_TEMPLATE = 1; 340 }

1 #include <linux/spinlock.h> 2 void ldv_spin_lock_irqsave(spinlock_t *lock, unsigned long flags); 3 void ldv_spin_lock_nested(spinlock_t *lock, int subclass); 4 void ldv_spin_lock_nest_lock(spinlock_t *lock, void *map); 5 void ldv_spin_lock_irqsave_nested(spinlock_t *lock, int subclass); 6 int ldv_spin_trylock_irqsave(spinlock_t *lock, unsigned long flags); 7 void ldv_spin_lock(spinlock_t *lock); 8 void ldv_spin_lock_bh(spinlock_t *lock); 9 void ldv_spin_lock_irq(spinlock_t *lock); 10 int ldv_spin_trylock(spinlock_t *lock); 11 int ldv_spin_trylock_bh(spinlock_t *lock); 12 int ldv_spin_trylock_irq(spinlock_t *lock); 13 void ldv_spin_unlock(spinlock_t *lock); 14 void ldv_spin_unlock_bh(spinlock_t *lock); 15 void ldv_spin_unlock_irq(spinlock_t *lock); 16 void ldv_spin_unlock_irqrestore(spinlock_t *lock, unsigned long flags); 17 void ldv_spin_unlock_wait(spinlock_t *lock); 18 int ldv_spin_is_locked(spinlock_t *lock); 19 int ldv_spin_is_contended(spinlock_t *lock); 20 int ldv_spin_can_lock(spinlock_t *lock); 21 int ldv_atomic_dec_and_lock(spinlock_t *lock, atomic_t *atomic); 22 #define ldv_atomic_dec_and_lock_macro(atomic,lock) ldv_atomic_dec_and_lock(lock,atomic) 23 24 // Michal Marschall 25 // Numer indeksu: 291693 26 27 #include <linux/module.h> 28 #include <linux/kernel.h> 29 #include <linux/pci.h> 30 #include <linux/interrupt.h> 31 #include <linux/fs.h> 32 #include <linux/errno.h> 33 #include <linux/spinlock.h> 34 #include <linux/wait.h> 35 #include <linux/sched.h> 36 #include "circ_dma_buf.h" 37 38 MODULE_AUTHOR("Michal Marschall"); 39 MODULE_LICENSE("GPL"); 40 41 #define MAX_DEVICES 64 42 43 // types of character devices: 44 enum chardev_type {TYPE_SRC1, TYPE_SRC2, TYPE_DST}; 45 const int TYPES = 3; // total number of types 46 47 #define XORDEV_MINOR(number, type) ((number) * TYPES + (type)) 48 #define GET_NUMBER(minor) ((minor) / TYPES) 49 #define GET_TYPE(minor) ((minor) % TYPES) 50 51 // names used in various places: 52 #define NAME_CLASS "xordev_class" 53 #define NAME_PCI "xordev_pci" 54 #define NAME_CHAR "xordev_char" 55 56 // ids: 57 #define ID_VENDOR 0x1af4 58 #define ID_DEVICE 0x10fd 59 60 // xordev PCI registers: 61 static const int REG_INTR_EN = 0x00; 62 static const int REG_SRC1 = 0x04; 63 static const int REG_SRC2 = 0x08; 64 static const int REG_DST = 0x0c; 65 static const int REG_COUNT = 0x10; 66 static const int REG_INTR_COUNT = 0x14; 67 68 // values of REG_INTR_EN: 69 static const int IRQ_OFF = 0; 70 static const int IRQ_ON = 1; 71 72 struct xordev_device { 73 int present; 74 int number; // from 0 to MAX_DEVICES - 1 75 struct pci_dev *pcidev; 76 void __iomem *iomem; 77 int sent; // bytes sent to xor 78 wait_queue_head_t wait_queue; 79 spinlock_t lock; 80 struct char_device *src1; 81 struct char_device *src2; 82 struct char_device *dst; 83 }; 84 85 struct char_device { 86 dev_t node; 87 struct xordev_device *xordev; 88 struct circ_dma_buf *buffer; 89 struct mutex mutex; 90 // for src* devices only: 91 int end; 92 }; 93 94 static int xordev_major; 95 static struct class *xordev_class; 96 struct xordev_device devices[MAX_DEVICES]; 97 98 static int cdb_copy_to_user(void *source, void *dest, int count) { // for argument to cdb_copy_* 99 return count - copy_to_user(dest, source, count); 100 } 101 102 static int cdb_copy_from_user(void *source, void *dest, int count) { // for argument to cdb_copy_* 103 return count - copy_from_user(dest, source, count); 104 } 105 106 static int update_buffer(struct char_device *chardev) { 107 struct circ_dma_buf *buffer = chardev->buffer; 108 unsigned long irq_flags; 109 int position, result; 110 111 ldv_spin_lock_irqsave(&buffer->lock, irq_flags); 112 position = chardev->xordev->sent - ioread32(chardev->xordev->iomem + REG_COUNT); 113 switch(GET_TYPE(chardev->node)) { 114 case TYPE_SRC1: 115 case TYPE_SRC2: 116 result = position - buffer->start; 117 if(result < 0) 118 result += CDB_SIZE; 119 if(cdb_inc_begin(buffer, result) != result) // this should never happen 120 result = -EIO; 121 break; 122 case TYPE_DST: 123 result = position - (buffer->start + buffer->fill); 124 if(result < 0) 125 result += CDB_SIZE; 126 if(cdb_inc_end(buffer, result) != result) // this should never happen 127 result = -EIO; 128 break; 129 default: 130 result = -EINVAL; 131 } 132 ldv_spin_unlock_irqrestore(&buffer->lock, irq_flags); 133 134 return result; 135 } 136 137 static void start_xoring(struct xordev_device *xordev) { 138 unsigned long irq_flags1, irq_flags2; 139 int ready, space, send_now; 140 141 ldv_spin_lock_irqsave(&xordev->lock, irq_flags1); 142 ldv_spin_lock_irqsave(&xordev->dst->buffer->lock, irq_flags2); 143 ready = min_t(int, xordev->src1->end, xordev->src2->end) - xordev->sent; 144 space = xordev->dst->buffer->start - xordev->sent; 145 if(space < 0 || (space == 0 && xordev->dst->buffer->fill == 0)) 146 space = CDB_SIZE - xordev->sent; 147 send_now = min_t(int, ready, space); 148 if(send_now > 0) { 149 xordev->sent += send_now; 150 iowrite32(send_now, xordev->iomem + REG_COUNT); 151 iowrite32(IRQ_ON, xordev->iomem + REG_INTR_EN); 152 wake_up_interruptible(&xordev->wait_queue); 153 } 154 ldv_spin_unlock_irqrestore(&xordev->dst->buffer->lock, irq_flags2); 155 ldv_spin_unlock_irqrestore(&xordev->lock, irq_flags1); 156 } 157 158 static void irq_one_byte(struct xordev_device *xordev) { 159 int count = ioread32(xordev->iomem + REG_COUNT); 160 if(count > 0) 161 iowrite32(count - 1, xordev->iomem + REG_INTR_COUNT); 162 } 163 164 static ssize_t xordev_read(struct file *file, char *buffer, size_t count, loff_t *file_pos) { 165 struct char_device *chardev = (struct char_device*)file->private_data; 166 struct xordev_device *xordev = chardev->xordev; 167 ssize_t result; 168 int ret_value; 169 170 ret_value = update_buffer(chardev); 171 if(ret_value < 0) 172 return ret_value; 173 mutex_lock(&chardev->mutex); 174 while(chardev->buffer->fill == 0) { 175 irq_one_byte(xordev); 176 mutex_unlock(&chardev->mutex); 177 if(wait_event_interruptible(xordev->wait_queue, chardev->buffer->fill > 0) < 0) 178 return -ERESTARTSYS; 179 mutex_lock(&chardev->mutex); 180 } 181 result = cdb_copy_from(chardev->buffer, (void *)buffer, count, cdb_copy_to_user); 182 if(result == 0) { // this should never happen 183 mutex_unlock(&chardev->mutex); 184 return -EIO; 185 } 186 *file_pos = file->f_pos + result; 187 mutex_unlock(&chardev->mutex); 188 start_xoring(xordev); 189 190 return result; 191 } 192 193 static ssize_t xordev_write(struct file *file, const char *buffer, size_t count, loff_t *file_pos) { 194 struct char_device *chardev = (struct char_device*)file->private_data; 195 struct xordev_device *xordev = chardev->xordev; 196 ssize_t result; 197 int ret_value; 198 unsigned long irq_flags; 199 200 ret_value = update_buffer(chardev); 201 if(ret_value < 0) 202 return ret_value; 203 mutex_lock(&chardev->mutex); 204 while(chardev->buffer->fill == CDB_SIZE) { 205 irq_one_byte(xordev); 206 mutex_unlock(&chardev->mutex); 207 if(wait_event_interruptible(xordev->wait_queue, chardev->buffer->fill < CDB_SIZE) < 0) 208 return -ERESTARTSYS; 209 mutex_lock(&chardev->mutex); 210 } 211 result = cdb_copy_to((void *)buffer, chardev->buffer, count, cdb_copy_from_user); 212 if(result == 0) { // this should never happen 213 mutex_unlock(&chardev->mutex); 214 return -ENOSPC; 215 } 216 *file_pos = file->f_pos + result; 217 mutex_unlock(&chardev->mutex); 218 ldv_spin_lock_irqsave(&xordev->lock, irq_flags); 219 chardev->end += result; 220 ldv_spin_unlock_irqrestore(&xordev->lock, irq_flags); 221 start_xoring(xordev); 222 223 return result; 224 } 225 226 static struct file_operations xordev_dst_fops; 227 static struct file_operations xordev_input_fops; // definitions below not visible 228 229 static int xordev_open(struct inode *inode, struct file *file) { 230 int minor = MINOR(inode->i_rdev); 231 struct xordev_device *xordev = &devices[GET_NUMBER(minor)]; 232 struct char_device *chardev; 233 struct file_operations *file_ops; 234 235 switch(GET_TYPE(minor)) { 236 case TYPE_SRC1: 237 file_ops = &xordev_input_fops; 238 chardev = xordev->src1; 239 break; 240 case TYPE_SRC2: 241 file_ops = &xordev_input_fops; 242 chardev = xordev->src2; 243 break; 244 case TYPE_DST: 245 file_ops = &xordev_dst_fops; 246 chardev = xordev->dst; 247 break; 248 default: // this should never happen 249 return -EINVAL; 250 } 251 252 file->f_op = file_ops; 253 file->private_data = (void *)chardev; 254 255 return 0; 256 } 257 258 static int xordev_release(struct inode *inode, struct file *file) { 259 return 0; 260 } 261 262 static struct file_operations xordev_fops = { 263 owner: THIS_MODULE, 264 open: xordev_open, 265 release: xordev_release 266 }; 267 268 static struct file_operations xordev_dst_fops = { 269 owner: THIS_MODULE, 270 read: xordev_read, 271 open: xordev_open, 272 release: xordev_release 273 }; 274 275 static struct file_operations xordev_input_fops = { 276 owner: THIS_MODULE, 277 write: xordev_write, 278 open: xordev_open, 279 release: xordev_release 280 }; 281 282 static struct char_device *chardev_alloc(struct xordev_device *xordev, int type) { 283 struct char_device *chardev = kmalloc(sizeof(struct char_device), GFP_KERNEL); 284 if(chardev == NULL) 285 return NULL; 286 chardev->node = MKDEV(xordev_major, XORDEV_MINOR(xordev->number, type)); 287 chardev->xordev = xordev; 288 chardev->buffer = cdb_alloc(&xordev->pcidev->dev); 289 if(chardev->buffer == NULL) { 290 kfree(chardev); // avoid memory leak 291 return NULL; 292 } 293 mutex_init(&chardev->mutex); 294 chardev->end = 0; 295 return chardev; 296 } 297 298 static void chardev_free(struct char_device *chardev) { 299 cdb_free(chardev->buffer); 300 kfree(chardev); 301 } 302 303 static void init_reg_values(struct xordev_device *xordev) { 304 iowrite32(0, xordev->iomem + REG_INTR_COUNT); 305 iowrite32(xordev->src1->buffer->handle, xordev->iomem + REG_SRC1); 306 iowrite32(xordev->src2->buffer->handle, xordev->iomem + REG_SRC2); 307 iowrite32(xordev->dst->buffer->handle, xordev->iomem + REG_DST); 308 } 309 310 static irqreturn_t xordev_irq_handler(int irq, void *device) { 311 struct xordev_device *xordev = (struct xordev_device *)pci_get_drvdata((struct pci_dev *)device); 312 int count, intr_count; 313 unsigned long irq_flags; 314 315 ldv_spin_lock_irqsave(&xordev->lock, irq_flags); 316 if(ioread32(xordev->iomem + REG_INTR_EN) != IRQ_ON) { 317 ldv_spin_unlock_irqrestore(&xordev->lock, irq_flags); 318 return IRQ_NONE; 319 } 320 count = ioread32(xordev->iomem + REG_COUNT); 321 intr_count = ioread32(xordev->iomem + REG_INTR_COUNT); 322 if(count > intr_count) { 323 ldv_spin_unlock_irqrestore(&xordev->lock, irq_flags); 324 return IRQ_NONE; 325 } 326 327 // handle interrupt: 328 update_buffer(xordev->src1); 329 update_buffer(xordev->src2); 330 update_buffer(xordev->dst); 331 if(count == 0) { 332 iowrite32(IRQ_OFF, xordev->iomem + REG_INTR_EN); 333 if(xordev->sent == CDB_SIZE) { 334 xordev->sent -= CDB_SIZE; 335 xordev->src1->end -= CDB_SIZE; 336 xordev->src2->end -= CDB_SIZE; 337 init_reg_values(xordev); 338 } 339 } 340 if(intr_count != 0) 341 iowrite32(0, xordev->iomem + REG_INTR_COUNT); 342 wake_up_interruptible(&xordev->wait_queue); 343 ldv_spin_unlock_irqrestore(&xordev->lock, irq_flags); 344 start_xoring(xordev); 345 346 return IRQ_HANDLED; 347 } 348 349 static int xordev_probe(struct pci_dev *pcidev, const struct pci_device_id *id) { 350 int i, ret_value; 351 struct xordev_device *xordev; 352 struct device *node; 353 354 // look for first not used device: 355 for(i = 0; i < MAX_DEVICES; ++i) 356 if(!devices[i].present) 357 break; 358 if(i == MAX_DEVICES) 359 return -ENOMEM; 360 xordev = &devices[i]; 361 362 // enable device, map region: 363 xordev->pcidev = pcidev; 364 ret_value = pci_enable_device(pcidev); 365 if(ret_value < 0) 366 return ret_value; 367 ret_value = pci_request_regions(pcidev, NAME_PCI); 368 if(ret_value < 0) 369 return ret_value; 370 xordev->iomem = pci_iomap(pcidev, 0, 0); 371 if(xordev->iomem == NULL) 372 return -EIO; 373 374 // enable DMA: 375 pci_set_master(pcidev); 376 ret_value = pci_set_dma_mask(pcidev, DMA_BIT_MASK(32)); 377 if(ret_value < 0) 378 return ret_value; 379 ret_value = pci_set_consistent_dma_mask(pcidev, DMA_BIT_MASK(32)); 380 if(ret_value < 0) 381 return ret_value; 382 383 // allocate other resources: 384 xordev->src1 = chardev_alloc(xordev, TYPE_SRC1); 385 if(xordev->src1 == NULL) 386 return -ENOMEM; 387 xordev->src2 = chardev_alloc(xordev, TYPE_SRC2); 388 if(xordev->src2 == NULL) 389 return -ENOMEM; 390 xordev->dst = chardev_alloc(xordev, TYPE_DST); 391 if(xordev->dst == NULL) 392 return -ENOMEM; 393 init_waitqueue_head(&xordev->wait_queue); 394 395 // set initial values: 396 spin_lock_init(&xordev->lock); 397 xordev->sent = 0; 398 pci_set_drvdata(pcidev, (void *)xordev); 399 init_reg_values(xordev); 400 iowrite32(IRQ_OFF, xordev->iomem + REG_INTR_EN); 401 402 // request interrupts: 403 ret_value = request_irq(pcidev->irq, xordev_irq_handler, IRQF_SHARED, NAME_PCI, (void *)pcidev); 404 if(ret_value < 0) 405 return ret_value; 406 407 // create nodes in /dev: 408 node = device_create(xordev_class, NULL, xordev->src1->node, NULL, "xor%ds1", xordev->number); 409 if(IS_ERR(node)) 410 return -EIO; 411 node = device_create(xordev_class, NULL, xordev->src2->node, NULL, "xor%ds2", xordev->number); 412 if(IS_ERR(node)) 413 return -EIO; 414 node = device_create(xordev_class, NULL, xordev->dst->node, NULL, "xor%dd", xordev->number); 415 if(IS_ERR(node)) 416 return -EIO; 417 418 xordev->present = 1; 419 420 return 0; 421 } 422 423 static void xordev_remove(struct pci_dev *pcidev) { 424 struct xordev_device *xordev = (struct xordev_device *)pci_get_drvdata(pcidev); 425 426 xordev->present = false; 427 428 // remove nodes from /dev: 429 device_destroy(xordev_class, xordev->src1->node); 430 device_destroy(xordev_class, xordev->src2->node); 431 device_destroy(xordev_class, xordev->dst->node); 432 433 // release resources, disable device: 434 free_irq(pcidev->irq, (void *)pcidev); 435 pci_iounmap(pcidev, xordev->iomem); 436 pci_release_regions(pcidev); 437 pci_disable_device(pcidev); 438 chardev_free(xordev->src1); 439 chardev_free(xordev->src2); 440 chardev_free(xordev->dst); 441 } 442 443 static struct pci_device_id xordev_id = { 444 PCI_DEVICE(ID_VENDOR, ID_DEVICE) 445 }; 446 447 static struct pci_driver xordev_driver = { 448 name: NAME_PCI, 449 id_table: &xordev_id, 450 probe: xordev_probe, 451 remove: xordev_remove, 452 }; 453 454 static int xordev_init_module(void) { 455 int i, ret_value; 456 457 // initialize array devices: 458 for(i = 0; i < MAX_DEVICES; ++i) { 459 devices[i].present = 0; 460 devices[i].number = i; 461 } 462 463 // register devices: 464 xordev_major = register_chrdev(0, NAME_CHAR, &xordev_fops); 465 if(xordev_major < 0) { 466 ret_value = xordev_major; 467 goto err_register_chrdev; 468 } 469 xordev_class = class_create(THIS_MODULE, NAME_CLASS); 470 if(IS_ERR(xordev_class)) { 471 ret_value = -EIO; 472 goto err_class_create; 473 } 474 ret_value = pci_register_driver(&xordev_driver); 475 if(ret_value < 0) 476 goto err_register_driver; 477 478 return 0; 479 480 