1/*
2 *  linux/fs/9p/vfs_file.c
3 *
4 * This file contians vfs file ops for 9P2000.
5 *
6 *  Copyright (C) 2004 by Eric Van Hensbergen <ericvh@gmail.com>
7 *  Copyright (C) 2002 by Ron Minnich <rminnich@lanl.gov>
8 *
9 *  This program is free software; you can redistribute it and/or modify
10 *  it under the terms of the GNU General Public License version 2
11 *  as published by the Free Software Foundation.
12 *
13 *  This program is distributed in the hope that it will be useful,
14 *  but WITHOUT ANY WARRANTY; without even the implied warranty of
15 *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
16 *  GNU General Public License for more details.
17 *
18 *  You should have received a copy of the GNU General Public License
19 *  along with this program; if not, write to:
20 *  Free Software Foundation
21 *  51 Franklin Street, Fifth Floor
22 *  Boston, MA  02111-1301  USA
23 *
24 */
25
26#include <linux/module.h>
27#include <linux/errno.h>
28#include <linux/fs.h>
29#include <linux/sched.h>
30#include <linux/file.h>
31#include <linux/stat.h>
32#include <linux/string.h>
33#include <linux/inet.h>
34#include <linux/list.h>
35#include <linux/pagemap.h>
36#include <linux/utsname.h>
37#include <asm/uaccess.h>
38#include <linux/idr.h>
39#include <linux/uio.h>
40#include <linux/slab.h>
41#include <net/9p/9p.h>
42#include <net/9p/client.h>
43
44#include "v9fs.h"
45#include "v9fs_vfs.h"
46#include "fid.h"
47#include "cache.h"
48
49static const struct vm_operations_struct v9fs_file_vm_ops;
50static const struct vm_operations_struct v9fs_mmap_file_vm_ops;
51
52/**
53 * v9fs_file_open - open a file (or directory)
54 * @inode: inode to be opened
55 * @file: file being opened
56 *
57 */
58
59int v9fs_file_open(struct inode *inode, struct file *file)
60{
61	int err;
62	struct v9fs_inode *v9inode;
63	struct v9fs_session_info *v9ses;
64	struct p9_fid *fid;
65	int omode;
66
67	p9_debug(P9_DEBUG_VFS, "inode: %p file: %p\n", inode, file);
68	v9inode = V9FS_I(inode);
69	v9ses = v9fs_inode2v9ses(inode);
70	if (v9fs_proto_dotl(v9ses))
71		omode = v9fs_open_to_dotl_flags(file->f_flags);
72	else
73		omode = v9fs_uflags2omode(file->f_flags,
74					v9fs_proto_dotu(v9ses));
75	fid = file->private_data;
76	if (!fid) {
77		fid = v9fs_fid_clone(file->f_path.dentry);
78		if (IS_ERR(fid))
79			return PTR_ERR(fid);
80
81		err = p9_client_open(fid, omode);
82		if (err < 0) {
83			p9_client_clunk(fid);
84			return err;
85		}
86		if ((file->f_flags & O_APPEND) &&
87			(!v9fs_proto_dotu(v9ses) && !v9fs_proto_dotl(v9ses)))
88			generic_file_llseek(file, 0, SEEK_END);
89	}
90
91	file->private_data = fid;
92	mutex_lock(&v9inode->v_mutex);
93	if ((v9ses->cache == CACHE_LOOSE || v9ses->cache == CACHE_FSCACHE) &&
94	    !v9inode->writeback_fid &&
95	    ((file->f_flags & O_ACCMODE) != O_RDONLY)) {
96		/*
97		 * clone a fid and add it to writeback_fid
98		 * we do it during open time instead of
99		 * page dirty time via write_begin/page_mkwrite
100		 * because we want write after unlink usecase
101		 * to work.
