1 // SPDX-License-Identifier: GPL-2.0-only
3 * This file contians vfs file ops for 9P2000.
5 * Copyright (C) 2004 by Eric Van Hensbergen <ericvh@gmail.com>
6 * Copyright (C) 2002 by Ron Minnich <rminnich@lanl.gov>
9 #include <linux/module.h>
10 #include <linux/errno.h>
12 #include <linux/sched.h>
13 #include <linux/file.h>
14 #include <linux/stat.h>
15 #include <linux/string.h>
16 #include <linux/inet.h>
17 #include <linux/list.h>
18 #include <linux/pagemap.h>
19 #include <linux/utsname.h>
20 #include <linux/uaccess.h>
21 #include <linux/idr.h>
22 #include <linux/uio.h>
23 #include <linux/slab.h>
24 #include <net/9p/9p.h>
25 #include <net/9p/client.h>
32 static const struct vm_operations_struct v9fs_file_vm_ops;
33 static const struct vm_operations_struct v9fs_mmap_file_vm_ops;
36 * v9fs_file_open - open a file (or directory)
37 * @inode: inode to be opened
38 * @file: file being opened
42 int v9fs_file_open(struct inode *inode, struct file *file)
45 struct v9fs_inode *v9inode;
46 struct v9fs_session_info *v9ses;
47 struct p9_fid *fid, *writeback_fid;
50 p9_debug(P9_DEBUG_VFS, "inode: %p file: %p\n", inode, file);
51 v9inode = V9FS_I(inode);
52 v9ses = v9fs_inode2v9ses(inode);
53 if (v9fs_proto_dotl(v9ses))
54 omode = v9fs_open_to_dotl_flags(file->f_flags);
56 omode = v9fs_uflags2omode(file->f_flags,
57 v9fs_proto_dotu(v9ses));
58 fid = file->private_data;
60 fid = v9fs_fid_clone(file_dentry(file));
64 err = p9_client_open(fid, omode);
69 if ((file->f_flags & O_APPEND) &&
70 (!v9fs_proto_dotu(v9ses) && !v9fs_proto_dotl(v9ses)))
71 generic_file_llseek(file, 0, SEEK_END);
74 file->private_data = fid;
75 mutex_lock(&v9inode->v_mutex);
76 if ((v9ses->cache == CACHE_LOOSE || v9ses->cache == CACHE_FSCACHE) &&
77 !v9inode->writeback_fid &&
78 ((file->f_flags & O_ACCMODE) != O_RDONLY)) {
80 * clone a fid and add it to writeback_fid
81 * we do it during open time instead of
82 * page dirty time via write_begin/page_mkwrite
83 * because we want write after unlink usecase
86 writeback_fid = v9fs_writeback_fid(file_dentry(file));
87 if (IS_ERR(writeback_fid)) {
88 err = PTR_ERR(writeback_fid);
89 mutex_unlock(&v9inode->v_mutex);
92 v9inode->writeback_fid = (void *) writeback_fid;
94 mutex_unlock(&v9inode->v_mutex);
95 if (v9ses->cache == CACHE_LOOSE || v9ses->cache == CACHE_FSCACHE)
96 fscache_use_cookie(v9fs_inode_cookie(v9inode),
97 file->f_mode & FMODE_WRITE);
98 v9fs_open_fid_add(inode, fid);
101 p9_client_clunk(file->private_data);
102 file->private_data = NULL;
107 * v9fs_file_lock - lock a file (or directory)
108 * @filp: file to be locked
110 * @fl: file lock structure
112 * Bugs: this looks like a local only lock, we should extend into 9P
113 * by using open exclusive
116 static int v9fs_file_lock(struct file *filp, int cmd, struct file_lock *fl)
118 struct inode *inode = file_inode(filp);
120 p9_debug(P9_DEBUG_VFS, "filp: %p lock: %p\n", filp, fl);
122 if ((IS_SETLK(cmd) || IS_SETLKW(cmd)) && fl->fl_type != F_UNLCK) {
123 filemap_write_and_wait(inode->i_mapping);
124 invalidate_mapping_pages(&inode->i_data, 0, -1);
130 static int v9fs_file_do_lock(struct file *filp, int cmd, struct file_lock *fl)