err_register_driver: 481 pci_unregister_driver(&xordev_driver); 482 err_class_create: 483 class_destroy(xordev_class); 484 err_register_chrdev: 485 unregister_chrdev(xordev_major, NAME_CHAR); 486 return ret_value; 487 } 488 489 static void xordev_cleanup_module(void) { 490 pci_unregister_driver(&xordev_driver); 491 class_destroy(xordev_class); 492 unregister_chrdev(xordev_major, NAME_CHAR); 493 } 494 495 module_init(xordev_init_module); 496 module_exit(xordev_cleanup_module); 497 498 499 500 501 502 /* LDV_COMMENT_BEGIN_MAIN */ 503 #ifdef LDV_MAIN1_sequence_infinite_withcheck_stateful 504 505 /*###########################################################################*/ 506 507 /*############## Driver Environment Generator 0.2 output ####################*/ 508 509 /*###########################################################################*/ 510 511 512 513 /* LDV_COMMENT_FUNCTION_DECLARE_LDV Special function for LDV verifier. Test if all kernel resources are correctly released by driver before driver will be unloaded. */ 514 void ldv_check_final_state(void); 515 516 /* LDV_COMMENT_FUNCTION_DECLARE_LDV Special function for LDV verifier. Test correct return result. */ 517 void ldv_check_return_value(int res); 518 519 /* LDV_COMMENT_FUNCTION_DECLARE_LDV Special function for LDV verifier. Initializes the model. */ 520 void ldv_initialize(void); 521 522 /* LDV_COMMENT_FUNCTION_DECLARE_LDV Special function for LDV verifier. Returns arbitrary interger value. */ 523 int nondet_int(void); 524 525 /* LDV_COMMENT_VAR_DECLARE_LDV Special variable for LDV verifier. */ 526 int LDV_IN_INTERRUPT; 527 528 /* LDV_COMMENT_FUNCTION_MAIN Main function for LDV verifier. */ 529 void ldv_main1_sequence_infinite_withcheck_stateful(void) { 530 531 532 533 /* LDV_COMMENT_BEGIN_VARIABLE_DECLARATION_PART */ 534 /*============================= VARIABLE DECLARATION PART =============================*/ 535 /** CALLBACK SECTION request_irq **/ 536 /* content: static irqreturn_t xordev_irq_handler(int irq, void *device)*/ 537 /* LDV_COMMENT_BEGIN_PREP */ 538 #define MAX_DEVICES 64 539 #define XORDEV_MINOR(number, type) ((number) * TYPES + (type)) 540 #define GET_NUMBER(minor) ((minor) / TYPES) 541 #define GET_TYPE(minor) ((minor) % TYPES) 542 #define NAME_CLASS "xordev_class" 543 #define NAME_PCI "xordev_pci" 544 #define NAME_CHAR "xordev_char" 545 #define ID_VENDOR 0x1af4 546 #define ID_DEVICE 0x10fd 547 /* LDV_COMMENT_END_PREP */ 548 /* LDV_COMMENT_VAR_DECLARE Variable declaration for function "xordev_irq_handler" */ 549 int var_xordev_irq_handler_12_p0; 550 /* LDV_COMMENT_VAR_DECLARE Variable declaration for function "xordev_irq_handler" */ 551 void * var_xordev_irq_handler_12_p1; 552 553 554 555 556 /* LDV_COMMENT_END_VARIABLE_DECLARATION_PART */ 557 /* LDV_COMMENT_BEGIN_VARIABLE_INITIALIZING_PART */ 558 /*============================= VARIABLE INITIALIZING PART =============================*/ 559 LDV_IN_INTERRUPT=1; 560 561 562 563 564 /* LDV_COMMENT_END_VARIABLE_INITIALIZING_PART */ 565 /* LDV_COMMENT_BEGIN_FUNCTION_CALL_SECTION */ 566 /*============================= FUNCTION CALL SECTION =============================*/ 567 /* LDV_COMMENT_FUNCTION_CALL Initialize LDV model. */ 568 ldv_initialize(); 569 570 /** INIT: init_type: ST_MODULE_INIT **/ 571 /* content: static int xordev_init_module(void)*/ 572 /* LDV_COMMENT_BEGIN_PREP */ 573 #define MAX_DEVICES 64 574 #define XORDEV_MINOR(number, type) ((number) * TYPES + (type)) 575 #define GET_NUMBER(minor) ((minor) / TYPES) 576 #define GET_TYPE(minor) ((minor) % TYPES) 577 #define NAME_CLASS "xordev_class" 578 #define NAME_PCI "xordev_pci" 579 #define NAME_CHAR "xordev_char" 580 #define ID_VENDOR 0x1af4 581 #define ID_DEVICE 0x10fd 582 /* LDV_COMMENT_END_PREP */ 583 /* LDV_COMMENT_FUNCTION_CALL Kernel calls driver init function after driver loading to kernel. This function declared as "MODULE_INIT(function name)". */ 584 if(xordev_init_module()) 585 goto ldv_final; 586 587 588 589 while( nondet_int() 590 ) { 591 592 switch(nondet_int()) { 593 594 case 0: { 595 596 /** CALLBACK SECTION request_irq **/ 597 LDV_IN_INTERRUPT=2; 598 599 /* content: static irqreturn_t xordev_irq_handler(int irq, void *device)*/ 600 /* LDV_COMMENT_BEGIN_PREP */ 601 #define MAX_DEVICES 64 602 #define XORDEV_MINOR(number, type) ((number) * TYPES + (type)) 603 #define GET_NUMBER(minor) ((minor) / TYPES) 604 #define GET_TYPE(minor) ((minor) % TYPES) 605 #define NAME_CLASS "xordev_class" 606 #define NAME_PCI "xordev_pci" 607 #define NAME_CHAR "xordev_char" 608 #define ID_VENDOR 0x1af4 609 #define ID_DEVICE 0x10fd 610 /* LDV_COMMENT_END_PREP */ 611 /* LDV_COMMENT_FUNCTION_CALL */ 612 xordev_irq_handler( var_xordev_irq_handler_12_p0, var_xordev_irq_handler_12_p1); 613 LDV_IN_INTERRUPT=1; 614 615 616 617 } 618 619 break; 620 default: break; 621 622 } 623 624 } 625 626 ldv_module_exit: 627 628 /** INIT: init_type: ST_MODULE_EXIT **/ 629 /* content: static void xordev_cleanup_module(void)*/ 630 /* LDV_COMMENT_BEGIN_PREP */ 631 #define MAX_DEVICES 64 632 #define XORDEV_MINOR(number, type) ((number) * TYPES + (type)) 633 #define GET_NUMBER(minor) ((minor) / TYPES) 634 #define GET_TYPE(minor) ((minor) % TYPES) 635 #define NAME_CLASS "xordev_class" 636 #define NAME_PCI "xordev_pci" 637 #define NAME_CHAR "xordev_char" 638 #define ID_VENDOR 0x1af4 639 #define ID_DEVICE 0x10fd 640 /* LDV_COMMENT_END_PREP */ 641 /* LDV_COMMENT_FUNCTION_CALL Kernel calls driver release function before driver will be uploaded from kernel. This function declared as "MODULE_EXIT(function name)". */ 642 xordev_cleanup_module(); 643 644 /* LDV_COMMENT_FUNCTION_CALL Checks that all resources and locks are correctly released before the driver will be unloaded. */ 645 ldv_final: ldv_check_final_state(); 646 647 /* LDV_COMMENT_END_FUNCTION_CALL_SECTION */ 648 return; 649 650 } 651 #endif 652 653 /* LDV_COMMENT_END_MAIN */

1 #ifndef _LINUX_ERR_H 2 #define _LINUX_ERR_H 3 4 #include <linux/compiler.h> 5 6 #include <asm/errno.h> 7 8 /* 9 * Kernel pointers have redundant information, so we can use a 10 * scheme where we can return either an error code or a dentry 11 * pointer with the same return value. 12 * 13 * This should be a per-architecture thing, to allow different 14 * error and pointer decisions. 15 */ 16 #define MAX_ERRNO 4095 17 18 #ifndef __ASSEMBLY__ 19 20 #define IS_ERR_VALUE(x) unlikely((x) >= (unsigned long)-MAX_ERRNO) 21 22 static inline void * __must_check ERR_PTR(long error) 23 { 24 return (void *) error; 25 } 26 27 static inline long __must_check PTR_ERR(const void *ptr) 28 { 29 return (long) ptr; 30 } 31 32 static inline long __must_check IS_ERR(const void *ptr) 33 { 34 return IS_ERR_VALUE((unsigned long)ptr); 35 } 36 37 static inline long __must_check IS_ERR_OR_NULL(const void *ptr) 38 { 39 return !ptr || IS_ERR_VALUE((unsigned long)ptr); 40 } 41 42 /** 43 * ERR_CAST - Explicitly cast an error-valued pointer to another pointer type 44 * @ptr: The pointer to cast. 45 * 46 * Explicitly cast an error-valued pointer to another pointer type in such a 47 * way as to make it clear that's what's going on. 48 */ 49 static inline void * __must_check ERR_CAST(const void *ptr) 50 { 51 /* cast away the const */ 52 return (void *) ptr; 53 } 54 55 static inline int __must_check PTR_RET(const void *ptr) 56 { 57 if (IS_ERR(ptr)) 58 return PTR_ERR(ptr); 59 else 60 return 0; 61 } 62 63 #endif 64 65 #endif /* _LINUX_ERR_H */

1 #ifndef _LINUX_FS_H 2 #define _LINUX_FS_H 3 4 /* 5 * This file has definitions for some important file table 6 * structures etc. 7 */ 8 9 #include <linux/limits.h> 10 #include <linux/ioctl.h> 11 #include <linux/blk_types.h> 12 #include <linux/types.h> 13 14 /* 15 * It's silly to have NR_OPEN bigger than NR_FILE, but you can change 16 * the file limit at runtime and only root can increase the per-process 17 * nr_file rlimit, so it's safe to set up a ridiculously high absolute 18 * upper limit on files-per-process. 19 * 20 * Some programs (notably those using select()) may have to be 21 * recompiled to take full advantage of the new limits.. 22 */ 23 24 /* Fixed constants first: */ 25 #undef NR_OPEN 26 #define INR_OPEN_CUR 1024 /* Initial setting for nfile rlimits */ 27 #define INR_OPEN_MAX 4096 /* Hard limit for nfile rlimits */ 28 29 #define BLOCK_SIZE_BITS 10 30 #define BLOCK_SIZE (1<<BLOCK_SIZE_BITS) 31 32 #define SEEK_SET 0 /* seek relative to beginning of file */ 33 #define SEEK_CUR 1 /* seek relative to current file position */ 34 #define SEEK_END 2 /* seek relative to end of file */ 35 #define SEEK_MAX SEEK_END 36 37 struct fstrim_range { 38 __u64 start; 39 __u64 len; 40 __u64 minlen; 41 }; 42 43 /* And dynamically-tunable limits and defaults: */ 44 struct files_stat_struct { 45 unsigned long nr_files; /* read only */ 46 unsigned long nr_free_files; /* read only */ 47 unsigned long max_files; /* tunable */ 48 }; 49 50 struct inodes_stat_t { 51 int nr_inodes; 52 int nr_unused; 53 int dummy[5]; /* padding for sysctl ABI compatibility */ 54 }; 55 56 57 #define NR_FILE 8192 /* this can well be larger on a larger system */ 58 59 #define MAY_EXEC 1 60 #define MAY_WRITE 2 61 #define MAY_READ 4 62 #define MAY_APPEND 8 63 #define MAY_ACCESS 16 64 #define MAY_OPEN 32 65 #define MAY_CHDIR 64 66 67 /* 68 * flags in file.f_mode. Note that FMODE_READ and FMODE_WRITE must correspond 69 * to O_WRONLY and O_RDWR via the strange trick in __dentry_open() 70 */ 71 72 /* file is open for reading */ 73 #define FMODE_READ ((__force fmode_t)0x1) 74 /* file is open for writing */ 75 #define FMODE_WRITE ((__force fmode_t)0x2) 76 /* file is seekable */ 77 #define FMODE_LSEEK ((__force fmode_t)0x4) 78 /* file can be accessed using pread */ 79 #define FMODE_PREAD ((__force fmode_t)0x8) 80 /* file can be accessed using pwrite */ 81 #define FMODE_PWRITE ((__force fmode_t)0x10) 82 /* File is opened for execution with sys_execve / sys_uselib */ 83 #define FMODE_EXEC ((__force fmode_t)0x20) 84 /* File is opened with O_NDELAY (only set for block devices) */ 85 #define FMODE_NDELAY ((__force fmode_t)0x40) 86 /* File is opened with O_EXCL (only set for block devices) */ 87 #define FMODE_EXCL ((__force fmode_t)0x80) 88 /* File is opened using open(.., 3, ..) and is writeable only for ioctls 89 (specialy hack for floppy.c) */ 90 #define FMODE_WRITE_IOCTL ((__force fmode_t)0x100) 91 92 /* 93 * Don't update ctime and mtime. 94 * 95 * Currently a special hack for the XFS open_by_handle ioctl, but we'll 96 * hopefully graduate it to a proper O_CMTIME flag supported by open(2) soon. 97 */ 98 #define FMODE_NOCMTIME ((__force fmode_t)0x800) 99 100 /* Expect random access pattern */ 101 #define FMODE_RANDOM ((__force fmode_t)0x1000) 102 103 /* File is huge (eg. /dev/kmem): treat loff_t as unsigned */ 104 #define FMODE_UNSIGNED_OFFSET ((__force fmode_t)0x2000) 105 106 /* File is opened with O_PATH; almost nothing can be done with it */ 107 #define FMODE_PATH ((__force fmode_t)0x4000) 108 109 /* File was opened by fanotify and shouldn't generate fanotify events */ 110 #define FMODE_NONOTIFY ((__force fmode_t)0x1000000) 111 112 /* 113 * The below are the various read and write types that we support. Some of 114 * them include behavioral modifiers that send information down to the 115 * block layer and IO scheduler. Terminology: 116 * 117 * The block layer uses device plugging to defer IO a little bit, in 118 * the hope that we will see more IO very shortly. This increases 119 * coalescing of adjacent IO and thus reduces the number of IOs we 120 * have to send to the device. It also allows for better queuing, 121 * if the IO isn't mergeable. If the caller is going to be waiting 122 * for the IO, then he must ensure that the device is unplugged so 123 * that the IO is dispatched to the driver. 124 * 125 * All IO is handled async in Linux. This is fine for background 126 * writes, but for reads or writes that someone waits for completion 127 * on, we want to notify the block layer and IO scheduler so that they 128 * know about it. That allows them to make better scheduling 129 * decisions. So when the below references 'sync' and 'async', it 130 * is referencing this priority hint. 131 * 132 * With that in mind, the available types are: 133 * 134 * READ A normal read operation. Device will be plugged. 135 * READ_SYNC A synchronous read. Device is not plugged, caller can 136 * immediately wait on this read without caring about 137 * unplugging. 138 * READA Used for read-ahead operations. Lower priority, and the 139 * block layer could (in theory) choose to ignore this 140 * request if it runs into resource problems. 141 * WRITE A normal async write. Device will be plugged. 142 * WRITE_SYNC Synchronous write. Identical to WRITE, but passes down 143 * the hint that someone will be waiting on this IO 144 * shortly. The write equivalent of READ_SYNC. 145 * WRITE_ODIRECT Special case write for O_DIRECT only. 146 * WRITE_FLUSH Like WRITE_SYNC but with preceding cache flush. 147 * WRITE_FUA Like WRITE_SYNC but data is guaranteed to be on 148 * non-volatile media on completion. 149 * WRITE_FLUSH_FUA Combination of WRITE_FLUSH and FUA. The IO is preceded 150 * by a cache flush and data is guaranteed to be on 151 * non-volatile media on completion. 152 * 153 */ 154 #define RW_MASK REQ_WRITE 155 #define RWA_MASK REQ_RAHEAD 156 157 #define READ 0 158 #define WRITE RW_MASK 159 #define READA RWA_MASK 160 161 #define READ_SYNC (READ | REQ_SYNC) 162 #define READ_META (READ | REQ_META) 163 #define WRITE_SYNC (WRITE | REQ_SYNC | REQ_NOIDLE) 164 #define WRITE_ODIRECT (WRITE | REQ_SYNC) 165 #define WRITE_META (WRITE | REQ_META) 166 #define WRITE_FLUSH (WRITE | REQ_SYNC | REQ_NOIDLE | REQ_FLUSH) 167 #define WRITE_FUA (WRITE | REQ_SYNC | REQ_NOIDLE | REQ_FUA) 168 #define WRITE_FLUSH_FUA (WRITE | REQ_SYNC | REQ_NOIDLE | REQ_FLUSH | REQ_FUA) 169 170 #define SEL_IN 1 171 #define SEL_OUT 2 172 #define SEL_EX 4 173 174 /* public flags for file_system_type */ 175 #define FS_REQUIRES_DEV 1 176 #define FS_BINARY_MOUNTDATA 2 177 #define FS_HAS_SUBTYPE 4 178 #define FS_REVAL_DOT 16384 /* Check the paths ".", ".." for staleness */ 179 #define FS_RENAME_DOES_D_MOVE 32768 /* FS will handle d_move() 180 * during rename() internally. 181 */ 182 183 /* 184 * These are the fs-independent mount-flags: up to 32 flags are supported 185 */ 186 #define MS_RDONLY 1 /* Mount read-only */ 187 #define MS_NOSUID 2 /* Ignore suid and sgid bits */ 188 #define MS_NODEV 4 /* Disallow access to device special files */ 189 #define MS_NOEXEC 8 /* Disallow program execution */ 190 #define MS_SYNCHRONOUS 16 /* Writes are synced at once */ 191 #define MS_REMOUNT 32 /* Alter flags of a mounted FS */ 192 #define MS_MANDLOCK 64 /* Allow mandatory locks on an FS */ 193 #define MS_DIRSYNC 128 /* Directory modifications are synchronous */ 194 #define MS_NOATIME 1024 /* Do not update access times. */ 195 #define MS_NODIRATIME 2048 /* Do not update directory access times */ 196 #define MS_BIND 4096 197 #define MS_MOVE 8192 198 #define MS_REC 16384 199 #define MS_VERBOSE 32768 /* War is peace. Verbosity is silence. 