102		 */
103		fid = v9fs_writeback_fid(file->f_path.dentry);
104		if (IS_ERR(fid)) {
105			err = PTR_ERR(fid);
106			mutex_unlock(&v9inode->v_mutex);
107			goto out_error;
108		}
109		v9inode->writeback_fid = (void *) fid;
110	}
111	mutex_unlock(&v9inode->v_mutex);
112	if (v9ses->cache == CACHE_LOOSE || v9ses->cache == CACHE_FSCACHE)
113		v9fs_cache_inode_set_cookie(inode, file);
114	return 0;
115out_error:
116	p9_client_clunk(file->private_data);
117	file->private_data = NULL;
118	return err;
119}
120
121/**
122 * v9fs_file_lock - lock a file (or directory)
123 * @filp: file to be locked
124 * @cmd: lock command
125 * @fl: file lock structure
126 *
127 * Bugs: this looks like a local only lock, we should extend into 9P
128 *       by using open exclusive
129 */
130
131static int v9fs_file_lock(struct file *filp, int cmd, struct file_lock *fl)
132{
133	int res = 0;
134	struct inode *inode = file_inode(filp);
135
136	p9_debug(P9_DEBUG_VFS, "filp: %p lock: %p\n", filp, fl);
137
138	/* No mandatory locks */
139	if (__mandatory_lock(inode) && fl->fl_type != F_UNLCK)
140		return -ENOLCK;
141
142	if ((IS_SETLK(cmd) || IS_SETLKW(cmd)) && fl->fl_type != F_UNLCK) {
143		filemap_write_and_wait(inode->i_mapping);
144		invalidate_mapping_pages(&inode->i_data, 0, -1);
145	}
146
147	return res;
148}
149
150static int v9fs_file_do_lock(struct file *filp, int cmd, struct file_lock *fl)
151{
152	struct p9_flock flock;
153	struct p9_fid *fid;
154	uint8_t status = P9_LOCK_ERROR;
155	int res = 0;
156	unsigned char fl_type;
157
158	fid = filp->private_data;
159	BUG_ON(fid == NULL);
160
161	if ((fl->fl_flags & FL_POSIX) != FL_POSIX)
162		BUG();
163
164	res = posix_lock_file_wait(filp, fl);
165	if (res < 0)
166		goto out;
167
168	/* convert posix lock to p9 tlock args */
169	memset(&flock, 0, sizeof(flock));
170	/* map the lock type */
171	switch (fl->fl_type) {
172	case F_RDLCK:
173		flock.type = P9_LOCK_TYPE_RDLCK;
174		break;
175	case F_WRLCK:
176		flock.type = P9_LOCK_TYPE_WRLCK;
177		break;
178	case F_UNLCK:
179		flock.type = P9_LOCK_TYPE_UNLCK;
180		break;
181	}
182	flock.start = fl->fl_start;
183	if (fl->fl_end == OFFSET_MAX)
184		flock.length = 0;
185	else
186		flock.length = fl->fl_end - fl->fl_start + 1;
187	flock.proc_id = fl->fl_pid;
188	flock.client_id = fid->clnt->name;
189	if (IS_SETLKW(cmd))
190		flock.flags = P9_LOCK_FLAGS_BLOCK;
191
192	/*
193	 * if its a blocked request and we get P9_LOCK_BLOCKED as the status
194	 * for lock request, keep on trying
195	 */
196	for (;;) {
197		res = p9_client_lock_dotl(fid, &flock, &status);
198		if (res < 0)
199			goto out_unlock;
200
201		if (status != P9_LOCK_BLOCKED)
202			break;
203		if (status == P9_LOCK_BLOCKED && !IS_SETLKW(cmd))
204			break;
205		if (schedule_timeout_interruptible(P9_LOCK_TIMEOUT) != 0)
206			break;
207	}
208
209	/* map 9p status to VFS status */
210	switch (status) {
211	case P9_LOCK_SUCCESS:
212		res = 0;
213		break;
214	case P9_LOCK_BLOCKED:
215		res = -EAGAIN;
216		break;
217	default:
218		WARN_ONCE(1, "unknown lock status code: %d\n", status);
219		/* fallthough */
220	case P9_LOCK_ERROR:
221	case P9_LOCK_GRACE:
222		res = -ENOLCK;
223		break;
224	}
225
226out_unlock:
227	/*
228	 * incase server returned error for lock request, revert
229	 * it locally
230	 */
231	if (res < 0 && fl->fl_type != F_UNLCK) {
232		fl_type = fl->fl_type;
233		fl->fl_type = F_UNLCK;
234		res = posix_lock_file_wait(filp, fl);
235		fl->fl_type = fl_type;
236	}
237out:
238	return res;
239}
240
241static int v9fs_file_getlock(struct file *filp, struct file_lock *fl)
242{
243	struct p9_getlock glock;
244	struct p9_fid *fid;
245	int res = 0;
246
247	fid = filp->private_data;
248	BUG_ON(fid == NULL);
249
250	posix_test_lock(filp, fl);
251	/*
252	 * if we have a conflicting lock locally, no need to validate
253	 * with server
254	 */