132 struct p9_flock flock;
134 uint8_t status = P9_LOCK_ERROR;
136 unsigned char fl_type;
137 struct v9fs_session_info *v9ses;
139 fid = filp->private_data;
142 BUG_ON((fl->fl_flags & FL_POSIX) != FL_POSIX);
144 res = locks_lock_file_wait(filp, fl);
148 /* convert posix lock to p9 tlock args */
149 memset(&flock, 0, sizeof(flock));
150 /* map the lock type */
151 switch (fl->fl_type) {
153 flock.type = P9_LOCK_TYPE_RDLCK;
156 flock.type = P9_LOCK_TYPE_WRLCK;
159 flock.type = P9_LOCK_TYPE_UNLCK;
162 flock.start = fl->fl_start;
163 if (fl->fl_end == OFFSET_MAX)
166 flock.length = fl->fl_end - fl->fl_start + 1;
167 flock.proc_id = fl->fl_pid;
168 flock.client_id = fid->clnt->name;
170 flock.flags = P9_LOCK_FLAGS_BLOCK;
172 v9ses = v9fs_inode2v9ses(file_inode(filp));
175 * if its a blocked request and we get P9_LOCK_BLOCKED as the status
176 * for lock request, keep on trying
179 res = p9_client_lock_dotl(fid, &flock, &status);
183 if (status != P9_LOCK_BLOCKED)
185 if (status == P9_LOCK_BLOCKED && !IS_SETLKW(cmd))
187 if (schedule_timeout_interruptible(v9ses->session_lock_timeout)
191 * p9_client_lock_dotl overwrites flock.client_id with the
192 * server message, free and reuse the client name
194 if (flock.client_id != fid->clnt->name) {
195 kfree(flock.client_id);
196 flock.client_id = fid->clnt->name;
200 /* map 9p status to VFS status */
202 case P9_LOCK_SUCCESS:
205 case P9_LOCK_BLOCKED:
209 WARN_ONCE(1, "unknown lock status code: %d\n", status);
219 * incase server returned error for lock request, revert
222 if (res < 0 && fl->fl_type != F_UNLCK) {
223 fl_type = fl->fl_type;
224 fl->fl_type = F_UNLCK;
225 /* Even if this fails we want to return the remote error */
226 locks_lock_file_wait(filp, fl);
227 fl->fl_type = fl_type;
229 if (flock.client_id != fid->clnt->name)
230 kfree(flock.client_id);
235 static int v9fs_file_getlock(struct file *filp, struct file_lock *fl)
237 struct p9_getlock glock;
241 fid = filp->private_data;
244 posix_test_lock(filp, fl);
246 * if we have a conflicting lock locally, no need to validate
249 if (fl->fl_type != F_UNLCK)
252 /* convert posix lock to p9 tgetlock args */
253 memset(&glock, 0, sizeof(glock));
254 glock.type = P9_LOCK_TYPE_UNLCK;
255 glock.start = fl->fl_start;
256 if (fl->fl_end == OFFSET_MAX)
259 glock.length = fl->fl_end - fl->fl_start + 1;
260 glock.proc_id = fl->fl_pid;
261 glock.client_id = fid->clnt->name;
263 res = p9_client_getlock_dotl(fid, &glock);
266 /* map 9p lock type to os lock type */
267 switch (glock.type) {
268 case P9_LOCK_TYPE_RDLCK:
269 fl->fl_type = F_RDLCK;
271 case P9_LOCK_TYPE_WRLCK:
272 fl->fl_type = F_WRLCK;
274 case P9_LOCK_TYPE_UNLCK:
275 fl->fl_type = F_UNLCK;
278 if (glock.type != P9_LOCK_TYPE_UNLCK) {
279 fl->fl_start = glock.start;
280 if (glock.length == 0)
281 fl->fl_end = OFFSET_MAX;
283 fl->fl_end = glock.start + glock.length - 1;
284 fl->fl_pid = -glock.proc_id;
287 if (glock.client_id != fid->clnt->name)
288 kfree(glock.client_id);
293 * v9fs_file_lock_dotl - lock a file (or directory)
294 * @filp: file to be locked
296 * @fl: file lock structure