200 MS_VERBOSE is deprecated. */ 201 #define MS_SILENT 32768 202 #define MS_POSIXACL (1<<16) /* VFS does not apply the umask */ 203 #define MS_UNBINDABLE (1<<17) /* change to unbindable */ 204 #define MS_PRIVATE (1<<18) /* change to private */ 205 #define MS_SLAVE (1<<19) /* change to slave */ 206 #define MS_SHARED (1<<20) /* change to shared */ 207 #define MS_RELATIME (1<<21) /* Update atime relative to mtime/ctime. */ 208 #define MS_KERNMOUNT (1<<22) /* this is a kern_mount call */ 209 #define MS_I_VERSION (1<<23) /* Update inode I_version field */ 210 #define MS_STRICTATIME (1<<24) /* Always perform atime updates */ 211 #define MS_NOSEC (1<<28) 212 #define MS_BORN (1<<29) 213 #define MS_ACTIVE (1<<30) 214 #define MS_NOUSER (1<<31) 215 216 /* 217 * Superblock flags that can be altered by MS_REMOUNT 218 */ 219 #define MS_RMT_MASK (MS_RDONLY|MS_SYNCHRONOUS|MS_MANDLOCK|MS_I_VERSION) 220 221 /* 222 * Old magic mount flag and mask 223 */ 224 #define MS_MGC_VAL 0xC0ED0000 225 #define MS_MGC_MSK 0xffff0000 226 227 /* Inode flags - they have nothing to superblock flags now */ 228 229 #define S_SYNC 1 /* Writes are synced at once */ 230 #define S_NOATIME 2 /* Do not update access times */ 231 #define S_APPEND 4 /* Append-only file */ 232 #define S_IMMUTABLE 8 /* Immutable file */ 233 #define S_DEAD 16 /* removed, but still open directory */ 234 #define S_NOQUOTA 32 /* Inode is not counted to quota */ 235 #define S_DIRSYNC 64 /* Directory modifications are synchronous */ 236 #define S_NOCMTIME 128 /* Do not update file c/mtime */ 237 #define S_SWAPFILE 256 /* Do not truncate: swapon got its bmaps */ 238 #define S_PRIVATE 512 /* Inode is fs-internal */ 239 #define S_IMA 1024 /* Inode has an associated IMA struct */ 240 #define S_AUTOMOUNT 2048 /* Automount/referral quasi-directory */ 241 #define S_NOSEC 4096 /* no suid or xattr security attributes */ 242 243 /* 244 * Note that nosuid etc flags are inode-specific: setting some file-system 245 * flags just means all the inodes inherit those flags by default. It might be 246 * possible to override it selectively if you really wanted to with some 247 * ioctl() that is not currently implemented. 248 * 249 * Exception: MS_RDONLY is always applied to the entire file system. 250 * 251 * Unfortunately, it is possible to change a filesystems flags with it mounted 252 * with files in use. This means that all of the inodes will not have their 253 * i_flags updated. Hence, i_flags no longer inherit the superblock mount 254 * flags, so these have to be checked separately. -- rmk@arm.uk.linux.org 255 */ 256 #define __IS_FLG(inode,flg) ((inode)->i_sb->s_flags & (flg)) 257 258 #define IS_RDONLY(inode) ((inode)->i_sb->s_flags & MS_RDONLY) 259 #define IS_SYNC(inode) (__IS_FLG(inode, MS_SYNCHRONOUS) || \ 260 ((inode)->i_flags & S_SYNC)) 261 #define IS_DIRSYNC(inode) (__IS_FLG(inode, MS_SYNCHRONOUS|MS_DIRSYNC) || \ 262 ((inode)->i_flags & (S_SYNC|S_DIRSYNC))) 263 #define IS_MANDLOCK(inode) __IS_FLG(inode, MS_MANDLOCK) 264 #define IS_NOATIME(inode) __IS_FLG(inode, MS_RDONLY|MS_NOATIME) 265 #define IS_I_VERSION(inode) __IS_FLG(inode, MS_I_VERSION) 266 267 #define IS_NOQUOTA(inode) ((inode)->i_flags & S_NOQUOTA) 268 #define IS_APPEND(inode) ((inode)->i_flags & S_APPEND) 269 #define IS_IMMUTABLE(inode) ((inode)->i_flags & S_IMMUTABLE) 270 #define IS_POSIXACL(inode) __IS_FLG(inode, MS_POSIXACL) 271 272 #define IS_DEADDIR(inode) ((inode)->i_flags & S_DEAD) 273 #define IS_NOCMTIME(inode) ((inode)->i_flags & S_NOCMTIME) 274 #define IS_SWAPFILE(inode) ((inode)->i_flags & S_SWAPFILE) 275 #define IS_PRIVATE(inode) ((inode)->i_flags & S_PRIVATE) 276 #define IS_IMA(inode) ((inode)->i_flags & S_IMA) 277 #define IS_AUTOMOUNT(inode) ((inode)->i_flags & S_AUTOMOUNT) 278 #define IS_NOSEC(inode) ((inode)->i_flags & S_NOSEC) 279 280 /* the read-only stuff doesn't really belong here, but any other place is 281 probably as bad and I don't want to create yet another include file. */ 282 283 #define BLKROSET _IO(0x12,93) /* set device read-only (0 = read-write) */ 284 #define BLKROGET _IO(0x12,94) /* get read-only status (0 = read_write) */ 285 #define BLKRRPART _IO(0x12,95) /* re-read partition table */ 286 #define BLKGETSIZE _IO(0x12,96) /* return device size /512 (long *arg) */ 287 #define BLKFLSBUF _IO(0x12,97) /* flush buffer cache */ 288 #define BLKRASET _IO(0x12,98) /* set read ahead for block device */ 289 #define BLKRAGET _IO(0x12,99) /* get current read ahead setting */ 290 #define BLKFRASET _IO(0x12,100)/* set filesystem (mm/filemap.c) read-ahead */ 291 #define BLKFRAGET _IO(0x12,101)/* get filesystem (mm/filemap.c) read-ahead */ 292 #define BLKSECTSET _IO(0x12,102)/* set max sectors per request (ll_rw_blk.c) */ 293 #define BLKSECTGET _IO(0x12,103)/* get max sectors per request (ll_rw_blk.c) */ 294 #define BLKSSZGET _IO(0x12,104)/* get block device sector size */ 295 #if 0 296 #define BLKPG _IO(0x12,105)/* See blkpg.h */ 297 298 /* Some people are morons. Do not use sizeof! */ 299 300 #define BLKELVGET _IOR(0x12,106,size_t)/* elevator get */ 301 #define BLKELVSET _IOW(0x12,107,size_t)/* elevator set */ 302 /* This was here just to show that the number is taken - 303 probably all these _IO(0x12,*) ioctls should be moved to blkpg.h. */ 304 #endif 305 /* A jump here: 108-111 have been used for various private purposes. */ 306 #define BLKBSZGET _IOR(0x12,112,size_t) 307 #define BLKBSZSET _IOW(0x12,113,size_t) 308 #define BLKGETSIZE64 _IOR(0x12,114,size_t) /* return device size in bytes (u64 *arg) */ 309 #define BLKTRACESETUP _IOWR(0x12,115,struct blk_user_trace_setup) 310 #define BLKTRACESTART _IO(0x12,116) 311 #define BLKTRACESTOP _IO(0x12,117) 312 #define BLKTRACETEARDOWN _IO(0x12,118) 313 #define BLKDISCARD _IO(0x12,119) 314 #define BLKIOMIN _IO(0x12,120) 315 #define BLKIOOPT _IO(0x12,121) 316 #define BLKALIGNOFF _IO(0x12,122) 317 #define BLKPBSZGET _IO(0x12,123) 318 #define BLKDISCARDZEROES _IO(0x12,124) 319 #define BLKSECDISCARD _IO(0x12,125) 320 321 #define BMAP_IOCTL 1 /* obsolete - kept for compatibility */ 322 #define FIBMAP _IO(0x00,1) /* bmap access */ 323 #define FIGETBSZ _IO(0x00,2) /* get the block size used for bmap */ 324 #define FIFREEZE _IOWR('X', 119, int) /* Freeze */ 325 #define FITHAW _IOWR('X', 120, int) /* Thaw */ 326 #define FITRIM _IOWR('X', 121, struct fstrim_range) /* Trim */ 327 328 #define FS_IOC_GETFLAGS _IOR('f', 1, long) 329 #define FS_IOC_SETFLAGS _IOW('f', 2, long) 330 #define FS_IOC_GETVERSION _IOR('v', 1, long) 331 #define FS_IOC_SETVERSION _IOW('v', 2, long) 332 #define FS_IOC_FIEMAP _IOWR('f', 11, struct fiemap) 333 #define FS_IOC32_GETFLAGS _IOR('f', 1, int) 334 #define FS_IOC32_SETFLAGS _IOW('f', 2, int) 335 #define FS_IOC32_GETVERSION _IOR('v', 1, int) 336 #define FS_IOC32_SETVERSION _IOW('v', 2, int) 337 338 /* 339 * Inode flags (FS_IOC_GETFLAGS / FS_IOC_SETFLAGS) 340 */ 341 #define FS_SECRM_FL 0x00000001 /* Secure deletion */ 342 #define FS_UNRM_FL 0x00000002 /* Undelete */ 343 #define FS_COMPR_FL 0x00000004 /* Compress file */ 344 #define FS_SYNC_FL 0x00000008 /* Synchronous updates */ 345 #define FS_IMMUTABLE_FL 0x00000010 /* Immutable file */ 346 #define FS_APPEND_FL 0x00000020 /* writes to file may only append */ 347 #define FS_NODUMP_FL 0x00000040 /* do not dump file */ 348 #define FS_NOATIME_FL 0x00000080 /* do not update atime */ 349 /* Reserved for compression usage... */ 350 #define FS_DIRTY_FL 0x00000100 351 #define FS_COMPRBLK_FL 0x00000200 /* One or more compressed clusters */ 352 #define FS_NOCOMP_FL 0x00000400 /* Don't compress */ 353 #define FS_ECOMPR_FL 0x00000800 /* Compression error */ 354 /* End compression flags --- maybe not all used */ 355 #define FS_BTREE_FL 0x00001000 /* btree format dir */ 356 #define FS_INDEX_FL 0x00001000 /* hash-indexed directory */ 357 #define FS_IMAGIC_FL 0x00002000 /* AFS directory */ 358 #define FS_JOURNAL_DATA_FL 0x00004000 /* Reserved for ext3 */ 359 #define FS_NOTAIL_FL 0x00008000 /* file tail should not be merged */ 360 #define FS_DIRSYNC_FL 0x00010000 /* dirsync behaviour (directories only) */ 361 #define FS_TOPDIR_FL 0x00020000 /* Top of directory hierarchies*/ 362 #define FS_EXTENT_FL 0x00080000 /* Extents */ 363 #define FS_DIRECTIO_FL 0x00100000 /* Use direct i/o */ 364 #define FS_NOCOW_FL 0x00800000 /* Do not cow file */ 365 #define FS_RESERVED_FL 0x80000000 /* reserved for ext2 lib */ 366 367 #define FS_FL_USER_VISIBLE 0x0003DFFF /* User visible flags */ 368 #define FS_FL_USER_MODIFIABLE 0x000380FF /* User modifiable flags */ 369 370 371 #define SYNC_FILE_RANGE_WAIT_BEFORE 1 372 #define SYNC_FILE_RANGE_WRITE 2 373 #define SYNC_FILE_RANGE_WAIT_AFTER 4 374 375 #ifdef __KERNEL__ 376 377 #include <linux/linkage.h> 378 #include <linux/wait.h> 379 #include <linux/types.h> 380 #include <linux/kdev_t.h> 381 #include <linux/dcache.h> 382 #include <linux/path.h> 383 #include <linux/stat.h> 384 #include <linux/cache.h> 385 #include <linux/list.h> 386 #include <linux/radix-tree.h> 387 #include <linux/prio_tree.h> 388 #include <linux/init.h> 389 #include <linux/pid.h> 390 #include <linux/mutex.h> 391 #include <linux/capability.h> 392 #include <linux/semaphore.h> 393 #include <linux/fiemap.h> 394 #include <linux/rculist_bl.h> 395 396 #include <asm/atomic.h> 397 #include <asm/byteorder.h> 398 399 struct export_operations; 400 struct hd_geometry; 401 struct iovec; 402 struct nameidata; 403 struct kiocb; 404 struct kobject; 405 struct pipe_inode_info; 406 struct poll_table_struct; 407 struct kstatfs; 408 struct vm_area_struct; 409 struct vfsmount; 410 struct cred; 411 412 extern void __init inode_init(void); 413 extern void __init inode_init_early(void); 414 extern void __init files_init(unsigned long); 415 416 extern struct files_stat_struct files_stat; 417 extern unsigned long get_max_files(void); 418 extern int sysctl_nr_open; 419 extern struct inodes_stat_t inodes_stat; 420 extern int leases_enable, lease_break_time; 421 422 struct buffer_head; 423 typedef int (get_block_t)(struct inode *inode, sector_t iblock, 424 struct buffer_head *bh_result, int create); 425 typedef void (dio_iodone_t)(struct kiocb *iocb, loff_t offset, 426 ssize_t bytes, void *private, int ret, 427 bool is_async); 428 429 /* 430 * Attribute flags. These should be or-ed together to figure out what 431 * has been changed! 432 */ 433 #define ATTR_MODE (1 << 0) 434 #define ATTR_UID (1 << 1) 435 #define ATTR_GID (1 << 2) 436 #define ATTR_SIZE (1 << 3) 437 #define ATTR_ATIME (1 << 4) 438 #define ATTR_MTIME (1 << 5) 439 #define ATTR_CTIME (1 << 6) 440 #define ATTR_ATIME_SET (1 << 7) 441 #define ATTR_MTIME_SET (1 << 8) 442 #define ATTR_FORCE (1 << 9) /* Not a change, but a change it */ 443 #define ATTR_ATTR_FLAG (1 << 10) 444 #define ATTR_KILL_SUID (1 << 11) 445 #define ATTR_KILL_SGID (1 << 12) 446 #define ATTR_FILE (1 << 13) 447 #define ATTR_KILL_PRIV (1 << 14) 448 #define ATTR_OPEN (1 << 15) /* Truncating from open(O_TRUNC) */ 449 #define ATTR_TIMES_SET (1 << 16) 450 451 /* 452 * This is the Inode Attributes structure, used for notify_change(). It 453 * uses the above definitions as flags, to know which values have changed. 454 * Also, in this manner, a Filesystem can look at only the values it cares 455 * about. Basically, these are the attributes that the VFS layer can 456 * request to change from the FS layer. 457 * 458 * Derek Atkins <warlord@MIT.EDU> 94-10-20 459 */ 460 struct iattr { 461 unsigned int ia_valid; 462 umode_t ia_mode; 463 uid_t ia_uid; 464 gid_t ia_gid; 465 loff_t ia_size; 466 struct timespec ia_atime; 467 struct timespec ia_mtime; 468 struct timespec ia_ctime; 469 470 /* 471 * Not an attribute, but an auxiliary info for filesystems wanting to 472 * implement an ftruncate() like method. NOTE: filesystem should 473 * check for (ia_valid & ATTR_FILE), and not for (ia_file != NULL). 474 */ 475 struct file *ia_file; 476 }; 477 478 /* 479 * Includes for diskquotas. 480 */ 481 #include <linux/quota.h> 482 483 /** 484 * enum positive_aop_returns - aop return codes with specific semantics 485 * 486 * @AOP_WRITEPAGE_ACTIVATE: Informs the caller that page writeback has 487 * completed, that the page is still locked, and 488 * should be considered active. The VM uses this hint 489 * to return the page to the active list -- it won't 490 * be a candidate for writeback again in the near 491 * future. Other callers must be careful to unlock 492 * the page if they get this return. Returned by 493 * writepage(); 494 * 495 * @AOP_TRUNCATED_PAGE: The AOP method that was handed a locked page has 496 * unlocked it and the page might have been truncated. 497 * The caller should back up to acquiring a new page and 498 * trying again. The aop will be taking reasonable 499 * precautions not to livelock. If the caller held a page 500 * reference, it should drop it before retrying. Returned 501 * by readpage(). 502 * 503 * address_space_operation functions return these large constants to indicate 504 * special semantics to the caller. These are much larger than the bytes in a 505 * page to allow for functions that return the number of bytes operated on in a 506 * given page. 507 */ 508 509 enum positive_aop_returns { 510 AOP_WRITEPAGE_ACTIVATE = 0x80000, 511 AOP_TRUNCATED_PAGE = 0x80001, 512 }; 513 514 #define AOP_FLAG_UNINTERRUPTIBLE 0x0001 /* will not do a short write */ 515 #define AOP_FLAG_CONT_EXPAND 0x0002 /* called from cont_expand */ 516 #define AOP_FLAG_NOFS 0x0004 /* used by filesystem to direct 517 * helper code (eg buffer layer) 518 * to clear GFP_FS from alloc */ 519 520 /* 521 * oh the beauties of C type declarations. 