1 // SPDX-License-Identifier: GPL-2.0-only
3 * linux/fs/9p/vfs_file.c
5 * This file contians vfs file ops for 9P2000.
7 * Copyright (C) 2004 by Eric Van Hensbergen <ericvh@gmail.com>
8 * Copyright (C) 2002 by Ron Minnich <rminnich@lanl.gov>
11 #include <linux/module.h>
12 #include <linux/errno.h>
14 #include <linux/sched.h>
15 #include <linux/file.h>
16 #include <linux/stat.h>
17 #include <linux/string.h>
18 #include <linux/inet.h>
19 #include <linux/list.h>
20 #include <linux/pagemap.h>
21 #include <linux/utsname.h>
22 #include <linux/uaccess.h>
23 #include <linux/idr.h>
24 #include <linux/uio.h>
25 #include <linux/slab.h>
26 #include <net/9p/9p.h>
27 #include <net/9p/client.h>
34 static const struct vm_operations_struct v9fs_file_vm_ops
;
35 static const struct vm_operations_struct v9fs_mmap_file_vm_ops
;
38 * v9fs_file_open - open a file (or directory)
39 * @inode: inode to be opened
40 * @file: file being opened
44 int v9fs_file_open(struct inode
*inode
, struct file
*file
)
47 struct v9fs_inode
*v9inode
;
48 struct v9fs_session_info
*v9ses
;
52 p9_debug(P9_DEBUG_VFS
, "inode: %p file: %p\n", inode
, file
);
53 v9inode
= V9FS_I(inode
);
54 v9ses
= v9fs_inode2v9ses(inode
);
55 if (v9fs_proto_dotl(v9ses
))
56 omode
= v9fs_open_to_dotl_flags(file
->f_flags
);
58 omode
= v9fs_uflags2omode(file
->f_flags
,
59 v9fs_proto_dotu(v9ses
));
60 fid
= file
->private_data
;
62 fid
= v9fs_fid_clone(file_dentry(file
));
66 err
= p9_client_open(fid
, omode
);
71 if ((file
->f_flags
& O_APPEND
) &&
72 (!v9fs_proto_dotu(v9ses
) && !v9fs_proto_dotl(v9ses
)))
73 generic_file_llseek(file
, 0, SEEK_END
);
76 file
->private_data
= fid
;
77 mutex_lock(&v9inode
->v_mutex
);
78 if ((v9ses
->cache
== CACHE_LOOSE
|| v9ses
->cache
== CACHE_FSCACHE
) &&
79 !v9inode
->writeback_fid
&&
80 ((file
->f_flags
& O_ACCMODE
) != O_RDONLY
)) {
82 * clone a fid and add it to writeback_fid
83 * we do it during open time instead of
84 * page dirty time via write_begin/page_mkwrite
85 * because we want write after unlink usecase
88 fid
= v9fs_writeback_fid(file_dentry(file
));
91 mutex_unlock(&v9inode
->v_mutex
);
94 v9inode
->writeback_fid
= (void *) fid
;
96 mutex_unlock(&v9inode
->v_mutex
);
97 if (v9ses
->cache
== CACHE_LOOSE
|| v9ses
->cache
== CACHE_FSCACHE
)
98 v9fs_cache_inode_set_cookie(inode
, file
);
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
)
119 struct inode
*inode
= file_inode(filp
);
121 p9_debug(P9_DEBUG_VFS
, "filp: %p lock: %p\n", filp
, fl
);
123 /* No mandatory locks */
124 if (__mandatory_lock(inode
) && fl
->fl_type
!= F_UNLCK
)
127 if ((IS_SETLK(cmd
) || IS_SETLKW(cmd
)) && fl
->fl_type
!= F_UNLCK
) {
128 filemap_write_and_wait(inode
->i_mapping
);
129 invalidate_mapping_pages(&inode
->i_data
, 0, -1);
135 static int v9fs_file_do_lock(struct file
*filp
, int cmd
, struct file_lock
*fl
)
137 struct p9_flock flock
;
139 uint8_t status
= P9_LOCK_ERROR
;
141 unsigned char fl_type
;
142 struct v9fs_session_info
*v9ses
;
144 fid
= filp
->private_data
;
147 if ((fl
->fl_flags
& FL_POSIX
) != FL_POSIX
)
150 res
= locks_lock_file_wait(filp
