1 // SPDX-License-Identifier: GPL-2.0 2 /* 3 * linux/fs/nfs/file.c 4 * 5 * Copyright (C) 1992 Rick Sladkey 6 */ 7 #include <linux/fs.h> 8 #include <linux/file.h> 9 #include <linux/falloc.h> 10 #include <linux/mount.h> 11 #include <linux/nfs_fs.h> 12 #include "delegation.h" 13 #include "internal.h" 14 #include "iostat.h" 15 #include "fscache.h" 16 #include "pnfs.h" 17 18 #include "nfstrace.h" 19 20 #ifdef CONFIG_NFS_V4_2 21 #include "nfs42.h" 22 #endif 23 24 #define NFSDBG_FACILITY NFSDBG_FILE 25 26 static int 27 nfs4_file_open(struct inode *inode, struct file *filp) 28 { 29 struct nfs_open_context *ctx; 30 struct dentry *dentry = file_dentry(filp); 31 struct dentry *parent = NULL; 32 struct inode *dir; 33 unsigned openflags = filp->f_flags; 34 struct iattr attr; 35 int err; 36 37 /* 38 * If no cached dentry exists or if it's negative, NFSv4 handled the 39 * opens in ->lookup() or ->create(). 40 * 41 * We only get this far for a cached positive dentry. We skipped 42 * revalidation, so handle it here by dropping the dentry and returning 43 * -EOPENSTALE. The VFS will retry the lookup/create/open. 44 */ 45 46 dprintk("NFS: open file(%pd2)\n", dentry); 47 48 err = nfs_check_flags(openflags); 49 if (err) 50 return err; 51 52 if ((openflags & O_ACCMODE) == 3) 53 return nfs_open(inode, filp); 54 55 /* We can't create new files here */ 56 openflags &= ~(O_CREAT|O_EXCL); 57 58 parent = dget_parent(dentry); 59 dir = d_inode(parent); 60 61 ctx = alloc_nfs_open_context(file_dentry(filp), filp->f_mode, filp); 62 err = PTR_ERR(ctx); 63 if (IS_ERR(ctx)) 64 goto out; 65 66 attr.ia_valid = ATTR_OPEN; 67 if (openflags & O_TRUNC) { 68 attr.ia_valid |= ATTR_SIZE; 69 attr.ia_size = 0; 70 filemap_write_and_wait(inode->i_mapping); 71 } 72 73 inode = NFS_PROTO(dir)->open_context(dir, ctx, openflags, &attr, NULL); 74 if (IS_ERR(inode)) { 75 err = PTR_ERR(inode); 76 switch (err) { 77 default: 78 goto out_put_ctx; 79 case -ENOENT: 80 case -ESTALE: 81 case -EISDIR: 82 case -ENOTDIR: 83 case -ELOOP: 84 goto out_drop; 85 } 86 } 87 if (inode != d_inode(dentry)) 88 goto out_drop; 89 90 nfs_set_verifier(dentry, nfs_save_change_attribute(dir)); 91 nfs_file_set_open_context(filp, ctx); 92 nfs_fscache_open_file(inode, filp); 93 err = 0; 94 95 out_put_ctx: 96 put_nfs_open_context(ctx); 97 out: 98 dput(parent); 99 return err; 100 101 out_drop: 102 d_drop(dentry); 103 err = -EOPENSTALE; 104 goto out_put_ctx; 105 } 106 107 /* 108 * Flush all dirty pages, and check for write errors. 