522 */ 523 struct page; 524 struct address_space; 525 struct writeback_control; 526 527 struct iov_iter { 528 const struct iovec *iov; 529 unsigned long nr_segs; 530 size_t iov_offset; 531 size_t count; 532 }; 533 534 size_t iov_iter_copy_from_user_atomic(struct page *page, 535 struct iov_iter *i, unsigned long offset, size_t bytes); 536 size_t iov_iter_copy_from_user(struct page *page, 537 struct iov_iter *i, unsigned long offset, size_t bytes); 538 void iov_iter_advance(struct iov_iter *i, size_t bytes); 539 int iov_iter_fault_in_readable(struct iov_iter *i, size_t bytes); 540 size_t iov_iter_single_seg_count(struct iov_iter *i); 541 542 static inline void iov_iter_init(struct iov_iter *i, 543 const struct iovec *iov, unsigned long nr_segs, 544 size_t count, size_t written) 545 { 546 i->iov = iov; 547 i->nr_segs = nr_segs; 548 i->iov_offset = 0; 549 i->count = count + written; 550 551 iov_iter_advance(i, written); 552 } 553 554 static inline size_t iov_iter_count(struct iov_iter *i) 555 { 556 return i->count; 557 } 558 559 /* 560 * "descriptor" for what we're up to with a read. 561 * This allows us to use the same read code yet 562 * have multiple different users of the data that 563 * we read from a file. 564 * 565 * The simplest case just copies the data to user 566 * mode. 567 */ 568 typedef struct { 569 size_t written; 570 size_t count; 571 union { 572 char __user *buf; 573 void *data; 574 } arg; 575 int error; 576 } read_descriptor_t; 577 578 typedef int (*read_actor_t)(read_descriptor_t *, struct page *, 579 unsigned long, unsigned long); 580 581 struct address_space_operations { 582 int (*writepage)(struct page *page, struct writeback_control *wbc); 583 int (*readpage)(struct file *, struct page *); 584 585 /* Write back some dirty pages from this mapping. */ 586 int (*writepages)(struct address_space *, struct writeback_control *); 587 588 /* Set a page dirty. Return true if this dirtied it */ 589 int (*set_page_dirty)(struct page *page); 590 591 int (*readpages)(struct file *filp, struct address_space *mapping, 592 struct list_head *pages, unsigned nr_pages); 593 594 int (*write_begin)(struct file *, struct address_space *mapping, 595 loff_t pos, unsigned len, unsigned flags, 596 struct page **pagep, void **fsdata); 597 int (*write_end)(struct file *, struct address_space *mapping, 598 loff_t pos, unsigned len, unsigned copied, 599 struct page *page, void *fsdata); 600 601 /* Unfortunately this kludge is needed for FIBMAP. Don't use it */ 602 sector_t (*bmap)(struct address_space *, sector_t); 603 void (*invalidatepage) (struct page *, unsigned long); 604 int (*releasepage) (struct page *, gfp_t); 605 void (*freepage)(struct page *); 606 ssize_t (*direct_IO)(int, struct kiocb *, const struct iovec *iov, 607 loff_t offset, unsigned long nr_segs); 608 int (*get_xip_mem)(struct address_space *, pgoff_t, int, 609 void **, unsigned long *); 610 /* migrate the contents of a page to the specified target */ 611 int (*migratepage) (struct address_space *, 612 struct page *, struct page *); 613 int (*launder_page) (struct page *); 614 int (*is_partially_uptodate) (struct page *, read_descriptor_t *, 615 unsigned long); 616 int (*error_remove_page)(struct address_space *, struct page *); 617 }; 618 619 extern const struct address_space_operations empty_aops; 620 621 /* 622 * pagecache_write_begin/pagecache_write_end must be used by general code 623 * to write into the pagecache. 624 */ 625 int pagecache_write_begin(struct file *, struct address_space *mapping, 626 loff_t pos, unsigned len, unsigned flags, 627 struct page **pagep, void **fsdata); 628 629 int pagecache_write_end(struct file *, struct address_space *mapping, 630 loff_t pos, unsigned len, unsigned copied, 631 struct page *page, void *fsdata); 632 633 struct backing_dev_info; 634 struct address_space { 635 struct inode *host; /* owner: inode, block_device */ 636 struct radix_tree_root page_tree; /* radix tree of all pages */ 637 spinlock_t tree_lock; /* and lock protecting it */ 638 unsigned int i_mmap_writable;/* count VM_SHARED mappings */ 639 struct prio_tree_root i_mmap; /* tree of private and shared mappings */ 640 struct list_head i_mmap_nonlinear;/*list VM_NONLINEAR mappings */ 641 struct mutex i_mmap_mutex; /* protect tree, count, list */ 642 /* Protected by tree_lock together with the radix tree */ 643 unsigned long nrpages; /* number of total pages */ 644 pgoff_t writeback_index;/* writeback starts here */ 645 const struct address_space_operations *a_ops; /* methods */ 646 unsigned long flags; /* error bits/gfp mask */ 647 struct backing_dev_info *backing_dev_info; /* device readahead, etc */ 648 spinlock_t private_lock; /* for use by the address_space */ 649 struct list_head private_list; /* ditto */ 650 struct address_space *assoc_mapping; /* ditto */ 651 } __attribute__((aligned(sizeof(long)))); 652 /* 653 * On most architectures that alignment is already the case; but 654 * must be enforced here for CRIS, to let the least significant bit 655 * of struct page's "mapping" pointer be used for PAGE_MAPPING_ANON. 656 */ 657 658 struct block_device { 659 dev_t bd_dev; /* not a kdev_t - it's a search key */ 660 int bd_openers; 661 struct inode * bd_inode; /* will die */ 662 struct super_block * bd_super; 663 struct mutex bd_mutex; /* open/close mutex */ 664 struct list_head bd_inodes; 665 void * bd_claiming; 666 void * bd_holder; 667 int bd_holders; 668 bool bd_write_holder; 669 #ifdef CONFIG_SYSFS 670 struct list_head bd_holder_disks; 671 #endif 672 struct block_device * bd_contains; 673 unsigned bd_block_size; 674 struct hd_struct * bd_part; 675 /* number of times partitions within this device have been opened. */ 676 unsigned bd_part_count; 677 int bd_invalidated; 678 struct gendisk * bd_disk; 679 struct list_head bd_list; 680 /* 681 * Private data. You must have bd_claim'ed the block_device 682 * to use this. NOTE: bd_claim allows an owner to claim 683 * the same device multiple times, the owner must take special 684 * care to not mess up bd_private for that case. 685 */ 686 unsigned long bd_private; 687 688 /* The counter of freeze processes */ 689 int bd_fsfreeze_count; 690 /* Mutex for freeze */ 691 struct mutex bd_fsfreeze_mutex; 692 }; 693 694 /* 695 * Radix-tree tags, for tagging dirty and writeback pages within the pagecache 696 * radix trees 697 */ 698 #define PAGECACHE_TAG_DIRTY 0 699 #define PAGECACHE_TAG_WRITEBACK 1 700 #define PAGECACHE_TAG_TOWRITE 2 701 702 int mapping_tagged(struct address_space *mapping, int tag); 703 704 /* 705 * Might pages of this file be mapped into userspace? 706 */ 707 static inline int mapping_mapped(struct address_space *mapping) 708 { 709 return !prio_tree_empty(&mapping->i_mmap) || 710 !list_empty(&mapping->i_mmap_nonlinear); 711 } 712 713 /* 714 * Might pages of this file have been modified in userspace? 715 * Note that i_mmap_writable counts all VM_SHARED vmas: do_mmap_pgoff 716 * marks vma as VM_SHARED if it is shared, and the file was opened for 717 * writing i.e. vma may be mprotected writable even if now readonly. 718 */ 719 static inline int mapping_writably_mapped(struct address_space *mapping) 720 { 721 return mapping->i_mmap_writable != 0; 722 } 723 724 /* 725 * Use sequence counter to get consistent i_size on 32-bit processors. 726 */ 727 #if BITS_PER_LONG==32 && defined(CONFIG_SMP) 728 #include <linux/seqlock.h> 729 #define __NEED_I_SIZE_ORDERED 730 #define i_size_ordered_init(inode) seqcount_init(&inode->i_size_seqcount) 731 #else 732 #define i_size_ordered_init(inode) do { } while (0) 733 #endif 734 735 struct posix_acl; 736 #define ACL_NOT_CACHED ((void *)(-1)) 737 738 struct inode { 739 /* RCU path lookup touches following: */ 740 umode_t i_mode; 741 uid_t i_uid; 742 gid_t i_gid; 743 const struct inode_operations *i_op; 744 struct super_block *i_sb; 745 746 spinlock_t i_lock; /* i_blocks, i_bytes, maybe i_size */ 747 unsigned int i_flags; 748 unsigned long i_state; 749 #ifdef CONFIG_SECURITY 750 void *i_security; 751 #endif 752 struct mutex i_mutex; 753 754 755 unsigned long dirtied_when; /* jiffies of first dirtying */ 756 757 struct hlist_node i_hash; 758 struct list_head i_wb_list; /* backing dev IO list */ 759 struct list_head i_lru; /* inode LRU list */ 760 struct list_head i_sb_list; 761 union { 762 struct list_head i_dentry; 763 struct rcu_head i_rcu; 764 }; 765 unsigned long i_ino; 766 atomic_t i_count; 767 unsigned int i_nlink; 768 dev_t i_rdev; 769 unsigned int i_blkbits; 770 u64 i_version; 771 loff_t i_size; 772 #ifdef __NEED_I_SIZE_ORDERED 773 seqcount_t i_size_seqcount; 774 #endif 775 struct timespec i_atime; 776 struct timespec i_mtime; 777 struct timespec i_ctime; 778 blkcnt_t i_blocks; 779 unsigned short i_bytes; 780 struct rw_semaphore i_alloc_sem; 781 const struct file_operations *i_fop; /* former ->i_op->default_file_ops */ 782 struct file_lock *i_flock; 783 struct address_space *i_mapping; 784 struct address_space i_data; 785 #ifdef CONFIG_QUOTA 786 struct dquot *i_dquot[MAXQUOTAS]; 787 #endif 788 struct list_head i_devices; 789 union { 790 struct pipe_inode_info *i_pipe; 791 struct block_device *i_bdev; 792 struct cdev *i_cdev; 793 }; 794 795 __u32 i_generation; 796 797 #ifdef CONFIG_FSNOTIFY 798 __u32 i_fsnotify_mask; /* all events this inode cares about */ 799 struct hlist_head i_fsnotify_marks; 800 #endif 801 802 #ifdef CONFIG_IMA 803 atomic_t i_readcount; /* struct files open RO */ 804 #endif 805 atomic_t i_writecount; 806 #ifdef CONFIG_FS_POSIX_ACL 807 struct posix_acl *i_acl; 808 struct posix_acl *i_default_acl; 809 #endif 810 void *i_private; /* fs or device private pointer */ 811 }; 812 813 static inline int inode_unhashed(struct inode *inode) 814 { 815 return hlist_unhashed(&inode->i_hash); 816 } 817 818 /* 819 * inode->i_mutex nesting subclasses for the lock validator: 820 * 821 * 0: the object of the current VFS operation 822 * 1: parent 823 * 2: child/target 824 * 3: quota file 825 * 826 * The locking order between these classes is 827 * parent -> child -> normal -> xattr -> quota 828 */ 829 enum inode_i_mutex_lock_class 830 { 831 I_MUTEX_NORMAL, 832 I_MUTEX_PARENT, 833 I_MUTEX_CHILD, 834 I_MUTEX_XATTR, 835 I_MUTEX_QUOTA 836 }; 837 838 /* 839 * NOTE: in a 32bit arch with a preemptable kernel and 840 * an UP compile the i_size_read/write must be atomic 841 * with respect to the local cpu (unlike with preempt disabled), 842 * but they don't need to be atomic with respect to other cpus like in 843 * true SMP (so they need either to either locally disable irq around 844 * the read or for example on x86 they can be still implemented as a 845 * cmpxchg8b without the need of the lock prefix). For SMP compiles 846 * and 64bit archs it makes no difference if preempt is enabled or not. 847 */ 848 static inline loff_t i_size_read(const struct inode *inode) 849 { 850 #if BITS_PER_LONG==32 && defined(CONFIG_SMP) 851 loff_t i_size; 852 unsigned int seq; 853 854 do { 855 seq = read_seqcount_begin(&inode->i_size_seqcount); 856 i_size = inode->i_size; 857 } while (read_seqcount_retry(&inode->i_size_seqcount, seq)); 858 return i_size; 859 #elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPT) 860 loff_t i_size; 861 862 preempt_disable(); 863 i_size = inode->i_size; 864 preempt_enable(); 865 return i_size; 866 #else 867 return inode->i_size; 868 #endif 869 } 870 871 /* 872 * NOTE: unlike i_size_read(), i_size_write() does need locking around it 873 * (normally i_mutex), otherwise on 32bit/SMP an update of i_size_seqcount 874 * can be lost, resulting in subsequent i_size_read() calls spinning forever. 875 */ 876 static inline void i_size_write(struct inode *inode, loff_t i_size) 877 { 878 #if BITS_PER_LONG==32 && defined(CONFIG_SMP) 879 write_seqcount_begin(&inode->i_size_seqcount); 880 inode->i_size = i_size; 881 write_seqcount_end(&inode->i_size_seqcount); 882 #elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPT) 883 preempt_disable(); 884 inode->i_size = i_size; 885 preempt_enable(); 886 #else 887 inode->i_size = i_size; 888 #endif 889 } 890 891 static inline unsigned iminor(const struct inode *inode) 892 { 893 return MINOR(inode->i_rdev); 894 } 895 896 static inline unsigned imajor(const struct inode *inode) 897 { 898 return MAJOR(inode->i_rdev); 899 } 900 901 extern struct block_device *I_BDEV(struct inode *inode); 902 903 struct fown_struct { 904 rwlock_t lock; /* protects pid, uid, euid fields */ 905 struct pid *pid; /* pid or -pgrp where SIGIO should be sent */ 906 enum pid_type pid_type; /* Kind of process group SIGIO should be sent to */ 907 uid_t uid, euid; /* uid/euid of process setting the owner */ 908 int signum; /* posix.1b rt signal to be delivered on IO */ 909 }; 910 911 /* 912 * Track a single file's readahead state 913 */ 914 struct file_ra_state { 915 pgoff_t start; /* where readahead started */ 916 unsigned int size; /* # of readahead pages */ 917 unsigned int async_size; /* do asynchronous readahead when 918 there are only # of pages ahead */ 919 920 unsigned int ra_pages; /* Maximum readahead window */ 921 unsigned int mmap_miss; /* Cache miss stat for mmap accesses */ 922 loff_t prev_pos; /* Cache last read() position */ 923 }; 924 925 /* 926 * Check if @index falls in the readahead windows. 927 */ 928 static inline int ra_has_index(struct file_ra_state *ra, pgoff_t index) 929 { 930 return (index >= ra->start && 931 index < ra->start + ra->size); 932 } 933 934 #define FILE_MNT_WRITE_TAKEN 1 935 #define FILE_MNT_WRITE_RELEASED 2 936 937 struct file { 938 /* 939 * fu_list becomes invalid after file_free is called and queued via 940 * fu_rcuhead for RCU freeing 941 */ 942 union { 943 struct list_head fu_list; 944 struct rcu_head fu_rcuhead; 945 } f_u; 946 struct path f_path; 947 #define f_dentry f_path.dentry 948 #define f_vfsmnt f_path.mnt 949 const struct file_operations *f_op; 950 spinlock_t f_lock; /* f_ep_links, f_flags, no IRQ */ 951 #ifdef CONFIG_SMP 952 int f_sb_list_cpu; 953 #endif 954 atomic_long_t f_count; 955 unsigned int f_flags; 956 fmode_t f_mode; 957 loff_t f_pos; 958 struct fown_struct f_owner; 959 const struct cred *f_cred; 960 struct file_ra_state f_ra; 961 962 u64 f_version; 963 #ifdef CONFIG_SECURITY 964 void *f_security; 965 #endif 966 /* needed for tty driver, and maybe others */ 967 void *private_data; 968 969 #ifdef CONFIG_EPOLL 970 /* Used by fs/eventpoll.c to link all the hooks to this file */ 971 struct list_head f_ep_links; 972 #endif /* #ifdef CONFIG_EPOLL */ 973 struct address_space *f_mapping; 974 #ifdef CONFIG_DEBUG_WRITECOUNT 975 unsigned long f_mnt_write_state; 976 #endif 977 }; 978 979 struct file_handle { 980 __u32 handle_bytes; 981 int handle_type; 982 /* file identifier */ 983 unsigned char f_handle[0]; 984 }; 985 986 #define get_file(x) atomic_long_inc(&(x)->f_count) 987 #define fput_atomic(x) atomic_long_add_unless(&(x)->f_count, -1, 1) 988 #define file_count(x) atomic_long_read(&(x)->f_count) 989 990 #ifdef CONFIG_DEBUG_WRITECOUNT 991 static inline void file_take_write(struct file *f) 992 { 993 WARN_ON(f->f_mnt_write_state != 0); 994 f->f_mnt_write_state = FILE_MNT_WRITE_TAKEN; 995 } 996 static inline void file_release_write(struct file *f) 997 { 998 f->f_mnt_write_state |= FILE_MNT_WRITE_RELEASED; 999 } 1000 static inline void file_reset_write(struct file *f) 1001 { 1002 f->f_mnt_write_state = 0; 1003 } 1004 static inline void file_check_state(struct file *f) 1005 { 1006 /* 1007 * At this point, either both or neither of these bits 1008 * should be set. 1009 */ 1010 WARN_ON(f->f_mnt_write_state == FILE_MNT_WRITE_TAKEN); 1011 WARN_ON(f->f_mnt_write_state == FILE_MNT_WRITE_RELEASED); 1012 } 1013 static inline int file_check_writeable(struct file *f) 1014 { 1015 if (f->f_mnt_write_state == FILE_MNT_WRITE_TAKEN) 1016 return 0; 1017 printk(KERN_WARNING "writeable file with no " 1018 "mnt_want_write()\n"); 1019 WARN_ON(1); 1020 return -EINVAL; 1021 } 1022 #else /* !CONFIG_DEBUG_WRITECOUNT */ 1023 static inline void file_take_write(struct file *filp) {} 1024 static inline void file_release_write(struct file *filp) {} 1025 static inline void file_reset_write(struct file *filp) {} 1026 static inline void file_check_state(struct file *filp) {} 1027 static inline int file_check_writeable(struct file *filp) 1028 { 1029 return 0; 1030 } 1031 #endif /* CONFIG_DEBUG_WRITECOUNT */ 1032 1033 #define MAX_NON_LFS ((1UL<<31) - 1) 1034 1035 /* Page cache limit. The filesystems should put that into their s_maxbytes 1036 limits, otherwise bad things can happen in VM. */ 1037 #if BITS_PER_LONG==32 1038 #define MAX_LFS_FILESIZE (((u64)PAGE_CACHE_SIZE << (BITS_PER_LONG-1))-1) 1039 #elif BITS_PER_LONG==64 1040 #define MAX_LFS_FILESIZE 0x7fffffffffffffffUL 1041 #endif 1042 1043 #define FL_POSIX 1 1044 #define FL_FLOCK 2 1045 #define FL_ACCESS 8 /* not trying to lock, just looking */ 1046 #define FL_EXISTS 16 /* when unlocking, test for existence */ 1047 #define FL_LEASE 32 /* lease held on this file */ 1048 #define FL_CLOSE 64 /* unlock on close */ 1049 #define FL_SLEEP 128 /* A blocking lock */ 1050 1051 /* 1052 * Special return value from posix_lock_file() and vfs_lock_file() for 1053 * asynchronous locking. 