300 static int v9fs_file_lock_dotl(struct file *filp, int cmd, struct file_lock *fl)
302 struct inode *inode = file_inode(filp);
305 p9_debug(P9_DEBUG_VFS, "filp: %p cmd:%d lock: %p name: %pD\n",
306 filp, cmd, fl, filp);
308 if ((IS_SETLK(cmd) || IS_SETLKW(cmd)) && fl->fl_type != F_UNLCK) {
309 filemap_write_and_wait(inode->i_mapping);
310 invalidate_mapping_pages(&inode->i_data, 0, -1);
313 if (IS_SETLK(cmd) || IS_SETLKW(cmd))
314 ret = v9fs_file_do_lock(filp, cmd, fl);
315 else if (IS_GETLK(cmd))
316 ret = v9fs_file_getlock(filp, fl);
323 * v9fs_file_flock_dotl - lock a file
324 * @filp: file to be locked
326 * @fl: file lock structure
330 static int v9fs_file_flock_dotl(struct file *filp, int cmd,
331 struct file_lock *fl)
333 struct inode *inode = file_inode(filp);
336 p9_debug(P9_DEBUG_VFS, "filp: %p cmd:%d lock: %p name: %pD\n",
337 filp, cmd, fl, filp);
339 if (!(fl->fl_flags & FL_FLOCK))
342 if ((IS_SETLK(cmd) || IS_SETLKW(cmd)) && fl->fl_type != F_UNLCK) {
343 filemap_write_and_wait(inode->i_mapping);
344 invalidate_mapping_pages(&inode->i_data, 0, -1);
346 /* Convert flock to posix lock */
347 fl->fl_flags |= FL_POSIX;
348 fl->fl_flags ^= FL_FLOCK;
350 if (IS_SETLK(cmd) | IS_SETLKW(cmd))
351 ret = v9fs_file_do_lock(filp, cmd, fl);
359 * v9fs_file_read_iter - read from a file
360 * @iocb: The operation parameters
361 * @to: The buffer to read into
365 v9fs_file_read_iter(struct kiocb *iocb, struct iov_iter *to)
367 struct p9_fid *fid = iocb->ki_filp->private_data;
370 p9_debug(P9_DEBUG_VFS, "count %zu offset %lld\n",
371 iov_iter_count(to), iocb->ki_pos);
373 if (iocb->ki_filp->f_flags & O_NONBLOCK)
374 ret = p9_client_read_once(fid, iocb->ki_pos, to, &err);
376 ret = p9_client_read(fid, iocb->ki_pos, to, &err);
385 * v9fs_file_write_iter - write to a file
386 * @iocb: The operation parameters
387 * @from: The data to write
391 v9fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
393 struct file *file = iocb->ki_filp;
398 retval = generic_write_checks(iocb, from);
402 origin = iocb->ki_pos;
403 retval = p9_client_write(file->private_data, iocb->ki_pos, from, &err);
405 struct inode *inode = file_inode(file);
407 unsigned long pg_start, pg_end;
409 pg_start = origin >> PAGE_SHIFT;
410 pg_end = (origin + retval - 1) >> PAGE_SHIFT;
411 if (inode->i_mapping && inode->i_mapping->nrpages)
412 invalidate_inode_pages2_range(inode->i_mapping,
414 iocb->ki_pos += retval;
415 i_size = i_size_read(inode);
416 if (iocb->ki_pos > i_size) {
417 inode_add_bytes(inode, iocb->ki_pos - i_size);
419 * Need to serialize against i_size_write() in
422 v9fs_i_size_write(inode, iocb->ki_pos);
429 static int v9fs_file_fsync(struct file *filp, loff_t start, loff_t end,
433 struct inode *inode = filp->f_mapping->host;
434 struct p9_wstat wstat;
437 retval = file_write_and_wait_range(filp, start, end);
442 p9_debug(P9_DEBUG_VFS, "filp %p datasync %x\n", filp, datasync);
444 fid = filp->private_data;
445 v9fs_blank_wstat(&wstat);
447 retval = p9_client_wstat(fid, &wstat);
453 int v9fs_file_fsync_dotl(struct file *filp, loff_t start, loff_t end,
457 struct inode *inode = filp->f_mapping->host;