109 */ 110 static int 111 nfs4_file_flush(struct file *file, fl_owner_t id) 112 { 113 struct inode *inode = file_inode(file); 114 115 dprintk("NFS: flush(%pD2)\n", file); 116 117 nfs_inc_stats(inode, NFSIOS_VFSFLUSH); 118 if ((file->f_mode & FMODE_WRITE) == 0) 119 return 0; 120 121 /* 122 * If we're holding a write delegation, then check if we're required 123 * to flush the i/o on close. If not, then just start the i/o now. 124 */ 125 if (!nfs4_delegation_flush_on_close(inode)) 126 return filemap_fdatawrite(file->f_mapping); 127 128 /* Flush writes to the server and return any errors */ 129 return nfs_wb_all(inode); 130 } 131 132 #ifdef CONFIG_NFS_V4_2 133 static ssize_t __nfs4_copy_file_range(struct file *file_in, loff_t pos_in, 134 struct file *file_out, loff_t pos_out, 135 size_t count, unsigned int flags) 136 { 137 struct nfs42_copy_notify_res *cn_resp = NULL; 138 struct nl4_server *nss = NULL; 139 nfs4_stateid *cnrs = NULL; 140 ssize_t ret; 141 bool sync = false; 142 143 /* Only offload copy if superblock is the same */ 144 if (file_in->f_op != &nfs4_file_operations) 145 return -EXDEV; 146 if (!nfs_server_capable(file_inode(file_out), NFS_CAP_COPY)) 147 return -EOPNOTSUPP; 148 if (file_inode(file_in) == file_inode(file_out)) 149 return -EOPNOTSUPP; 150 /* if the copy size if smaller than 2 RPC payloads, make it 151 * synchronous 152 */ 153 if (count <= 2 * NFS_SERVER(file_inode(file_in))->rsize) 154 sync = true; 155 retry: 156 if (!nfs42_files_from_same_server(file_in, file_out)) { 157 /* for inter copy, if copy size if smaller than 12 RPC 158 * payloads, fallback to traditional copy. There are 159 * 14 RPCs during an NFSv4.x mount between source/dest 160 * servers. 161 */ 162 if (sync || 163 count <= 14 * NFS_SERVER(file_inode(file_in))->rsize) 164 return -EOPNOTSUPP; 165 cn_resp = kzalloc(sizeof(struct nfs42_copy_notify_res), 166 GFP_NOFS); 167 if (unlikely(cn_resp == NULL)) 168 return -ENOMEM; 169 170 ret = nfs42_proc_copy_notify(file_in, file_out, cn_resp); 171 if (ret) { 172 ret = -EOPNOTSUPP; 173 goto out; 174 } 175 nss = &cn_resp->cnr_src; 176 cnrs = &cn_resp->cnr_stateid; 177 } 178 ret = nfs42_proc_copy(file_in, pos_in, file_out, pos_out, count, 179 nss, cnrs, sync); 180 out: 181 if (!nfs42_files_from_same_server(file_in, file_out)) 182 kfree(cn_resp); 183 if (ret == -EAGAIN) 184 goto retry; 185 return ret; 186 } 187 188 static ssize_t nfs4_copy_file_range(struct file *file_in, loff_t pos_in, 189 struct file *file_out, loff_t pos_out, 190 size_t count, unsigned int flags) 191 { 192 ssize_t ret; 193 194 ret = __nfs4_copy_file_range(file_in, pos_in, file_out, pos_out, count, 195 flags); 196 if (ret == -EOPNOTSUPP || ret == -EXDEV) 197 ret = generic_copy_file_range(file_in, pos_in, file_out, 198 pos_out, count, flags); 199 return ret; 200 } 201 202 static loff_t nfs4_file_llseek(struct file *filep, loff_t offset, int whence) 203 { 204 loff_t ret; 205 206 switch (whence) { 207 case SEEK_HOLE: 208 case SEEK_DATA: 209 ret = nfs42_proc_llseek(filep, offset, whence); 210 if (ret != -ENOTSUPP) 211 return ret; 212 /* Fall through */ 213 default: 214 return nfs_file_llseek(filep, offset, whence); 215 } 216 } 217 218 static long nfs42_fallocate(struct file *filep, int mode, loff_t offset, loff_t len) 219 { 220 struct inode *inode = file_inode(filep); 221 long ret; 222 223 if (!S_ISREG(inode->i_mode)) 