1054 */ 1055 #define FILE_LOCK_DEFERRED 1 1056 1057 /* 1058 * The POSIX file lock owner is determined by 1059 * the "struct files_struct" in the thread group 1060 * (or NULL for no owner - BSD locks). 1061 * 1062 * Lockd stuffs a "host" pointer into this. 1063 */ 1064 typedef struct files_struct *fl_owner_t; 1065 1066 struct file_lock_operations { 1067 void (*fl_copy_lock)(struct file_lock *, struct file_lock *); 1068 void (*fl_release_private)(struct file_lock *); 1069 }; 1070 1071 struct lock_manager_operations { 1072 int (*fl_compare_owner)(struct file_lock *, struct file_lock *); 1073 void (*fl_notify)(struct file_lock *); /* unblock callback */ 1074 int (*fl_grant)(struct file_lock *, struct file_lock *, int); 1075 void (*fl_release_private)(struct file_lock *); 1076 void (*fl_break)(struct file_lock *); 1077 int (*fl_change)(struct file_lock **, int); 1078 }; 1079 1080 struct lock_manager { 1081 struct list_head list; 1082 }; 1083 1084 void locks_start_grace(struct lock_manager *); 1085 void locks_end_grace(struct lock_manager *); 1086 int locks_in_grace(void); 1087 1088 /* that will die - we need it for nfs_lock_info */ 1089 #include <linux/nfs_fs_i.h> 1090 1091 struct file_lock { 1092 struct file_lock *fl_next; /* singly linked list for this inode */ 1093 struct list_head fl_link; /* doubly linked list of all locks */ 1094 struct list_head fl_block; /* circular list of blocked processes */ 1095 fl_owner_t fl_owner; 1096 unsigned char fl_flags; 1097 unsigned char fl_type; 1098 unsigned int fl_pid; 1099 struct pid *fl_nspid; 1100 wait_queue_head_t fl_wait; 1101 struct file *fl_file; 1102 loff_t fl_start; 1103 loff_t fl_end; 1104 1105 struct fasync_struct * fl_fasync; /* for lease break notifications */ 1106 unsigned long fl_break_time; /* for nonblocking lease breaks */ 1107 1108 const struct file_lock_operations *fl_ops; /* Callbacks for filesystems */ 1109 const struct lock_manager_operations *fl_lmops; /* Callbacks for lockmanagers */ 1110 union { 1111 struct nfs_lock_info nfs_fl; 1112 struct nfs4_lock_info nfs4_fl; 1113 struct { 1114 struct list_head link; /* link in AFS vnode's pending_locks list */ 1115 int state; /* state of grant or error if -ve */ 1116 } afs; 1117 } fl_u; 1118 }; 1119 1120 /* The following constant reflects the upper bound of the file/locking space */ 1121 #ifndef OFFSET_MAX 1122 #define INT_LIMIT(x) (~((x)1 << (sizeof(x)*8 - 1))) 1123 #define OFFSET_MAX INT_LIMIT(loff_t) 1124 #define OFFT_OFFSET_MAX INT_LIMIT(off_t) 1125 #endif 1126 1127 #include <linux/fcntl.h> 1128 1129 extern void send_sigio(struct fown_struct *fown, int fd, int band); 1130 1131 #ifdef CONFIG_FILE_LOCKING 1132 extern int fcntl_getlk(struct file *, struct flock __user *); 1133 extern int fcntl_setlk(unsigned int, struct file *, unsigned int, 1134 struct flock __user *); 1135 1136 #if BITS_PER_LONG == 32 1137 extern int fcntl_getlk64(struct file *, struct flock64 __user *); 1138 extern int fcntl_setlk64(unsigned int, struct file *, unsigned int, 1139 struct flock64 __user *); 1140 #endif 1141 1142 extern int fcntl_setlease(unsigned int fd, struct file *filp, long arg); 1143 extern int fcntl_getlease(struct file *filp); 1144 1145 /* fs/locks.c */ 1146 void locks_free_lock(struct file_lock *fl); 1147 extern void locks_init_lock(struct file_lock *); 1148 extern struct file_lock * locks_alloc_lock(void); 1149 extern void locks_copy_lock(struct file_lock *, struct file_lock *); 1150 extern void __locks_copy_lock(struct file_lock *, const struct file_lock *); 1151 extern void locks_remove_posix(struct file *, fl_owner_t); 1152 extern void locks_remove_flock(struct file *); 1153 extern void locks_release_private(struct file_lock *); 1154 extern void posix_test_lock(struct file *, struct file_lock *); 1155 extern int posix_lock_file(struct file *, struct file_lock *, struct file_lock *); 1156 extern int posix_lock_file_wait(struct file *, struct file_lock *); 1157 extern int posix_unblock_lock(struct file *, struct file_lock *); 1158 extern int vfs_test_lock(struct file *, struct file_lock *); 1159 extern int vfs_lock_file(struct file *, unsigned int, struct file_lock *, struct file_lock *); 1160 extern int vfs_cancel_lock(struct file *filp, struct file_lock *fl); 1161 extern int flock_lock_file_wait(struct file *filp, struct file_lock *fl); 1162 extern int __break_lease(struct inode *inode, unsigned int flags); 1163 extern void lease_get_mtime(struct inode *, struct timespec *time); 1164 extern int generic_setlease(struct file *, long, struct file_lock **); 1165 extern int vfs_setlease(struct file *, long, struct file_lock **); 1166 extern int lease_modify(struct file_lock **, int); 1167 extern int lock_may_read(struct inode *, loff_t start, unsigned long count); 1168 extern int lock_may_write(struct inode *, loff_t start, unsigned long count); 1169 extern void lock_flocks(void); 1170 extern void unlock_flocks(void); 1171 #else /* !CONFIG_FILE_LOCKING */ 1172 static inline int fcntl_getlk(struct file *file, struct flock __user *user) 1173 { 1174 return -EINVAL; 1175 } 1176 1177 static inline int fcntl_setlk(unsigned int fd, struct file *file, 1178 unsigned int cmd, struct flock __user *user) 1179 { 1180 return -EACCES; 1181 } 1182 1183 #if BITS_PER_LONG == 32 1184 static inline int fcntl_getlk64(struct file *file, struct flock64 __user *user) 1185 { 1186 return -EINVAL; 1187 } 1188 1189 static inline int fcntl_setlk64(unsigned int fd, struct file *file, 1190 unsigned int cmd, struct flock64 __user *user) 1191 { 1192 return -EACCES; 1193 } 1194 #endif 1195 static inline int fcntl_setlease(unsigned int fd, struct file *filp, long arg) 1196 { 1197 return 0; 1198 } 1199 1200 static inline int fcntl_getlease(struct file *filp) 1201 { 1202 return 0; 1203 } 1204 1205 static inline void locks_init_lock(struct file_lock *fl) 1206 { 1207 return; 1208 } 1209 1210 static inline void __locks_copy_lock(struct file_lock *new, struct file_lock *fl) 1211 { 1212 return; 1213 } 1214 1215 static inline void locks_copy_lock(struct file_lock *new, struct file_lock *fl) 1216 { 1217 return; 1218 } 1219 1220 static inline void locks_remove_posix(struct file *filp, fl_owner_t owner) 1221 { 1222 return; 1223 } 1224 1225 static inline void locks_remove_flock(struct file *filp) 1226 { 1227 return; 1228 } 1229 1230 static inline void posix_test_lock(struct file *filp, struct file_lock *fl) 1231 { 1232 return; 1233 } 1234 1235 static inline int posix_lock_file(struct file *filp, struct file_lock *fl, 1236 struct file_lock *conflock) 1237 { 1238 return -ENOLCK; 1239 } 1240 1241 static inline int posix_lock_file_wait(struct file *filp, struct file_lock *fl) 1242 { 1243 return -ENOLCK; 1244 } 1245 1246 static inline int posix_unblock_lock(struct file *filp, 1247 struct file_lock *waiter) 1248 { 1249 return -ENOENT; 1250 } 1251 1252 static inline int vfs_test_lock(struct file *filp, struct file_lock *fl) 1253 { 1254 return 0; 1255 } 1256 1257 static inline int vfs_lock_file(struct file *filp, unsigned int cmd, 1258 struct file_lock *fl, struct file_lock *conf) 1259 { 1260 return -ENOLCK; 1261 } 1262 1263 static inline int vfs_cancel_lock(struct file *filp, struct file_lock *fl) 1264 { 1265 return 0; 1266 } 1267 1268 static inline int flock_lock_file_wait(struct file *filp, 1269 struct file_lock *request) 1270 { 1271 return -ENOLCK; 1272 } 1273 1274 static inline int __break_lease(struct inode *inode, unsigned int mode) 1275 { 1276 return 0; 1277 } 1278 1279 static inline void lease_get_mtime(struct inode *inode, struct timespec *time) 1280 { 1281 return; 1282 } 1283 1284 static inline int generic_setlease(struct file *filp, long arg, 1285 struct file_lock **flp) 1286 { 1287 return -EINVAL; 1288 } 1289 1290 static inline int vfs_setlease(struct file *filp, long arg, 1291 struct file_lock **lease) 1292 { 1293 return -EINVAL; 1294 } 1295 1296 static inline int lease_modify(struct file_lock **before, int arg) 1297 { 1298 return -EINVAL; 1299 } 1300 1301 static inline int lock_may_read(struct inode *inode, loff_t start, 1302 unsigned long len) 1303 { 1304 return 1; 1305 } 1306 1307 static inline int lock_may_write(struct inode *inode, loff_t start, 1308 unsigned long len) 1309 { 1310 return 1; 1311 } 1312 1313 static inline void lock_flocks(void) 1314 { 1315 } 1316 1317 static inline void unlock_flocks(void) 1318 { 1319 } 1320 1321 #endif /* !CONFIG_FILE_LOCKING */ 1322 1323 1324 struct fasync_struct { 1325 spinlock_t fa_lock; 1326 int magic; 1327 int fa_fd; 1328 struct fasync_struct *fa_next; /* singly linked list */ 1329 struct file *fa_file; 1330 struct rcu_head fa_rcu; 1331 }; 1332 1333 #define FASYNC_MAGIC 0x4601 1334 1335 /* SMP safe fasync helpers: */ 1336 extern int fasync_helper(int, struct file *, int, struct fasync_struct **); 1337 extern struct fasync_struct *fasync_insert_entry(int, struct file *, struct fasync_struct **, struct fasync_struct *); 1338 extern int fasync_remove_entry(struct file *, struct fasync_struct **); 1339 extern struct fasync_struct *fasync_alloc(void); 1340 extern void fasync_free(struct fasync_struct *); 1341 1342 /* can be called from interrupts */ 1343 extern void kill_fasync(struct fasync_struct **, int, int); 1344 1345 extern int __f_setown(struct file *filp, struct pid *, enum pid_type, int force); 1346 extern int f_setown(struct file *filp, unsigned long arg, int force); 1347 extern void f_delown(struct file *filp); 1348 extern pid_t f_getown(struct file *filp); 1349 extern int send_sigurg(struct fown_struct *fown); 1350 1351 /* 1352 * Umount options 1353 */ 1354 1355 #define MNT_FORCE 0x00000001 /* Attempt to forcibily umount */ 1356 #define MNT_DETACH 0x00000002 /* Just detach from the tree */ 1357 #define MNT_EXPIRE 0x00000004 /* Mark for expiry */ 1358 #define UMOUNT_NOFOLLOW 0x00000008 /* Don't follow symlink on umount */ 1359 #define UMOUNT_UNUSED 0x80000000 /* Flag guaranteed to be unused */ 1360 1361 extern struct list_head super_blocks; 1362 extern spinlock_t sb_lock; 1363 1364 struct super_block { 1365 struct list_head s_list; /* Keep this first */ 1366 dev_t s_dev; /* search index; _not_ kdev_t */ 1367 unsigned char s_dirt; 1368 unsigned char s_blocksize_bits; 1369 unsigned long s_blocksize; 1370 loff_t s_maxbytes; /* Max file size */ 1371 struct file_system_type *s_type; 1372 const struct super_operations *s_op; 1373 const struct dquot_operations *dq_op; 1374 const struct quotactl_ops *s_qcop; 1375 const struct export_operations *s_export_op; 1376 unsigned long s_flags; 1377 unsigned long s_magic; 1378 struct dentry *s_root; 1379 struct rw_semaphore s_umount; 1380 struct mutex s_lock; 1381 int s_count; 1382 atomic_t s_active; 1383 #ifdef CONFIG_SECURITY 1384 void *s_security; 1385 #endif 1386 const struct xattr_handler **s_xattr; 1387 1388 struct list_head s_inodes; /* all inodes */ 1389 struct hlist_bl_head s_anon; /* anonymous dentries for (nfs) exporting */ 1390 #ifdef CONFIG_SMP 1391 struct list_head __percpu *s_files; 1392 #else 1393 struct list_head s_files; 1394 #endif 1395 /* s_dentry_lru, s_nr_dentry_unused protected by dcache.c lru locks */ 1396 struct list_head s_dentry_lru; /* unused dentry lru */ 1397 int s_nr_dentry_unused; /* # of dentry on lru */ 1398 1399 struct block_device *s_bdev; 1400 struct backing_dev_info *s_bdi; 1401 struct mtd_info *s_mtd; 1402 struct list_head s_instances; 1403 struct quota_info s_dquot; /* Diskquota specific options */ 1404 1405 int s_frozen; 1406 wait_queue_head_t s_wait_unfrozen; 1407 1408 char s_id[32]; /* Informational name */ 1409 u8 s_uuid[16]; /* UUID */ 1410 1411 void *s_fs_info; /* Filesystem private info */ 1412 fmode_t s_mode; 1413 1414 /* Granularity of c/m/atime in ns. 1415 Cannot be worse than a second */ 1416 u32 s_time_gran; 1417 1418 /* 1419 * The next field is for VFS *only*. No filesystems have any business 1420 * even looking at it. You had been warned. 1421 */ 1422 struct mutex s_vfs_rename_mutex; /* Kludge */ 1423 1424 /* 1425 * Filesystem subtype. If non-empty the filesystem type field 1426 * in /proc/mounts will be "type.subtype" 1427 */ 1428 char *s_subtype; 1429 1430 /* 1431 * Saved mount options for lazy filesystems using 1432 * generic_show_options() 1433 */ 1434 char __rcu *s_options; 1435 const struct dentry_operations *s_d_op; /* default d_op for dentries */ 1436 1437 /* 1438 * Saved pool identifier for cleancache (-1 means none) 1439 */ 1440 int cleancache_poolid; 1441 }; 1442 1443 extern struct timespec current_fs_time(struct super_block *sb); 1444 1445 /* 1446 * Snapshotting support. 1447 */ 1448 enum { 1449 SB_UNFROZEN = 0, 1450 SB_FREEZE_WRITE = 1, 1451 SB_FREEZE_TRANS = 2, 1452 }; 1453 1454 #define vfs_check_frozen(sb, level) \ 1455 wait_event((sb)->s_wait_unfrozen, ((sb)->s_frozen < (level))) 1456 1457 #define get_fs_excl() atomic_inc(&current->fs_excl) 1458 #define put_fs_excl() atomic_dec(&current->fs_excl) 1459 #define has_fs_excl() atomic_read(&current->fs_excl) 1460 1461 /* 1462 * until VFS tracks user namespaces for inodes, just make all files 1463 * belong to init_user_ns 1464 */ 1465 extern struct user_namespace init_user_ns; 1466 #define inode_userns(inode) (&init_user_ns) 1467 extern bool inode_owner_or_capable(const struct inode *inode); 1468 1469 /* not quite ready to be deprecated, but... */ 1470 extern void lock_super(struct super_block *); 1471 extern void unlock_super(struct super_block *); 1472 1473 /* 1474 * VFS helper functions.. 1475 */ 1476 extern int vfs_create(struct inode *, struct dentry *, int, struct nameidata *); 1477 extern int vfs_mkdir(struct inode *, struct dentry *, int); 1478 extern int vfs_mknod(struct inode *, struct dentry *, int, dev_t); 1479 extern int vfs_symlink(struct inode *, struct dentry *, const char *); 1480 extern int vfs_link(struct dentry *, struct inode *, struct dentry *); 1481 extern int vfs_rmdir(struct inode *, struct dentry *); 1482 extern int vfs_unlink(struct inode *, struct dentry *); 1483 extern int vfs_rename(struct inode *, struct dentry *, struct inode *, struct dentry *); 1484 1485 /* 1486 * VFS dentry helper functions. 1487 */ 1488 extern void dentry_unhash(struct dentry *dentry); 1489 1490 /* 1491 * VFS file helper functions. 1492 */ 1493 extern int file_permission(struct file *, int); 1494 extern void inode_init_owner(struct inode *inode, const struct inode *dir, 1495 mode_t mode); 1496 /* 1497 * VFS FS_IOC_FIEMAP helper definitions. 1498 */ 1499 struct fiemap_extent_info { 1500 unsigned int fi_flags; /* Flags as passed from user */ 1501 unsigned int fi_extents_mapped; /* Number of mapped extents */ 1502 unsigned int fi_extents_max; /* Size of fiemap_extent array */ 1503 struct fiemap_extent __user *fi_extents_start; /* Start of 1504 fiemap_extent array */ 1505 }; 1506 int fiemap_fill_next_extent(struct fiemap_extent_info *info, u64 logical, 1507 u64 phys, u64 len, u32 flags); 1508 int fiemap_check_flags(struct fiemap_extent_info *fieinfo, u32 fs_flags); 1509 1510 /* 1511 * File types 1512 * 1513 * NOTE! These match bits 12..15 of stat.st_mode 1514 * (ie "(i_mode >> 12) & 15"). 