, fl
);
154 /* convert posix lock to p9 tlock args */
155 memset(&flock
, 0, sizeof(flock
));
156 /* map the lock type */
157 switch (fl
->fl_type
) {
159 flock
.type
= P9_LOCK_TYPE_RDLCK
;
162 flock
.type
= P9_LOCK_TYPE_WRLCK
;
165 flock
.type
= P9_LOCK_TYPE_UNLCK
;
168 flock
.start
= fl
->fl_start
;
169 if (fl
->fl_end
== OFFSET_MAX
)
172 flock
.length
= fl
->fl_end
- fl
->fl_start
+ 1;
173 flock
.proc_id
= fl
->fl_pid
;
174 flock
.client_id
= fid
->clnt
->name
;
176 flock
.flags
= P9_LOCK_FLAGS_BLOCK
;
178 v9ses
= v9fs_inode2v9ses(file_inode(filp
));
181 * if its a blocked request and we get P9_LOCK_BLOCKED as the status
182 * for lock request, keep on trying
185 res
= p9_client_lock_dotl(fid
, &flock
, &status
);
189 if (status
!= P9_LOCK_BLOCKED
)
191 if (status
== P9_LOCK_BLOCKED
&& !IS_SETLKW(cmd
))
193 if (schedule_timeout_interruptible(v9ses
->session_lock_timeout
)
197 * p9_client_lock_dotl overwrites flock.client_id with the
198 * server message, free and reuse the client name
200 if (flock
.client_id
!= fid
->clnt
->name
) {
201 kfree(flock
.client_id
);
202 flock
.client_id
= fid
->clnt
->name
;
206 /* map 9p status to VFS status */
208 case P9_LOCK_SUCCESS
:
211 case P9_LOCK_BLOCKED
:
215 WARN_ONCE(1, "unknown lock status code: %d\n", status
);
225 * incase server returned error for lock request, revert
228 if (res
< 0 && fl
->fl_type
!= F_UNLCK
) {
229 fl_type
= fl
->fl_type
;
230 fl
->fl_type
= F_UNLCK
;
231 /* Even if this fails we want to return the remote error */
232 locks_lock_file_wait(filp
, fl
);
233 fl
->fl_type
= fl_type
;
235 if (flock
.client_id
!= fid
->clnt
->name
)
236 kfree(flock
.client_id
);
241 static int v9fs_file_getlock(struct file
*filp
, struct file_lock
*fl
)
243 struct p9_getlock glock
;
247 fid
= filp
->private_data
;
250 posix_test_lock(filp
, fl
);
252 * if we have a conflicting lock locally, no need to validate
255 if (fl
->fl_type
!= F_UNLCK
)
258 /* convert posix lock to p9 tgetlock args */
259 memset(&glock
, 0, sizeof(glock
));
260 glock
.type
= P9_LOCK_TYPE_UNLCK
;
261 glock
.start
= fl
->fl_start
;
262 if (fl
->fl_end
== OFFSET_MAX
)
265 glock
.length
= fl
->fl_end
- fl
->fl_start
+ 1;
266 glock
.proc_id
= fl
->fl_pid
;
267 glock
.client_id
= fid
->clnt
->name
;
269 res
= p9_client_getlock_dotl(fid
, &glock
);
272 /* map 9p lock type to os lock type */
273 switch (glock
.type
) {
274 case P9_LOCK_TYPE_RDLCK
:
275 fl
->fl_type
= F_RDLCK
;
277 case P9_LOCK_TYPE_WRLCK
:
278 fl
->fl_type
= F_WRLCK
;
280 case P9_LOCK_TYPE_UNLCK
:
281 fl
->fl_type
= F_UNLCK
;
284 if (glock
.type
!= P9_LOCK_TYPE_UNLCK
) {
285 fl
->fl_start
= glock
.start
;
286 if (glock
.length
== 0)
287 fl
->fl_end
= OFFSET_MAX
;
289 fl
->fl_end
= glock
.start
+ glock
.length
- 1;
290 fl
->fl_pid
= -glock
.proc_id
;
293 if (glock
.client_id
!= fid
->clnt
->name
)
294 kfree(glock
.client_id
);
299 * v9fs_file_lock_dotl - lock a file (or directory)
300 * @filp: file to be locked
302 * @fl: file lock structure
306 static int v9fs_file_lock_dotl(struct file
*filp
, int cmd
, struct file_lock