224 return -EOPNOTSUPP; 225 226 if ((mode != 0) && (mode != (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE))) 227 return -EOPNOTSUPP; 228 229 ret = inode_newsize_ok(inode, offset + len); 230 if (ret < 0) 231 return ret; 232 233 if (mode & FALLOC_FL_PUNCH_HOLE) 234 return nfs42_proc_deallocate(filep, offset, len); 235 return nfs42_proc_allocate(filep, offset, len); 236 } 237 238 static loff_t nfs42_remap_file_range(struct file *src_file, loff_t src_off, 239 struct file *dst_file, loff_t dst_off, loff_t count, 240 unsigned int remap_flags) 241 { 242 struct inode *dst_inode = file_inode(dst_file); 243 struct nfs_server *server = NFS_SERVER(dst_inode); 244 struct inode *src_inode = file_inode(src_file); 245 unsigned int bs = server->clone_blksize; 246 bool same_inode = false; 247 int ret; 248 249 /* NFS does not support deduplication. */ 250 if (remap_flags & REMAP_FILE_DEDUP) 251 return -EOPNOTSUPP; 252 253 if (remap_flags & ~REMAP_FILE_ADVISORY) 254 return -EINVAL; 255 256 /* check alignment w.r.t. clone_blksize */ 257 ret = -EINVAL; 258 if (bs) { 259 if (!IS_ALIGNED(src_off, bs) || !IS_ALIGNED(dst_off, bs)) 260 goto out; 261 if (!IS_ALIGNED(count, bs) && i_size_read(src_inode) != (src_off + count)) 262 goto out; 263 } 264 265 if (src_inode == dst_inode) 266 same_inode = true; 267 268 /* XXX: do we lock at all? what if server needs CB_RECALL_LAYOUT? */ 269 if (same_inode) { 270 inode_lock(src_inode); 271 } else if (dst_inode < src_inode) { 272 inode_lock_nested(dst_inode, I_MUTEX_PARENT); 273 inode_lock_nested(src_inode, I_MUTEX_CHILD); 274 } else { 275 inode_lock_nested(src_inode, I_MUTEX_PARENT); 276 inode_lock_nested(dst_inode, I_MUTEX_CHILD); 277 } 278 279 /* flush all pending writes on both src and dst so that server 280 * has the latest data */ 281 ret = nfs_sync_inode(src_inode); 282 if (ret) 283 goto out_unlock; 284 ret = nfs_sync_inode(dst_inode); 285 if (ret) 286 goto out_unlock; 287 288 ret = nfs42_proc_clone(src_file, dst_file, src_off, dst_off, count); 289 290 /* truncate inode page cache of the dst range so that future reads can fetch 291 * new data from server */ 292 if (!ret) 293 truncate_inode_pages_range(&dst_inode->i_data, dst_off, dst_off + count - 1); 294 295 out_unlock: 296 if (same_inode) { 297 inode_unlock(src_inode); 298 } else if (dst_inode < src_inode) { 299 inode_unlock(src_inode); 300 inode_unlock(dst_inode); 301 } else { 302 inode_unlock(dst_inode); 303 inode_unlock(src_inode); 304 } 305 out: 306 return ret < 0 ? ret : count; 307 } 308 309 static int read_name_gen = 1; 310 #define SSC_READ_NAME_BODY "ssc_read_%d" 311 312 struct file * 313 nfs42_ssc_open(struct vfsmount *ss_mnt, struct nfs_fh *src_fh, 314 nfs4_stateid *stateid) 315 { 316 struct nfs_fattr fattr; 317 struct file *filep, *res; 318 struct nfs_server *server; 319 struct inode *r_ino = NULL; 320 struct nfs_open_context *ctx; 321 struct nfs4_state_owner *sp; 322 char *read_name = NULL; 323 int len, status = 0; 324 325 server = NFS_SERVER(ss_mnt->mnt_root->d_inode); 326 327 nfs_fattr_init(&fattr); 328 329 status = nfs4_proc_getattr(server, src_fh, &fattr, NULL, NULL); 330 if (status < 0) { 331 res = ERR_PTR(status); 332 goto out; 333 } 334 335 res = ERR_PTR(-ENOMEM); 336 len = strlen(SSC_READ_NAME_BODY) + 16; 337 read_name = kzalloc(len, GFP_NOFS); 338 if (read_name == NULL) 339 goto out; 340 snprintf(read_name, len, SSC_READ_NAME_BODY, read_name_gen++); 341 342 r_ino = nfs_fhget(ss_mnt->mnt_root->d_inode->i_sb, src_fh, &fattr, 343 NULL); 344 if (IS_ERR(r_ino)) { 345 res = ERR_CAST(r_ino); 346 goto out_free_name; 347 } 348 349 filep = alloc_file_pseudo(r_ino, ss_mnt, read_name, FMODE_READ, 350 r_ino->i_fop); 351 if (IS_ERR(filep)) { 352 res = ERR_CAST(filep); 353 goto out_free_name; 354 } 355 filep->f_mode |= FMODE_READ; 356 357 ctx = alloc_nfs_open_context(filep->f_path.dentry, filep->f_mode, 358 filep); 359 if (IS_ERR(ctx)) { 360 res = ERR_CAST(ctx); 361 goto out_filep; 362 } 363 364 res = ERR_PTR(-EINVAL); 365 sp = nfs4_get_state_owner(server, ctx->cred, GFP_KERNEL); 366 if (sp == NULL) 367 goto out_ctx; 368 369 ctx->state = nfs4_get_open_state(r_ino, sp); 370 if (ctx->state == NULL) 371 goto out_stateowner; 372 373 set_bit(NFS_SRV_SSC_COPY_STATE, &ctx->state->flags); 374 set_bit(NFS_OPEN_STATE, &ctx->state->flags); 375 memcpy(&ctx->state->open_stateid.other, &stateid->other, 376 NFS4_STATEID_OTHER_SIZE); 377 update_open_stateid(ctx->state, stateid, NULL, filep->f_mode); 378 379 nfs_file_set_open_context(filep, ctx); 380 put_nfs_open_context(ctx); 381 382 file_ra_state_init(&filep->f_ra, filep->f_mapping->host->i_mapping); 383 res = filep; 384 out_free_name: 385 kfree(read_name); 386 out: 387 return res; 388 out_stateowner: 389 nfs4_put_state_owner(sp); 390 out_ctx: 391 put_nfs_open_context(ctx); 392 out_filep: 393 fput(filep); 394 goto out_free_name; 395 } 396 EXPORT_SYMBOL_GPL(nfs42_ssc_open); 397 void nfs42_ssc_close(struct file *filep) 398 { 399 struct nfs_open_context *ctx = nfs_file_open_context(filep); 400 401 ctx->state->flags = 0; 402 } 403 EXPORT_SYMBOL_GPL(nfs42_ssc_close); 404 #endif /* CONFIG_NFS_V4_2 */ 405 406 const struct file_operations nfs4_file_operations = { 407 .read_iter = nfs_file_read, 408 .write_iter = nfs_file_write, 409 .mmap = nfs_file_mmap, 410 .open = nfs4_file_open, 411 .flush = nfs4_file_flush, 412 .release = nfs_file_release, 413 .fsync = nfs_file_fsync, 414 .lock = nfs_lock, 415 .flock = nfs_flock, 416 .splice_read = generic_file_splice_read, 417 .splice_write = iter_file_splice_write, 418 .check_flags = nfs_check_flags, 419 .setlease = simple_nosetlease, 420 #ifdef CONFIG_NFS_V4_2 421 .copy_file_range = nfs4_copy_file_range, 422 .llseek = nfs4_file_llseek, 423 .fallocate = nfs42_fallocate, 424 .remap_file_range = nfs42_remap_file_range, 425 #else 426 .llseek = nfs_file_llseek, 427 #endif 428 }; 429