1515 */ 1516 #define DT_UNKNOWN 0 1517 #define DT_FIFO 1 1518 #define DT_CHR 2 1519 #define DT_DIR 4 1520 #define DT_BLK 6 1521 #define DT_REG 8 1522 #define DT_LNK 10 1523 #define DT_SOCK 12 1524 #define DT_WHT 14 1525 1526 /* 1527 * This is the "filldir" function type, used by readdir() to let 1528 * the kernel specify what kind of dirent layout it wants to have. 1529 * This allows the kernel to read directories into kernel space or 1530 * to have different dirent layouts depending on the binary type. 1531 */ 1532 typedef int (*filldir_t)(void *, const char *, int, loff_t, u64, unsigned); 1533 struct block_device_operations; 1534 1535 /* These macros are for out of kernel modules to test that 1536 * the kernel supports the unlocked_ioctl and compat_ioctl 1537 * fields in struct file_operations. */ 1538 #define HAVE_COMPAT_IOCTL 1 1539 #define HAVE_UNLOCKED_IOCTL 1 1540 1541 /* 1542 * NOTE: 1543 * all file operations except setlease can be called without 1544 * the big kernel lock held in all filesystems. 1545 */ 1546 struct file_operations { 1547 struct module *owner; 1548 loff_t (*llseek) (struct file *, loff_t, int); 1549 ssize_t (*read) (struct file *, char __user *, size_t, loff_t *); 1550 ssize_t (*write) (struct file *, const char __user *, size_t, loff_t *); 1551 ssize_t (*aio_read) (struct kiocb *, const struct iovec *, unsigned long, loff_t); 1552 ssize_t (*aio_write) (struct kiocb *, const struct iovec *, unsigned long, loff_t); 1553 int (*readdir) (struct file *, void *, filldir_t); 1554 unsigned int (*poll) (struct file *, struct poll_table_struct *); 1555 long (*unlocked_ioctl) (struct file *, unsigned int, unsigned long); 1556 long (*compat_ioctl) (struct file *, unsigned int, unsigned long); 1557 int (*mmap) (struct file *, struct vm_area_struct *); 1558 int (*open) (struct inode *, struct file *); 1559 int (*flush) (struct file *, fl_owner_t id); 1560 int (*release) (struct inode *, struct file *); 1561 int (*fsync) (struct file *, int datasync); 1562 int (*aio_fsync) (struct kiocb *, int datasync); 1563 int (*fasync) (int, struct file *, int); 1564 int (*lock) (struct file *, int, struct file_lock *); 1565 ssize_t (*sendpage) (struct file *, struct page *, int, size_t, loff_t *, int); 1566 unsigned long (*get_unmapped_area)(struct file *, unsigned long, unsigned long, unsigned long, unsigned long); 1567 int (*check_flags)(int); 1568 int (*flock) (struct file *, int, struct file_lock *); 1569 ssize_t (*splice_write)(struct pipe_inode_info *, struct file *, loff_t *, size_t, unsigned int); 1570 ssize_t (*splice_read)(struct file *, loff_t *, struct pipe_inode_info *, size_t, unsigned int); 1571 int (*setlease)(struct file *, long, struct file_lock **); 1572 long (*fallocate)(struct file *file, int mode, loff_t offset, 1573 loff_t len); 1574 }; 1575 1576 #define IPERM_FLAG_RCU 0x0001 1577 1578 struct inode_operations { 1579 struct dentry * (*lookup) (struct inode *,struct dentry *, struct nameidata *); 1580 void * (*follow_link) (struct dentry *, struct nameidata *); 1581 int (*permission) (struct inode *, int, unsigned int); 1582 int (*check_acl)(struct inode *, int, unsigned int); 1583 1584 int (*readlink) (struct dentry *, char __user *,int); 1585 void (*put_link) (struct dentry *, struct nameidata *, void *); 1586 1587 int (*create) (struct inode *,struct dentry *,int, struct nameidata *); 1588 int (*link) (struct dentry *,struct inode *,struct dentry *); 1589 int (*unlink) (struct inode *,struct dentry *); 1590 int (*symlink) (struct inode *,struct dentry *,const char *); 1591 int (*mkdir) (struct inode *,struct dentry *,int); 1592 int (*rmdir) (struct inode *,struct dentry *); 1593 int (*mknod) (struct inode *,struct dentry *,int,dev_t); 1594 int (*rename) (struct inode *, struct dentry *, 1595 struct inode *, struct dentry *); 1596 void (*truncate) (struct inode *); 1597 int (*setattr) (struct dentry *, struct iattr *); 1598 int (*getattr) (struct vfsmount *mnt, struct dentry *, struct kstat *); 1599 int (*setxattr) (struct dentry *, const char *,const void *,size_t,int); 1600 ssize_t (*getxattr) (struct dentry *, const char *, void *, size_t); 1601 ssize_t (*listxattr) (struct dentry *, char *, size_t); 1602 int (*removexattr) (struct dentry *, const char *); 1603 void (*truncate_range)(struct inode *, loff_t, loff_t); 1604 int (*fiemap)(struct inode *, struct fiemap_extent_info *, u64 start, 1605 u64 len); 1606 } ____cacheline_aligned; 1607 1608 struct seq_file; 1609 1610 ssize_t rw_copy_check_uvector(int type, const struct iovec __user * uvector, 1611 unsigned long nr_segs, unsigned long fast_segs, 1612 struct iovec *fast_pointer, 1613 struct iovec **ret_pointer); 1614 1615 extern ssize_t vfs_read(struct file *, char __user *, size_t, loff_t *); 1616 extern ssize_t vfs_write(struct file *, const char __user *, size_t, loff_t *); 1617 extern ssize_t vfs_readv(struct file *, const struct iovec __user *, 1618 unsigned long, loff_t *); 1619 extern ssize_t vfs_writev(struct file *, const struct iovec __user *, 1620 unsigned long, loff_t *); 1621 1622 struct super_operations { 1623 struct inode *(*alloc_inode)(struct super_block *sb); 1624 void (*destroy_inode)(struct inode *); 1625 1626 void (*dirty_inode) (struct inode *, int flags); 1627 int (*write_inode) (struct inode *, struct writeback_control *wbc); 1628 int (*drop_inode) (struct inode *); 1629 void (*evict_inode) (struct inode *); 1630 void (*put_super) (struct super_block *); 1631 void (*write_super) (struct super_block *); 1632 int (*sync_fs)(struct super_block *sb, int wait); 1633 int (*freeze_fs) (struct super_block *); 1634 int (*unfreeze_fs) (struct super_block *); 1635 int (*statfs) (struct dentry *, struct kstatfs *); 1636 int (*remount_fs) (struct super_block *, int *, char *); 1637 void (*umount_begin) (struct super_block *); 1638 1639 int (*show_options)(struct seq_file *, struct vfsmount *); 1640 int (*show_devname)(struct seq_file *, struct vfsmount *); 1641 int (*show_path)(struct seq_file *, struct vfsmount *); 1642 int (*show_stats)(struct seq_file *, struct vfsmount *); 1643 #ifdef CONFIG_QUOTA 1644 ssize_t (*quota_read)(struct super_block *, int, char *, size_t, loff_t); 1645 ssize_t (*quota_write)(struct super_block *, int, const char *, size_t, loff_t); 1646 #endif 1647 int (*bdev_try_to_free_page)(struct super_block*, struct page*, gfp_t); 1648 }; 1649 1650 /* 1651 * Inode state bits. Protected by inode->i_lock 1652 * 1653 * Three bits determine the dirty state of the inode, I_DIRTY_SYNC, 1654 * I_DIRTY_DATASYNC and I_DIRTY_PAGES. 1655 * 1656 * Four bits define the lifetime of an inode. Initially, inodes are I_NEW, 1657 * until that flag is cleared. I_WILL_FREE, I_FREEING and I_CLEAR are set at 1658 * various stages of removing an inode. 1659 * 1660 * Two bits are used for locking and completion notification, I_NEW and I_SYNC. 1661 * 1662 * I_DIRTY_SYNC Inode is dirty, but doesn't have to be written on 1663 * fdatasync(). i_atime is the usual cause. 1664 * I_DIRTY_DATASYNC Data-related inode changes pending. We keep track of 1665 * these changes separately from I_DIRTY_SYNC so that we 1666 * don't have to write inode on fdatasync() when only 1667 * mtime has changed in it. 1668 * I_DIRTY_PAGES Inode has dirty pages. Inode itself may be clean. 1669 * I_NEW Serves as both a mutex and completion notification. 1670 * New inodes set I_NEW. If two processes both create 1671 * the same inode, one of them will release its inode and 1672 * wait for I_NEW to be released before returning. 1673 * Inodes in I_WILL_FREE, I_FREEING or I_CLEAR state can 1674 * also cause waiting on I_NEW, without I_NEW actually 1675 * being set. find_inode() uses this to prevent returning 1676 * nearly-dead inodes. 1677 * I_WILL_FREE Must be set when calling write_inode_now() if i_count 1678 * is zero. I_FREEING must be set when I_WILL_FREE is 1679 * cleared. 1680 * I_FREEING Set when inode is about to be freed but still has dirty 1681 * pages or buffers attached or the inode itself is still 1682 * dirty. 1683 * I_CLEAR Added by end_writeback(). In this state the inode is clean 1684 * and can be destroyed. Inode keeps I_FREEING. 1685 * 1686 * Inodes that are I_WILL_FREE, I_FREEING or I_CLEAR are 1687 * prohibited for many purposes. iget() must wait for 1688 * the inode to be completely released, then create it 1689 * anew. Other functions will just ignore such inodes, 1690 * if appropriate. I_NEW is used for waiting. 1691 * 1692 * I_SYNC Synchonized write of dirty inode data. The bits is 1693 * set during data writeback, and cleared with a wakeup 1694 * on the bit address once it is done. 1695 * 1696 * Q: What is the difference between I_WILL_FREE and I_FREEING? 1697 */ 1698 #define I_DIRTY_SYNC (1 << 0) 1699 #define I_DIRTY_DATASYNC (1 << 1) 1700 #define I_DIRTY_PAGES (1 << 2) 1701 #define __I_NEW 3 1702 #define I_NEW (1 << __I_NEW) 1703 #define I_WILL_FREE (1 << 4) 1704 #define I_FREEING (1 << 5) 1705 #define I_CLEAR (1 << 6) 1706 #define __I_SYNC 7 1707 #define I_SYNC (1 << __I_SYNC) 1708 #define I_REFERENCED (1 << 8) 1709 1710 #define I_DIRTY (I_DIRTY_SYNC | I_DIRTY_DATASYNC | I_DIRTY_PAGES) 1711 1712 extern void __mark_inode_dirty(struct inode *, int); 1713 static inline void mark_inode_dirty(struct inode *inode) 1714 { 1715 __mark_inode_dirty(inode, I_DIRTY); 1716 } 1717 1718 static inline void mark_inode_dirty_sync(struct inode *inode) 1719 { 1720 __mark_inode_dirty(inode, I_DIRTY_SYNC); 1721 } 1722 1723 /** 1724 * inc_nlink - directly increment an inode's link count 1725 * @inode: inode 1726 * 1727 * This is a low-level filesystem helper to replace any 1728 * direct filesystem manipulation of i_nlink. Currently, 1729 * it is only here for parity with dec_nlink(). 1730 */ 1731 static inline void inc_nlink(struct inode *inode) 1732 { 1733 inode->i_nlink++; 1734 } 1735 1736 static inline void inode_inc_link_count(struct inode *inode) 1737 { 1738 inc_nlink(inode); 1739 mark_inode_dirty(inode); 1740 } 1741 1742 /** 1743 * drop_nlink - directly drop an inode's link count 1744 * @inode: inode 1745 * 1746 * This is a low-level filesystem helper to replace any 1747 * direct filesystem manipulation of i_nlink. In cases 1748 * where we are attempting to track writes to the 1749 * filesystem, a decrement to zero means an imminent 1750 * write when the file is truncated and actually unlinked 1751 * on the filesystem. 1752 */ 1753 static inline void drop_nlink(struct inode *inode) 1754 { 1755 inode->i_nlink--; 1756 } 1757 1758 /** 1759 * clear_nlink - directly zero an inode's link count 1760 * @inode: inode 1761 * 1762 * This is a low-level filesystem helper to replace any 1763 * direct filesystem manipulation of i_nlink. See 1764 * drop_nlink() for why we care about i_nlink hitting zero. 1765 */ 1766 static inline void clear_nlink(struct inode *inode) 1767 { 1768 inode->i_nlink = 0; 1769 } 1770 1771 static inline void inode_dec_link_count(struct inode *inode) 1772 { 1773 drop_nlink(inode); 1774 mark_inode_dirty(inode); 1775 } 1776 1777 /** 1778 * inode_inc_iversion - increments i_version 1779 * @inode: inode that need to be updated 1780 * 1781 * Every time the inode is modified, the i_version field will be incremented. 1782 * The filesystem has to be mounted with i_version flag 1783 */ 1784 1785 static inline void inode_inc_iversion(struct inode *inode) 1786 { 1787 spin_lock(&inode->i_lock); 1788 inode->i_version++; 1789 spin_unlock(&inode->i_lock); 1790 } 1791 1792 extern void touch_atime(struct vfsmount *mnt, struct dentry *dentry); 1793 static inline void file_accessed(struct file *file) 1794 { 1795 if (!(file->f_flags & O_NOATIME)) 1796 touch_atime(file->f_path.mnt, file->f_path.dentry); 1797 } 1798 1799 int sync_inode(struct inode *inode, struct writeback_control *wbc); 1800 int sync_inode_metadata(struct inode *inode, int wait); 1801 1802 struct file_system_type { 1803 const char *name; 1804 int fs_flags; 1805 struct dentry *(*mount) (struct file_system_type *, int, 1806 const char *, void *); 1807 void (*kill_sb) (struct super_block *); 1808 struct module *owner; 1809 struct file_system_type * next; 1810 struct list_head fs_supers; 1811 1812 struct lock_class_key s_lock_key; 1813 struct lock_class_key s_umount_key; 1814 struct lock_class_key s_vfs_rename_key; 1815 1816 struct lock_class_key i_lock_key; 1817 struct lock_class_key i_mutex_key; 1818 struct lock_class_key i_mutex_dir_key; 1819 struct lock_class_key i_alloc_sem_key; 1820 }; 1821 1822 extern struct dentry *mount_ns(struct file_system_type *fs_type, int flags, 1823 void *data, int (*fill_super)(struct super_block *, void *, int)); 1824 extern struct dentry *mount_bdev(struct file_system_type *fs_type, 1825 int flags, const char *dev_name, void *data, 1826 int (*fill_super)(struct super_block *, void *, int)); 1827 extern struct dentry *mount_single(struct file_system_type *fs_type, 1828 int flags, void *data, 1829 int (*fill_super)(struct super_block *, void *, int)); 1830 extern struct dentry *mount_nodev(struct file_system_type *fs_type, 1831 int flags, void *data, 1832 int (*fill_super)(struct super_block *, void *, int)); 1833 void generic_shutdown_super(struct super_block *sb); 1834 void kill_block_super(struct super_block *sb); 1835 void kill_anon_super(struct super_block *sb); 1836 void kill_litter_super(struct super_block *sb); 1837 void deactivate_super(struct super_block *sb); 1838 void deactivate_locked_super(struct super_block *sb); 1839 int set_anon_super(struct super_block *s, void *data); 1840 struct super_block *sget(struct file_system_type *type, 1841 int (*test)(struct super_block *,void *), 1842 int (*set)(struct super_block *,void *), 1843 void *data); 1844 extern struct dentry *mount_pseudo(struct file_system_type *, char *, 1845 const struct super_operations *ops, 1846 const struct dentry_operations *dops, 1847 unsigned long); 1848 1849 static inline void sb_mark_dirty(struct super_block *sb) 1850 { 1851 sb->s_dirt = 1; 1852 } 1853 static inline void sb_mark_clean(struct super_block *sb) 1854 { 1855 sb->s_dirt = 0; 1856 } 1857 static inline int sb_is_dirty(struct super_block *sb) 1858 { 1859 return sb->s_dirt; 1860 } 1861 1862 /* Alas, no aliases. Too much hassle with bringing module.h everywhere */ 1863 #define fops_get(fops) \ 1864 (((fops) && try_module_get((fops)->owner) ? (fops) : NULL)) 1865 #define fops_put(fops) \ 1866 do { if (fops) module_put((fops)->owner); } while(0) 1867 1868 extern int register_filesystem(struct file_system_type *); 1869 extern int unregister_filesystem(struct file_system_type *); 1870 extern struct vfsmount *kern_mount_data(struct file_system_type *, void *data); 1871 #define kern_mount(type) kern_mount_data(type, NULL) 1872 extern int may_umount_tree(struct vfsmount *); 1873 extern int may_umount(struct vfsmount *); 1874 extern long do_mount(char *, char *, char *, unsigned long, void *); 1875 extern struct vfsmount *collect_mounts(struct path *); 1876 extern void drop_collected_mounts(struct vfsmount *); 1877 extern int iterate_mounts(int (*)(struct vfsmount *, void *), void *, 1878 struct vfsmount *); 1879 extern int vfs_statfs(struct path *, struct kstatfs *); 1880 extern int user_statfs(const char __user *, struct kstatfs *); 1881 extern int fd_statfs(int, struct kstatfs *); 1882 extern int statfs_by_dentry(struct dentry *, struct kstatfs *); 1883 extern int freeze_super(struct super_block *super); 1884 extern int thaw_super(struct super_block *super); 1885 1886 extern int current_umask(void); 1887 1888 /* /sys/fs */ 1889 extern struct kobject *fs_kobj; 1890 1891 #define MAX_RW_COUNT (INT_MAX & PAGE_CACHE_MASK) 1892 extern int rw_verify_area(int, struct file *, loff_t *, size_t); 1893 1894 #define FLOCK_VERIFY_READ 1 1895 #define FLOCK_VERIFY_WRITE 2 1896 1897 #ifdef CONFIG_FILE_LOCKING 1898 extern int locks_mandatory_locked(struct inode *); 1899 extern int locks_mandatory_area(int, struct inode *, struct file *, loff_t, size_t); 1900 1901 /* 1902 * Candidates for mandatory locking have the setgid bit set 1903 * but no group execute bit - an otherwise meaningless combination. 