*fl
)
308 struct inode
*inode
= file_inode(filp
);
311 p9_debug(P9_DEBUG_VFS
, "filp: %p cmd:%d lock: %p name: %pD\n",
312 filp
, cmd
, fl
, filp
);
314 /* No mandatory locks */
315 if (__mandatory_lock(inode
) && fl
->fl_type
!= F_UNLCK
)
318 if ((IS_SETLK(cmd
) || IS_SETLKW(cmd
)) && fl
->fl_type
!= F_UNLCK
) {
319 filemap_write_and_wait(inode
->i_mapping
);
320 invalidate_mapping_pages(&inode
->i_data
, 0, -1);
323 if (IS_SETLK(cmd
) || IS_SETLKW(cmd
))
324 ret
= v9fs_file_do_lock(filp
, cmd
, fl
);
325 else if (IS_GETLK(cmd
))
326 ret
= v9fs_file_getlock(filp
, fl
);
334 * v9fs_file_flock_dotl - lock a file
335 * @filp: file to be locked
337 * @fl: file lock structure
341 static int v9fs_file_flock_dotl(struct file
*filp
, int cmd
,
342 struct file_lock
*fl
)
344 struct inode
*inode
= file_inode(filp
);
347 p9_debug(P9_DEBUG_VFS
, "filp: %p cmd:%d lock: %p name: %pD\n",
348 filp
, cmd
, fl
, filp
);
350 /* No mandatory locks */
351 if (__mandatory_lock(inode
) && fl
->fl_type
!= F_UNLCK
)
354 if (!(fl
->fl_flags
& FL_FLOCK
))
357 if ((IS_SETLK(cmd
) || IS_SETLKW(cmd
)) && fl
->fl_type
!= F_UNLCK
) {
358 filemap_write_and_wait(inode
->i_mapping
);
359 invalidate_mapping_pages(&inode
->i_data
, 0, -1);
361 /* Convert flock to posix lock */
362 fl
->fl_flags
|= FL_POSIX
;
363 fl
->fl_flags
^= FL_FLOCK
;
365 if (IS_SETLK(cmd
) | IS_SETLKW(cmd
))
366 ret
= v9fs_file_do_lock(filp
, cmd
, fl
);
374 * v9fs_file_read - read from a file
375 * @filp: file pointer to read
376 * @udata: user data buffer to read data into
377 * @count: size of buffer
378 * @offset: offset at which to read data
383 v9fs_file_read_iter(struct kiocb
*iocb
, struct iov_iter
*to
)
385 struct p9_fid
*fid
= iocb
->ki_filp
->private_data
;
388 p9_debug(P9_DEBUG_VFS
, "count %zu offset %lld\n",
389 iov_iter_count(to
), iocb
->ki_pos
);
391 ret
= p9_client_read(fid
, iocb
->ki_pos
, to
, &err
);
400 * v9fs_file_write - write to a file
401 * @filp: file pointer to write
402 * @data: data buffer to write data from
403 * @count: size of buffer
404 * @offset: offset at which to write data
408 v9fs_file_write_iter(struct kiocb
*iocb
, struct iov_iter
*from
)
410 struct file
*file
= iocb
->ki_filp
;
415 retval
= generic_write_checks(iocb
, from
);
419 origin
= iocb
->ki_pos
;
420 retval
= p9_client_write(file
->private_data
, iocb
->ki_pos
, from
, &err
);
422 struct inode
*inode
= file_inode(file
);
424 unsigned long pg_start
, pg_end
;
425 pg_start
= origin
>> PAGE_SHIFT
;
426 pg_end
= (origin
+ retval
- 1) >> PAGE_SHIFT
;
427 if (inode
->i_mapping
&& inode
->i_mapping
->nrpages
)
428 invalidate_inode_pages2_range(inode
->i_mapping
,
430 iocb
->ki_pos
+= retval
;
431 i_size
= i_size_read(inode
);
432 if (iocb
->ki_pos
> i_size
) {
433 inode_add_bytes(inode
, iocb
->ki_pos
- i_size
);
435 * Need to serialize against i_size_write() in
438 v9fs_i_size_write(inode
, iocb
->ki_pos
);
445 static int v9fs_file_fsync(struct file
*filp
, loff_t start
, loff_t end
,
449 struct inode
*inode
= filp
->f_mapping
->host
;
450 struct p9_wstat wstat
;
453 retval