460 retval = file_write_and_wait_range(filp, start, end);
465 p9_debug(P9_DEBUG_VFS, "filp %p datasync %x\n", filp, datasync);
467 fid = filp->private_data;
469 retval = p9_client_fsync(fid, datasync);
476 v9fs_file_mmap(struct file *filp, struct vm_area_struct *vma)
481 retval = generic_file_mmap(filp, vma);
483 vma->vm_ops = &v9fs_file_vm_ops;
489 v9fs_mmap_file_mmap(struct file *filp, struct vm_area_struct *vma)
493 struct v9fs_inode *v9inode;
496 inode = file_inode(filp);
497 v9inode = V9FS_I(inode);
498 mutex_lock(&v9inode->v_mutex);
499 if (!v9inode->writeback_fid &&
500 (vma->vm_flags & VM_SHARED) &&
501 (vma->vm_flags & VM_WRITE)) {
503 * clone a fid and add it to writeback_fid
504 * we do it during mmap instead of
505 * page dirty time via write_begin/page_mkwrite
506 * because we want write after unlink usecase
509 fid = v9fs_writeback_fid(file_dentry(filp));
511 retval = PTR_ERR(fid);
512 mutex_unlock(&v9inode->v_mutex);
515 v9inode->writeback_fid = (void *) fid;
517 mutex_unlock(&v9inode->v_mutex);
519 retval = generic_file_mmap(filp, vma);
521 vma->vm_ops = &v9fs_mmap_file_vm_ops;
527 v9fs_vm_page_mkwrite(struct vm_fault *vmf)
529 struct v9fs_inode *v9inode;
530 struct folio *folio = page_folio(vmf->page);
531 struct file *filp = vmf->vma->vm_file;
532 struct inode *inode = file_inode(filp);
535 p9_debug(P9_DEBUG_VFS, "folio %p fid %lx\n",
536 folio, (unsigned long)filp->private_data);
538 v9inode = V9FS_I(inode);
540 /* Wait for the page to be written to the cache before we allow it to
541 * be modified. We then assume the entire page will need writing back.
543 #ifdef CONFIG_9P_FSCACHE
544 if (folio_test_fscache(folio) &&
545 folio_wait_fscache_killable(folio) < 0)
546 return VM_FAULT_NOPAGE;
549 /* Update file times before taking page lock */
550 file_update_time(filp);
552 BUG_ON(!v9inode->writeback_fid);
553 if (folio_lock_killable(folio) < 0)
554 return VM_FAULT_RETRY;
555 if (folio_mapping(folio) != inode->i_mapping)
557 folio_wait_stable(folio);
559 return VM_FAULT_LOCKED;
562 return VM_FAULT_NOPAGE;
566 * v9fs_mmap_file_read_iter - read from a file
567 * @iocb: The operation parameters
568 * @to: The buffer to read into
572 v9fs_mmap_file_read_iter(struct kiocb *iocb, struct iov_iter *to)
574 /* TODO: Check if there are dirty pages */
575 return v9fs_file_read_iter(iocb, to);
579 * v9fs_mmap_file_write_iter - write to a file
580 * @iocb: The operation parameters
581 * @from: The data to write
585 v9fs_mmap_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
588 * TODO: invalidate mmaps on filp's inode between
589 * offset and offset+count
591 return v9fs_file_write_iter(iocb, from);
594 static void v9fs_mmap_vm_close(struct vm_area_struct *vma)
598 struct writeback_control wbc = {
599 .nr_to_write = LONG_MAX,
600 .sync_mode = WB_SYNC_ALL,
601 .range_start = (loff_t)vma->vm_pgoff * PAGE_SIZE,
602 /* absolute end, byte at end included */
603 .range_end = (loff_t)vma->vm_pgoff * PAGE_SIZE +
604 (vma->vm_end - vma->vm_start - 1),
607 if (!(vma->vm_flags & VM_SHARED))
610 p9_debug(P9_DEBUG_VFS, "9p VMA close, %p, flushing", vma);
612 inode = file_inode(vma->vm_file);