1904 */ 1905 1906 static inline int __mandatory_lock(struct inode *ino) 1907 { 1908 return (ino->i_mode & (S_ISGID | S_IXGRP)) == S_ISGID; 1909 } 1910 1911 /* 1912 * ... and these candidates should be on MS_MANDLOCK mounted fs, 1913 * otherwise these will be advisory locks 1914 */ 1915 1916 static inline int mandatory_lock(struct inode *ino) 1917 { 1918 return IS_MANDLOCK(ino) && __mandatory_lock(ino); 1919 } 1920 1921 static inline int locks_verify_locked(struct inode *inode) 1922 { 1923 if (mandatory_lock(inode)) 1924 return locks_mandatory_locked(inode); 1925 return 0; 1926 } 1927 1928 static inline int locks_verify_truncate(struct inode *inode, 1929 struct file *filp, 1930 loff_t size) 1931 { 1932 if (inode->i_flock && mandatory_lock(inode)) 1933 return locks_mandatory_area( 1934 FLOCK_VERIFY_WRITE, inode, filp, 1935 size < inode->i_size ? size : inode->i_size, 1936 (size < inode->i_size ? inode->i_size - size 1937 : size - inode->i_size) 1938 ); 1939 return 0; 1940 } 1941 1942 static inline int break_lease(struct inode *inode, unsigned int mode) 1943 { 1944 if (inode->i_flock) 1945 return __break_lease(inode, mode); 1946 return 0; 1947 } 1948 #else /* !CONFIG_FILE_LOCKING */ 1949 static inline int locks_mandatory_locked(struct inode *inode) 1950 { 1951 return 0; 1952 } 1953 1954 static inline int locks_mandatory_area(int rw, struct inode *inode, 1955 struct file *filp, loff_t offset, 1956 size_t count) 1957 { 1958 return 0; 1959 } 1960 1961 static inline int __mandatory_lock(struct inode *inode) 1962 { 1963 return 0; 1964 } 1965 1966 static inline int mandatory_lock(struct inode *inode) 1967 { 1968 return 0; 1969 } 1970 1971 static inline int locks_verify_locked(struct inode *inode) 1972 { 1973 return 0; 1974 } 1975 1976 static inline int locks_verify_truncate(struct inode *inode, struct file *filp, 1977 size_t size) 1978 { 1979 return 0; 1980 } 1981 1982 static inline int break_lease(struct inode *inode, unsigned int mode) 1983 { 1984 return 0; 1985 } 1986 1987 #endif /* CONFIG_FILE_LOCKING */ 1988 1989 /* fs/open.c */ 1990 1991 extern int do_truncate(struct dentry *, loff_t start, unsigned int time_attrs, 1992 struct file *filp); 1993 extern int do_fallocate(struct file *file, int mode, loff_t offset, 1994 loff_t len); 1995 extern long do_sys_open(int dfd, const char __user *filename, int flags, 1996 int mode); 1997 extern struct file *filp_open(const char *, int, int); 1998 extern struct file *file_open_root(struct dentry *, struct vfsmount *, 1999 const char *, int); 2000 extern struct file * dentry_open(struct dentry *, struct vfsmount *, int, 2001 const struct cred *); 2002 extern int filp_close(struct file *, fl_owner_t id); 2003 extern char * getname(const char __user *); 2004 2005 /* fs/ioctl.c */ 2006 2007 extern int ioctl_preallocate(struct file *filp, void __user *argp); 2008 2009 /* fs/dcache.c */ 2010 extern void __init vfs_caches_init_early(void); 2011 extern void __init vfs_caches_init(unsigned long); 2012 2013 extern struct kmem_cache *names_cachep; 2014 2015 #define __getname_gfp(gfp) kmem_cache_alloc(names_cachep, (gfp)) 2016 #define __getname() __getname_gfp(GFP_KERNEL) 2017 #define __putname(name) kmem_cache_free(names_cachep, (void *)(name)) 2018 #ifndef CONFIG_AUDITSYSCALL 2019 #define putname(name) __putname(name) 2020 #else 2021 extern void putname(const char *name); 2022 #endif 2023 2024 #ifdef CONFIG_BLOCK 2025 extern int register_blkdev(unsigned int, const char *); 2026 extern void unregister_blkdev(unsigned int, const char *); 2027 extern struct block_device *bdget(dev_t); 2028 extern struct block_device *bdgrab(struct block_device *bdev); 2029 extern void bd_set_size(struct block_device *, loff_t size); 2030 extern void bd_forget(struct inode *inode); 2031 extern void bdput(struct block_device *); 2032 extern void invalidate_bdev(struct block_device *); 2033 extern int sync_blockdev(struct block_device *bdev); 2034 extern struct super_block *freeze_bdev(struct block_device *); 2035 extern void emergency_thaw_all(void); 2036 extern int thaw_bdev(struct block_device *bdev, struct super_block *sb); 2037 extern int fsync_bdev(struct block_device *); 2038 #else 2039 static inline void bd_forget(struct inode *inode) {} 2040 static inline int sync_blockdev(struct block_device *bdev) { return 0; } 2041 static inline void invalidate_bdev(struct block_device *bdev) {} 2042 2043 static inline struct super_block *freeze_bdev(struct block_device *sb) 2044 { 2045 return NULL; 2046 } 2047 2048 static inline int thaw_bdev(struct block_device *bdev, struct super_block *sb) 2049 { 2050 return 0; 2051 } 2052 #endif 2053 extern int sync_filesystem(struct super_block *); 2054 extern const struct file_operations def_blk_fops; 2055 extern const struct file_operations def_chr_fops; 2056 extern const struct file_operations bad_sock_fops; 2057 extern const struct file_operations def_fifo_fops; 2058 #ifdef CONFIG_BLOCK 2059 extern int ioctl_by_bdev(struct block_device *, unsigned, unsigned long); 2060 extern int blkdev_ioctl(struct block_device *, fmode_t, unsigned, unsigned long); 2061 extern long compat_blkdev_ioctl(struct file *, unsigned, unsigned long); 2062 extern int blkdev_get(struct block_device *bdev, fmode_t mode, void *holder); 2063 extern struct block_device *blkdev_get_by_path(const char *path, fmode_t mode, 2064 void *holder); 2065 extern struct block_device *blkdev_get_by_dev(dev_t dev, fmode_t mode, 2066 void *holder); 2067 extern int blkdev_put(struct block_device *bdev, fmode_t mode); 2068 #ifdef CONFIG_SYSFS 2069 extern int bd_link_disk_holder(struct block_device *bdev, struct gendisk *disk); 2070 extern void bd_unlink_disk_holder(struct block_device *bdev, 2071 struct gendisk *disk); 2072 #else 2073 static inline int bd_link_disk_holder(struct block_device *bdev, 2074 struct gendisk *disk) 2075 { 2076 return 0; 2077 } 2078 static inline void bd_unlink_disk_holder(struct block_device *bdev, 2079 struct gendisk *disk) 2080 { 2081 } 2082 #endif 2083 #endif 2084 2085 /* fs/char_dev.c */ 2086 #define CHRDEV_MAJOR_HASH_SIZE 255 2087 extern int alloc_chrdev_region(dev_t *, unsigned, unsigned, const char *); 2088 extern int register_chrdev_region(dev_t, unsigned, const char *); 2089 extern int __register_chrdev(unsigned int major, unsigned int baseminor, 2090 unsigned int count, const char *name, 2091 const struct file_operations *fops); 2092 extern void __unregister_chrdev(unsigned int major, unsigned int baseminor, 2093 unsigned int count, const char *name); 2094 extern void unregister_chrdev_region(dev_t, unsigned); 2095 extern void chrdev_show(struct seq_file *,off_t); 2096 2097 static inline int register_chrdev(unsigned int major, const char *name, 2098 const struct file_operations *fops) 2099 { 2100 return __register_chrdev(major, 0, 256, name, fops); 2101 } 2102 2103 static inline void unregister_chrdev(unsigned int major, const char *name) 2104 { 2105 __unregister_chrdev(major, 0, 256, name); 2106 } 2107 2108 /* fs/block_dev.c */ 2109 #define BDEVNAME_SIZE 32 /* Largest string for a blockdev identifier */ 2110 #define BDEVT_SIZE 10 /* Largest string for MAJ:MIN for blkdev */ 2111 2112 #ifdef CONFIG_BLOCK 2113 #define BLKDEV_MAJOR_HASH_SIZE 255 2114 extern const char *__bdevname(dev_t, char *buffer); 2115 extern const char *bdevname(struct block_device *bdev, char *buffer); 2116 extern struct block_device *lookup_bdev(const char *); 2117 extern void blkdev_show(struct seq_file *,off_t); 2118 2119 #else 2120 #define BLKDEV_MAJOR_HASH_SIZE 0 2121 #endif 2122 2123 extern void init_special_inode(struct inode *, umode_t, dev_t); 2124 2125 /* Invalid inode operations -- fs/bad_inode.c */ 2126 extern void make_bad_inode(struct inode *); 2127 extern int is_bad_inode(struct inode *); 2128 2129 extern const struct file_operations read_pipefifo_fops; 2130 extern const struct file_operations write_pipefifo_fops; 2131 extern const struct file_operations rdwr_pipefifo_fops; 2132 2133 extern int fs_may_remount_ro(struct super_block *); 2134 2135 #ifdef CONFIG_BLOCK 2136 /* 2137 * return READ, READA, or WRITE 2138 */ 2139 #define bio_rw(bio) ((bio)->bi_rw & (RW_MASK | RWA_MASK)) 2140 2141 /* 2142 * return data direction, READ or WRITE 2143 */ 2144 #define bio_data_dir(bio) ((bio)->bi_rw & 1) 2145 2146 extern void check_disk_size_change(struct gendisk *disk, 2147 struct block_device *bdev); 2148 extern int revalidate_disk(struct gendisk *); 2149 extern int check_disk_change(struct block_device *); 2150 extern int __invalidate_device(struct block_device *, bool); 2151 extern int invalidate_partition(struct gendisk *, int); 2152 #endif 2153 unsigned long invalidate_mapping_pages(struct address_space *mapping, 2154 pgoff_t start, pgoff_t end); 2155 2156 static inline void invalidate_remote_inode(struct inode *inode) 2157 { 2158 if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || 2159 S_ISLNK(inode->i_mode)) 2160 invalidate_mapping_pages(inode->i_mapping, 0, -1); 2161 } 2162 extern int invalidate_inode_pages2(struct address_space *mapping); 2163 extern int invalidate_inode_pages2_range(struct address_space *mapping, 2164 pgoff_t start, pgoff_t end); 2165 extern int write_inode_now(struct inode *, int); 2166 extern int filemap_fdatawrite(struct address_space *); 2167 extern int filemap_flush(struct address_space *); 2168 extern int filemap_fdatawait(struct address_space *); 2169 extern int filemap_fdatawait_range(struct address_space *, loff_t lstart, 2170 loff_t lend); 2171 extern int filemap_write_and_wait(struct address_space *mapping); 2172 extern int filemap_write_and_wait_range(struct address_space *mapping, 2173 loff_t lstart, loff_t lend); 2174 extern int __filemap_fdatawrite_range(struct address_space *mapping, 2175 loff_t start, loff_t end, int sync_mode); 2176 extern int filemap_fdatawrite_range(struct address_space *mapping, 2177 loff_t start, loff_t end); 2178 2179 extern int vfs_fsync_range(struct file *file, loff_t start, loff_t end, 2180 int datasync); 2181 extern int vfs_fsync(struct file *file, int datasync); 2182 extern int generic_write_sync(struct file *file, loff_t pos, loff_t count); 2183 extern void sync_supers(void); 2184 extern void emergency_sync(void); 2185 extern void emergency_remount(void); 2186 #ifdef CONFIG_BLOCK 2187 extern sector_t bmap(struct inode *, sector_t); 2188 #endif 2189 extern int notify_change(struct dentry *, struct iattr *); 2190 extern int inode_permission(struct inode *, int); 2191 extern int generic_permission(struct inode *, int, unsigned int, 2192 int (*check_acl)(struct inode *, int, unsigned int)); 2193 2194 static inline bool execute_ok(struct inode *inode) 2195 { 2196 return (inode->i_mode & S_IXUGO) || S_ISDIR(inode->i_mode); 2197 } 2198 2199 extern int get_write_access(struct inode *); 2200 extern int deny_write_access(struct file *); 2201 static inline void put_write_access(struct inode * inode) 2202 { 2203 atomic_dec(&inode->i_writecount); 2204 } 2205 static inline void allow_write_access(struct file *file) 2206 { 2207 if (file) 2208 atomic_inc(&file->f_path.dentry->d_inode->i_writecount); 2209 } 2210 #ifdef CONFIG_IMA 2211 static inline void i_readcount_dec(struct inode *inode) 2212 { 2213 BUG_ON(!atomic_read(&inode->i_readcount)); 2214 atomic_dec(&inode->i_readcount); 2215 } 2216 static inline void i_readcount_inc(struct inode *inode) 2217 { 2218 atomic_inc(&inode->i_readcount); 2219 } 2220 #else 2221 static inline void i_readcount_dec(struct inode *inode) 2222 { 2223 return; 2224 } 2225 static inline void i_readcount_inc(struct inode *inode) 2226 { 2227 return; 2228 } 2229 #endif 2230 extern int do_pipe_flags(int *, int); 2231 extern struct file *create_read_pipe(struct file *f, int flags); 2232 extern struct file *create_write_pipe(int flags); 2233 extern void free_write_pipe(struct file *); 2234 2235 extern int kernel_read(struct file *, loff_t, char *, unsigned long); 2236 extern struct file * open_exec(const char *); 2237 2238 /* fs/dcache.c -- generic fs support functions */ 2239 extern int is_subdir(struct dentry *, struct dentry *); 2240 extern int path_is_under(struct path *, struct path *); 2241 extern ino_t find_inode_number(struct dentry *, struct qstr *); 2242 2243 #include <linux/err.h> 2244 2245 /* needed for stackable file system support */ 2246 extern loff_t default_llseek(struct file *file, loff_t offset, int origin); 2247 2248 extern loff_t vfs_llseek(struct file *file, loff_t offset, int origin); 2249 2250 extern int inode_init_always(struct super_block *, struct inode *); 2251 extern void inode_init_once(struct inode *); 2252 extern void address_space_init_once(struct address_space *mapping); 2253 extern void ihold(struct inode * inode); 2254 extern void iput(struct inode *); 2255 extern struct inode * igrab(struct inode *); 2256 extern ino_t iunique(struct super_block *, ino_t); 2257 extern int inode_needs_sync(struct inode *inode); 2258 extern int generic_delete_inode(struct inode *inode); 2259 extern int generic_drop_inode(struct inode *inode); 2260 2261 extern struct inode *ilookup5_nowait(struct super_block *sb, 2262 unsigned long hashval, int (*test)(struct inode *, void *), 2263 void *data); 2264 extern struct inode *ilookup5(struct super_block *sb, unsigned long hashval, 2265 int (*test)(struct inode *, void *), void *data); 2266 extern struct inode *ilookup(struct super_block *sb, unsigned long ino); 2267 2268 extern struct inode * iget5_locked(struct super_block *, unsigned long, int (*test)(struct inode *, void *), int (*set)(struct inode *, void *), void *); 2269 extern struct inode * iget_locked(struct super_block *, unsigned long); 2270 extern int insert_inode_locked4(struct inode *, unsigned long, int (*test)(struct inode *, void *), void *); 2271 extern int insert_inode_locked(struct inode *); 2272 extern void unlock_new_inode(struct inode *); 2273 extern unsigned int get_next_ino(void); 2274 2275 extern void __iget(struct