= file_write_and_wait_range(filp
, start
, end
);
458 p9_debug(P9_DEBUG_VFS
, "filp %p datasync %x\n", filp
, datasync
);
460 fid
= filp
->private_data
;
461 v9fs_blank_wstat(&wstat
);
463 retval
= p9_client_wstat(fid
, &wstat
);
469 int v9fs_file_fsync_dotl(struct file
*filp
, loff_t start
, loff_t end
,
473 struct inode
*inode
= filp
->f_mapping
->host
;
476 retval
= file_write_and_wait_range(filp
, start
, end
);
481 p9_debug(P9_DEBUG_VFS
, "filp %p datasync %x\n", filp
, datasync
);
483 fid
= filp
->private_data
;
485 retval
= p9_client_fsync(fid
, datasync
);
492 v9fs_file_mmap(struct file
*filp
, struct vm_area_struct
*vma
)
497 retval
= generic_file_mmap(filp
, vma
);
499 vma
->vm_ops
= &v9fs_file_vm_ops
;
505 v9fs_mmap_file_mmap(struct file
*filp
, struct vm_area_struct
*vma
)
509 struct v9fs_inode
*v9inode
;
512 inode
= file_inode(filp
);
513 v9inode
= V9FS_I(inode
);
514 mutex_lock(&v9inode
->v_mutex
);
515 if (!v9inode
->writeback_fid
&&
516 (vma
->vm_flags
& VM_SHARED
) &&
517 (vma
->vm_flags
& VM_WRITE
)) {
519 * clone a fid and add it to writeback_fid
520 * we do it during mmap instead of
521 * page dirty time via write_begin/page_mkwrite
522 * because we want write after unlink usecase
525 fid
= v9fs_writeback_fid(file_dentry(filp
));
527 retval
= PTR_ERR(fid
);
528 mutex_unlock(&v9inode
->v_mutex
);
531 v9inode
->writeback_fid
= (void *) fid
;
533 mutex_unlock(&v9inode
->v_mutex
);
535 retval
= generic_file_mmap(filp
, vma
);
537 vma
->vm_ops
= &v9fs_mmap_file_vm_ops
;
543 v9fs_vm_page_mkwrite(struct vm_fault
*vmf
)
545 struct v9fs_inode
*v9inode
;
546 struct page
*page
= vmf
->page
;
547 struct file
*filp
= vmf
->vma
->vm_file
;
548 struct inode
*inode
= file_inode(filp
);
551 p9_debug(P9_DEBUG_VFS
, "page %p fid %lx\n",
552 page
, (unsigned long)filp
->private_data
);
554 /* Update file times before taking page lock */
555 file_update_time(filp
);
557 v9inode
= V9FS_I(inode
);
558 /* make sure the cache has finished storing the page */
559 v9fs_fscache_wait_on_page_write(inode
, page
);
560 BUG_ON(!v9inode
->writeback_fid
);
562 if (page
->mapping
!= inode
->i_mapping
)
564 wait_for_stable_page(page
);
566 return VM_FAULT_LOCKED
;
569 return VM_FAULT_NOPAGE
;
573 * v9fs_mmap_file_read - read from a file
574 * @filp: file pointer to read
575 * @data: user data buffer to read data into
576 * @count: size of buffer
577 * @offset: offset at which to read data
581 v9fs_mmap_file_read_iter(struct kiocb
*iocb
, struct iov_iter
*to
)
583 /* TODO: Check if there are dirty pages */
584 return v9fs_file_read_iter(iocb
, to
);
588 * v9fs_mmap_file_write - write to a file
589 * @filp: file pointer to write
590 * @data: data buffer to write data from
591 * @count: size of buffer
592 * @offset: offset at which to write data
596 v9fs_mmap_file_write_iter(struct kiocb
*iocb
, struct iov_iter
*from
)
599 * TODO: invalidate mmaps on filp's inode between
600 * offset and offset+count
602 return v9fs_file_write_iter(iocb
, from
);
605 static void v9fs_mmap_vm_close(struct vm_area_struct
*vma
)
609 struct writeback_control wbc