613 filemap_fdatawrite_wbc(inode->i_mapping, &wbc);
617 static const struct vm_operations_struct v9fs_file_vm_ops = {
618 .fault = filemap_fault,
619 .map_pages = filemap_map_pages,
620 .page_mkwrite = v9fs_vm_page_mkwrite,
623 static const struct vm_operations_struct v9fs_mmap_file_vm_ops = {
624 .close = v9fs_mmap_vm_close,
625 .fault = filemap_fault,
626 .map_pages = filemap_map_pages,
627 .page_mkwrite = v9fs_vm_page_mkwrite,
631 const struct file_operations v9fs_cached_file_operations = {
632 .llseek = generic_file_llseek,
633 .read_iter = generic_file_read_iter,
634 .write_iter = generic_file_write_iter,
635 .open = v9fs_file_open,
636 .release = v9fs_dir_release,
637 .lock = v9fs_file_lock,
638 .mmap = v9fs_file_mmap,
639 .splice_read = generic_file_splice_read,
640 .splice_write = iter_file_splice_write,
641 .fsync = v9fs_file_fsync,
644 const struct file_operations v9fs_cached_file_operations_dotl = {
645 .llseek = generic_file_llseek,
646 .read_iter = generic_file_read_iter,
647 .write_iter = generic_file_write_iter,
648 .open = v9fs_file_open,
649 .release = v9fs_dir_release,
650 .lock = v9fs_file_lock_dotl,
651 .flock = v9fs_file_flock_dotl,
652 .mmap = v9fs_file_mmap,
653 .splice_read = generic_file_splice_read,
654 .splice_write = iter_file_splice_write,
655 .fsync = v9fs_file_fsync_dotl,
658 const struct file_operations v9fs_file_operations = {
659 .llseek = generic_file_llseek,
660 .read_iter = v9fs_file_read_iter,
661 .write_iter = v9fs_file_write_iter,
662 .open = v9fs_file_open,
663 .release = v9fs_dir_release,
664 .lock = v9fs_file_lock,
665 .mmap = generic_file_readonly_mmap,
666 .splice_read = generic_file_splice_read,
667 .splice_write = iter_file_splice_write,
668 .fsync = v9fs_file_fsync,
671 const struct file_operations v9fs_file_operations_dotl = {
672 .llseek = generic_file_llseek,
673 .read_iter = v9fs_file_read_iter,
674 .write_iter = v9fs_file_write_iter,
675 .open = v9fs_file_open,
676 .release = v9fs_dir_release,
677 .lock = v9fs_file_lock_dotl,
678 .flock = v9fs_file_flock_dotl,
679 .mmap = generic_file_readonly_mmap,
680 .splice_read = generic_file_splice_read,
681 .splice_write = iter_file_splice_write,
682 .fsync = v9fs_file_fsync_dotl,
685 const struct file_operations v9fs_mmap_file_operations = {
686 .llseek = generic_file_llseek,
687 .read_iter = v9fs_mmap_file_read_iter,
688 .write_iter = v9fs_mmap_file_write_iter,
689 .open = v9fs_file_open,
690 .release = v9fs_dir_release,
691 .lock = v9fs_file_lock,
692 .mmap = v9fs_mmap_file_mmap,
693 .splice_read = generic_file_splice_read,
694 .splice_write = iter_file_splice_write,
695 .fsync = v9fs_file_fsync,
698 const struct file_operations v9fs_mmap_file_operations_dotl = {
699 .llseek = generic_file_llseek,
700 .read_iter = v9fs_mmap_file_read_iter,
701 .write_iter = v9fs_mmap_file_write_iter,
702 .open = v9fs_file_open,
703 .release = v9fs_dir_release,
704 .lock = v9fs_file_lock_dotl,
705 .flock = v9fs_file_flock_dotl,
706 .mmap = v9fs_mmap_file_mmap,
707 .splice_read = generic_file_splice_read,
708 .splice_write = iter_file_splice_write,
709 .fsync = v9fs_file_fsync_dotl,