inode * inode); 2276 extern void iget_failed(struct inode *); 2277 extern void end_writeback(struct inode *); 2278 extern void __destroy_inode(struct inode *); 2279 extern struct inode *new_inode(struct super_block *); 2280 extern void free_inode_nonrcu(struct inode *inode); 2281 extern int should_remove_suid(struct dentry *); 2282 extern int file_remove_suid(struct file *); 2283 2284 extern void __insert_inode_hash(struct inode *, unsigned long hashval); 2285 extern void remove_inode_hash(struct inode *); 2286 static inline void insert_inode_hash(struct inode *inode) 2287 { 2288 __insert_inode_hash(inode, inode->i_ino); 2289 } 2290 extern void inode_sb_list_add(struct inode *inode); 2291 2292 #ifdef CONFIG_BLOCK 2293 extern void submit_bio(int, struct bio *); 2294 extern int bdev_read_only(struct block_device *); 2295 #endif 2296 extern int set_blocksize(struct block_device *, int); 2297 extern int sb_set_blocksize(struct super_block *, int); 2298 extern int sb_min_blocksize(struct super_block *, int); 2299 2300 extern int generic_file_mmap(struct file *, struct vm_area_struct *); 2301 extern int generic_file_readonly_mmap(struct file *, struct vm_area_struct *); 2302 extern int file_read_actor(read_descriptor_t * desc, struct page *page, unsigned long offset, unsigned long size); 2303 int generic_write_checks(struct file *file, loff_t *pos, size_t *count, int isblk); 2304 extern ssize_t generic_file_aio_read(struct kiocb *, const struct iovec *, unsigned long, loff_t); 2305 extern ssize_t __generic_file_aio_write(struct kiocb *, const struct iovec *, unsigned long, 2306 loff_t *); 2307 extern ssize_t generic_file_aio_write(struct kiocb *, const struct iovec *, unsigned long, loff_t); 2308 extern ssize_t generic_file_direct_write(struct kiocb *, const struct iovec *, 2309 unsigned long *, loff_t, loff_t *, size_t, size_t); 2310 extern ssize_t generic_file_buffered_write(struct kiocb *, const struct iovec *, 2311 unsigned long, loff_t, loff_t *, size_t, ssize_t); 2312 extern ssize_t do_sync_read(struct file *filp, char __user *buf, size_t len, loff_t *ppos); 2313 extern ssize_t do_sync_write(struct file *filp, const char __user *buf, size_t len, loff_t *ppos); 2314 extern int generic_segment_checks(const struct iovec *iov, 2315 unsigned long *nr_segs, size_t *count, int access_flags); 2316 2317 /* fs/block_dev.c */ 2318 extern ssize_t blkdev_aio_write(struct kiocb *iocb, const struct iovec *iov, 2319 unsigned long nr_segs, loff_t pos); 2320 extern int blkdev_fsync(struct file *filp, int datasync); 2321 2322 /* fs/splice.c */ 2323 extern ssize_t generic_file_splice_read(struct file *, loff_t *, 2324 struct pipe_inode_info *, size_t, unsigned int); 2325 extern ssize_t default_file_splice_read(struct file *, loff_t *, 2326 struct pipe_inode_info *, size_t, unsigned int); 2327 extern ssize_t generic_file_splice_write(struct pipe_inode_info *, 2328 struct file *, loff_t *, size_t, unsigned int); 2329 extern ssize_t generic_splice_sendpage(struct pipe_inode_info *pipe, 2330 struct file *out, loff_t *, size_t len, unsigned int flags); 2331 extern long do_splice_direct(struct file *in, loff_t *ppos, struct file *out, 2332 size_t len, unsigned int flags); 2333 2334 extern void 2335 file_ra_state_init(struct file_ra_state *ra, struct address_space *mapping); 2336 extern loff_t noop_llseek(struct file *file, loff_t offset, int origin); 2337 extern loff_t no_llseek(struct file *file, loff_t offset, int origin); 2338 extern loff_t generic_file_llseek(struct file *file, loff_t offset, int origin); 2339 extern loff_t generic_file_llseek_unlocked(struct file *file, loff_t offset, 2340 int origin); 2341 extern int generic_file_open(struct inode * inode, struct file * filp); 2342 extern int nonseekable_open(struct inode * inode, struct file * filp); 2343 2344 #ifdef CONFIG_FS_XIP 2345 extern ssize_t xip_file_read(struct file *filp, char __user *buf, size_t len, 2346 loff_t *ppos); 2347 extern int xip_file_mmap(struct file * file, struct vm_area_struct * vma); 2348 extern ssize_t xip_file_write(struct file *filp, const char __user *buf, 2349 size_t len, loff_t *ppos); 2350 extern int xip_truncate_page(struct address_space *mapping, loff_t from); 2351 #else 2352 static inline int xip_truncate_page(struct address_space *mapping, loff_t from) 2353 { 2354 return 0; 2355 } 2356 #endif 2357 2358 #ifdef CONFIG_BLOCK 2359 typedef void (dio_submit_t)(int rw, struct bio *bio, struct inode *inode, 2360 loff_t file_offset); 2361 2362 enum { 2363 /* need locking between buffered and direct access */ 2364 DIO_LOCKING = 0x01, 2365 2366 /* filesystem does not support filling holes */ 2367 DIO_SKIP_HOLES = 0x02, 2368 }; 2369 2370 void dio_end_io(struct bio *bio, int error); 2371 2372 ssize_t __blockdev_direct_IO(int rw, struct kiocb *iocb, struct inode *inode, 2373 struct block_device *bdev, const struct iovec *iov, loff_t offset, 2374 unsigned long nr_segs, get_block_t get_block, dio_iodone_t end_io, 2375 dio_submit_t submit_io, int flags); 2376 2377 static inline ssize_t blockdev_direct_IO(int rw, struct kiocb *iocb, 2378 struct inode *inode, struct block_device *bdev, const struct iovec *iov, 2379 loff_t offset, unsigned long nr_segs, get_block_t get_block, 2380 dio_iodone_t end_io) 2381 { 2382 return __blockdev_direct_IO(rw, iocb, inode, bdev, iov, offset, 2383 nr_segs, get_block, end_io, NULL, 2384 DIO_LOCKING | DIO_SKIP_HOLES); 2385 } 2386 #endif 2387 2388 extern const struct file_operations generic_ro_fops; 2389 2390 #define special_file(m) (S_ISCHR(m)||S_ISBLK(m)||S_ISFIFO(m)||S_ISSOCK(m)) 2391 2392 extern int vfs_readlink(struct dentry *, char __user *, int, const char *); 2393 extern int vfs_follow_link(struct nameidata *, const char *); 2394 extern int page_readlink(struct dentry *, char __user *, int); 2395 extern void *page_follow_link_light(struct dentry *, struct nameidata *); 2396 extern void page_put_link(struct dentry *, struct nameidata *, void *); 2397 extern int __page_symlink(struct inode *inode, const char *symname, int len, 2398 int nofs); 2399 extern int page_symlink(struct inode *inode, const char *symname, int len); 2400 extern const struct inode_operations page_symlink_inode_operations; 2401 extern int generic_readlink(struct dentry *, char __user *, int); 2402 extern void generic_fillattr(struct inode *, struct kstat *); 2403 extern int vfs_getattr(struct vfsmount *, struct dentry *, struct kstat *); 2404 void __inode_add_bytes(struct inode *inode, loff_t bytes); 2405 void inode_add_bytes(struct inode *inode, loff_t bytes); 2406 void inode_sub_bytes(struct inode *inode, loff_t bytes); 2407 loff_t inode_get_bytes(struct inode *inode); 2408 void inode_set_bytes(struct inode *inode, loff_t bytes); 2409 2410 extern int vfs_readdir(struct file *, filldir_t, void *); 2411 2412 extern int vfs_stat(const char __user *, struct kstat *); 2413 extern int vfs_lstat(const char __user *, struct kstat *); 2414 extern int vfs_fstat(unsigned int, struct kstat *); 2415 extern int vfs_fstatat(int , const char __user *, struct kstat *, int); 2416 2417 extern int do_vfs_ioctl(struct file *filp, unsigned int fd, unsigned int cmd, 2418 unsigned long arg); 2419 extern int __generic_block_fiemap(struct inode *inode, 2420 struct fiemap_extent_info *fieinfo, 2421 loff_t start, loff_t len, 2422 get_block_t *get_block); 2423 extern int generic_block_fiemap(struct inode *inode, 2424 struct fiemap_extent_info *fieinfo, u64 start, 2425 u64 len, get_block_t *get_block); 2426 2427 extern void get_filesystem(struct file_system_type *fs); 2428 extern void put_filesystem(struct file_system_type *fs); 2429 extern struct file_system_type *get_fs_type(const char *name); 2430 extern struct super_block *get_super(struct block_device *); 2431 extern struct super_block *get_active_super(struct block_device *bdev); 2432 extern struct super_block *user_get_super(dev_t); 2433 extern void drop_super(struct super_block *sb); 2434 extern void iterate_supers(void (*)(struct super_block *, void *), void *); 2435 2436 extern int dcache_dir_open(struct inode *, struct file *); 2437 extern int dcache_dir_close(struct inode *, struct file *); 2438 extern loff_t dcache_dir_lseek(struct file *, loff_t, int); 2439 extern int dcache_readdir(struct file *, void *, filldir_t); 2440 extern int simple_setattr(struct dentry *, struct iattr *); 2441 extern int simple_getattr(struct vfsmount *, struct dentry *, struct kstat *); 2442 extern int simple_statfs(struct dentry *, struct kstatfs *); 2443 extern int simple_link(struct dentry *, struct inode *, struct dentry *); 2444 extern int simple_unlink(struct inode *, struct dentry *); 2445 extern int simple_rmdir(struct inode *, struct dentry *); 2446 extern int simple_rename(struct inode *, struct dentry *, struct inode *, struct dentry *); 2447 extern int noop_fsync(struct file *, int); 2448 extern int simple_empty(struct dentry *); 2449 extern int simple_readpage(struct file *file, struct page *page); 2450 extern int simple_write_begin(struct file *file, struct address_space *mapping, 2451 loff_t pos, unsigned len, unsigned flags, 2452 struct page **pagep, void **fsdata); 2453 extern int simple_write_end(struct file *file, struct address_space *mapping, 2454 loff_t pos, unsigned len, unsigned copied, 2455 struct page *page, void *fsdata); 2456 2457 extern struct dentry *simple_lookup(struct inode *, struct dentry *, struct nameidata *); 2458 extern ssize_t generic_read_dir(struct file *, char __user *, size_t, loff_t *); 2459 extern const struct file_operations simple_dir_operations; 2460 extern const struct inode_operations simple_dir_inode_operations; 2461 struct tree_descr { char *name; const struct file_operations *ops; int mode; }; 2462 struct dentry *d_alloc_name(struct dentry *, const char *); 2463 extern int simple_fill_super(struct super_block *, unsigned long, struct tree_descr *); 2464 extern int simple_pin_fs(struct file_system_type *, struct vfsmount **mount, int *count); 2465 extern void simple_release_fs(struct vfsmount **mount, int *count); 2466 2467 extern ssize_t simple_read_from_buffer(void __user *to, size_t count, 2468 loff_t *ppos, const void *from, size_t available); 2469 extern ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos, 2470 const void __user *from, size_t count); 2471 2472 extern int generic_file_fsync(struct file *, int); 2473 2474 extern int generic_check_addressable(unsigned, u64); 2475 2476 #ifdef CONFIG_MIGRATION 2477 extern int buffer_migrate_page(struct address_space *, 2478 struct page *, struct page *); 2479 #else 2480 #define buffer_migrate_page NULL 2481 #endif 2482 2483 extern int inode_change_ok(const struct inode *, struct iattr *); 2484 extern int inode_newsize_ok(const struct inode *, loff_t offset); 2485 extern void setattr_copy(struct inode *inode, const struct iattr *attr); 2486 2487 extern void file_update_time(struct file *file); 2488 2489 extern int generic_show_options(struct seq_file *m, struct vfsmount *mnt); 2490 extern void save_mount_options(struct super_block *sb, char *options); 2491 extern void replace_mount_options(struct super_block *sb, char *options); 2492 2493 static inline ino_t parent_ino(struct dentry *dentry) 2494 { 2495 ino_t res; 2496 2497 /* 2498 * Don't strictly need d_lock here? If the parent ino could change 2499 * then surely we'd have a deeper race in the caller? 2500 */ 2501 spin_lock(&dentry->d_lock); 2502 res = dentry->d_parent->d_inode->i_ino; 2503 spin_unlock(&dentry->d_lock); 2504 return res; 2505 } 2506 2507 /* Transaction based IO helpers */ 2508 2509 /* 2510 * An argresp is stored in an allocated page and holds the 2511 * size of the argument or response, along with its content 2512 */ 2513 struct simple_transaction_argresp { 2514 ssize_t size; 2515 char data[0]; 2516 }; 2517 2518 #define SIMPLE_TRANSACTION_LIMIT (PAGE_SIZE - sizeof(struct simple_transaction_argresp)) 2519 2520 char *simple_transaction_get(struct file *file, const char __user *buf, 2521 size_t size); 2522 ssize_t simple_transaction_read(struct file *file, char __user *buf, 2523 size_t size, loff_t *pos); 2524 int simple_transaction_release(struct inode *inode, struct file *file); 2525 2526 void simple_transaction_set(struct file *file, size_t n); 2527 2528 /* 2529 * simple attribute files 2530 * 2531 * These attributes behave similar to those in sysfs: 2532 * 2533 * Writing to an attribute immediately sets a value, an open file can be 2534 * written to multiple times. 2535 * 2536 * Reading from an attribute creates a buffer from the value that might get 2537 * read with multiple read calls. When the attribute has been read 2538 * completely, no further read calls are possible until the file is opened 2539 * again. 2540 * 2541 * All attributes contain a text representation of a numeric value 2542 * that are accessed with the get() and set() functions. 2543 */ 2544 #define DEFINE_SIMPLE_ATTRIBUTE(__fops, __get, __set, __fmt) \ 2545 static int __fops ## _open(struct inode *inode, struct file *file) \ 2546 { \ 2547 __simple_attr_check_format(__fmt, 0ull); \ 2548 return simple_attr_open(inode, file, __get, __set, __fmt); \ 2549 } \ 2550 static const struct file_operations __fops = { \ 2551 .owner = THIS_MODULE, \ 2552 .open = __fops ## _open, \ 2553 .release = simple_attr_release, \ 2554 .read = simple_attr_read, \ 2555 .write = simple_attr_write, \ 2556 .llseek = generic_file_llseek, \ 2557 }; 2558 2559 static inline void __attribute__((format(printf, 1, 2))) 2560 __simple_attr_check_format(const char *fmt, ...) 2561 { 2562 /* don't do anything, just let the compiler check the arguments; */ 2563 } 2564 2565 int simple_attr_open(struct inode *inode, struct file *file, 2566 int (*get)(void *, u64 *), int (*set)(void *, u64), 2567 const char *fmt); 2568 int simple_attr_release(struct inode *inode, struct file *file); 2569 ssize_t simple_attr_read(struct file *file, char __user *buf, 2570 size_t len, loff_t *ppos); 2571 ssize_t simple_attr_write(struct file *file, const char __user *buf, 2572 size_t len, loff_t *ppos); 2573 2574 struct ctl_table; 2575 int proc_nr_files(struct ctl_table *table, int write, 2576 void __user *buffer, size_t *lenp, loff_t *ppos); 2577 int proc_nr_dentry(struct ctl_table *table, int write, 2578 void __user *buffer, size_t *lenp, loff_t *ppos); 2579 int proc_nr_inodes(struct ctl_table *table, int write, 2580 void __user *buffer, size_t *lenp, loff_t *ppos); 2581 int __init get_filesystem_list(char *buf); 2582 2583 #define __FMODE_EXEC ((__force int) FMODE_EXEC) 2584 #define __FMODE_NONOTIFY ((__force int) FMODE_NONOTIFY) 2585 2586 #define ACC_MODE(x) ("\004\002\006\006"[(x)&O_ACCMODE]) 2587 #define OPEN_FMODE(flag) ((__force fmode_t)(((flag + 1) & O_ACCMODE) | \ 2588 (flag & __FMODE_NONOTIFY))) 2589 2590 static inline int is_sxid(mode_t mode) 2591 { 2592 return (mode & S_ISUID) || ((mode & S_ISGID) && (mode & S_IXGRP)); 2593 } 2594 2595 static inline void inode_has_no_xattr(struct inode *inode) 2596 { 2597 if (!is_sxid(inode->i_mode) && (inode->i_sb->s_flags & MS_NOSEC)) 2598 inode->i_flags |= S_NOSEC; 2599 } 2600 2601 #endif /* __KERNEL__ */ 2602 #endif /* _LINUX_FS_H */