= {
610 .nr_to_write
= LONG_MAX
,
611 .sync_mode
= WB_SYNC_ALL
,
612 .range_start
= vma
->vm_pgoff
* PAGE_SIZE
,
613 /* absolute end, byte at end included */
614 .range_end
= vma
->vm_pgoff
* PAGE_SIZE
+
615 (vma
->vm_end
- vma
->vm_start
- 1),
618 if (!(vma
->vm_flags
& VM_SHARED
))
621 p9_debug(P9_DEBUG_VFS
, "9p VMA close, %p, flushing", vma
);
623 inode
= file_inode(vma
->vm_file
);
625 if (!mapping_cap_writeback_dirty(inode
->i_mapping
))
629 sync_inode(inode
, &wbc
);
633 static const struct vm_operations_struct v9fs_file_vm_ops
= {
634 .fault
= filemap_fault
,
635 .map_pages
= filemap_map_pages
,
636 .page_mkwrite
= v9fs_vm_page_mkwrite
,
639 static const struct vm_operations_struct v9fs_mmap_file_vm_ops
= {
640 .close
= v9fs_mmap_vm_close
,
641 .fault
= filemap_fault
,
642 .map_pages
= filemap_map_pages
,
643 .page_mkwrite
= v9fs_vm_page_mkwrite
,
647 const struct file_operations v9fs_cached_file_operations
= {
648 .llseek
= generic_file_llseek
,
649 .read_iter
= generic_file_read_iter
,
650 .write_iter
= generic_file_write_iter
,
651 .open
= v9fs_file_open
,
652 .release
= v9fs_dir_release
,
653 .lock
= v9fs_file_lock
,
654 .mmap
= v9fs_file_mmap
,
655 .fsync
= v9fs_file_fsync
,
658 const struct file_operations v9fs_cached_file_operations_dotl
= {
659 .llseek
= generic_file_llseek
,
660 .read_iter
= generic_file_read_iter
,
661 .write_iter
= generic_file_write_iter
,
662 .open
= v9fs_file_open
,
663 .release
= v9fs_dir_release
,
664 .lock
= v9fs_file_lock_dotl
,
665 .flock
= v9fs_file_flock_dotl
,
666 .mmap
= v9fs_file_mmap
,
667 .fsync
= v9fs_file_fsync_dotl
,
670 const struct file_operations v9fs_file_operations
= {
671 .llseek
= generic_file_llseek
,
672 .read_iter
= v9fs_file_read_iter
,
673 .write_iter
= v9fs_file_write_iter
,
674 .open
= v9fs_file_open
,
675 .release
= v9fs_dir_release
,
676 .lock
= v9fs_file_lock
,
677 .mmap
= generic_file_readonly_mmap
,
678 .fsync
= v9fs_file_fsync
,
681 const struct file_operations v9fs_file_operations_dotl
= {
682 .llseek
= generic_file_llseek
,
683 .read_iter
= v9fs_file_read_iter
,
684 .write_iter
= v9fs_file_write_iter
,
685 .open
= v9fs_file_open
,
686 .release
= v9fs_dir_release
,
687 .lock
= v9fs_file_lock_dotl
,
688 .flock
= v9fs_file_flock_dotl
,
689 .mmap
= generic_file_readonly_mmap
,
690 .fsync
= v9fs_file_fsync_dotl
,
693 const struct file_operations v9fs_mmap_file_operations
= {
694 .llseek
= generic_file_llseek
,
695 .read_iter
= v9fs_mmap_file_read_iter
,
696 .write_iter
= v9fs_mmap_file_write_iter
,
697 .open
= v9fs_file_open
,
698 .release
= v9fs_dir_release
,
699 .lock
= v9fs_file_lock
,
700 .mmap
= v9fs_mmap_file_mmap
,
701 .fsync
= v9fs_file_fsync
,
704 const struct file_operations v9fs_mmap_file_operations_dotl
= {
705 .llseek
= generic_file_llseek
,
706 .read_iter
= v9fs_mmap_file_read_iter
,
707 .write_iter
= v9fs_mmap_file_write_iter
,
708 .open
= v9fs_file_open
,
709 .release
= v9fs_dir_release
,
710 .lock
= v9fs_file_lock_dotl
,
711 .flock
= v9fs_file_flock_dotl
,
712 .mmap
= v9fs_mmap_file_mmap
,
713 .fsync
= v9fs_file_fsync_dotl
,