255	if (fl->fl_type != F_UNLCK)
256		return res;
257
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)
263		glock.length = 0;
264	else
265		glock.length = fl->fl_end - fl->fl_start + 1;
266	glock.proc_id = fl->fl_pid;
267	glock.client_id = fid->clnt->name;
268
269	res = p9_client_getlock_dotl(fid, &glock);
270	if (res < 0)
271		return res;
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;
276		break;
277	case P9_LOCK_TYPE_WRLCK:
278		fl->fl_type = F_WRLCK;
279		break;
280	case P9_LOCK_TYPE_UNLCK:
281		fl->fl_type = F_UNLCK;
282		break;
283	}
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;
288		else
289			fl->fl_end = glock.start + glock.length - 1;
290		fl->fl_pid = glock.proc_id;
291	}
292	kfree(glock.client_id);
293	return res;
294}
295
296/**
297 * v9fs_file_lock_dotl - lock a file (or directory)
298 * @filp: file to be locked
299 * @cmd: lock command
300 * @fl: file lock structure
301 *
302 */
303
304static int v9fs_file_lock_dotl(struct file *filp, int cmd, struct file_lock *fl)
305{
306	struct inode *inode = file_inode(filp);
307	int ret = -ENOLCK;
308
309	p9_debug(P9_DEBUG_VFS, "filp: %p cmd:%d lock: %p name: %pD\n",
310		 filp, cmd, fl, filp);
311
312	/* No mandatory locks */
313	if (__mandatory_lock(inode) && fl->fl_type != F_UNLCK)
314		goto out_err;
315
316	if ((IS_SETLK(cmd) || IS_SETLKW(cmd)) && fl->fl_type != F_UNLCK) {
317		filemap_write_and_wait(inode->i_mapping);
318		invalidate_mapping_pages(&inode->i_data, 0, -1);
319	}
320
321	if (IS_SETLK(cmd) || IS_SETLKW(cmd))
322		ret = v9fs_file_do_lock(filp, cmd, fl);
323	else if (IS_GETLK(cmd))
324		ret = v9fs_file_getlock(filp, fl);
325	else
326		ret = -EINVAL;
327out_err:
328	return ret;
329}
330
331/**
332 * v9fs_file_flock_dotl - lock a file
333 * @filp: file to be locked
334 * @cmd: lock command
335 * @fl: file lock structure
336 *
337 */
338
339static int v9fs_file_flock_dotl(struct file *filp, int cmd,
340	struct file_lock *fl)
341{
342	struct inode *inode = file_inode(filp);
343	int ret = -ENOLCK;
344
345	p9_debug(P9_DEBUG_VFS, "filp: %p cmd:%d lock: %p name: %pD\n",
346		 filp, cmd, fl, filp);
347
348	/* No mandatory locks */
349	if (__mandatory_lock(inode) && fl->fl_type != F_UNLCK)
350		goto out_err;
351
352	if (!(fl->fl_flags & FL_FLOCK))
353		goto out_err;
354
355	if ((IS_SETLK(cmd) || IS_SETLKW(cmd)) && fl->fl_type != F_UNLCK) {
356		filemap_write_and_wait(inode->i_mapping);
357		invalidate_mapping_pages(&inode->i_data, 0, -1);
358	}
359	/* Convert flock to posix lock */
360	fl->fl_flags |= FL_POSIX;
361	fl->fl_flags ^= FL_FLOCK;
362
363	if (IS_SETLK(cmd) | IS_SETLKW(cmd))
364		ret = v9fs_file_do_lock(filp, cmd, fl);
365	else
366		ret = -EINVAL;
367out_err:
368	return ret;
369}
370
371/**
372 * v9fs_file_read - read from a file
373 * @filp: file pointer to read
374 * @udata: user data buffer to read data into
375 * @count: size of buffer
376 * @offset: offset at which to read data
377 *
378 */
379
380static ssize_t
381v9fs_file_read_iter(struct kiocb *iocb, struct iov_iter *to)
382{
383	struct p9_fid *fid = iocb->ki_filp->private_data;
384	int ret, err;
385
386	p9_debug(P9_DEBUG_VFS, "count %zu offset %lld\n",
387		 iov_iter_count(to), iocb->ki_pos);
388
389	ret = p9_client_read(fid, iocb->ki_pos, to, &err);
390	if (!ret)
391		return err;
392
393	iocb->ki_pos += ret;
394	return ret;
395}
396
397/**
398 * v9fs_file_write - write to a file
399 * @filp: file pointer to write
400 * @data: data buffer to write data from
401 * @count: size of buffer
402 * @offset: offset at which to write data
403 *
404 */
405static ssize_t
406v9fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
407{
408	struct file *file = iocb->ki_filp;
409	ssize_t retval;
410	loff_t origin;
411	int err = 0;
412
413	retval = generic_write_checks(iocb, from);
414	if (retval <= 0)
415		return retval;
416
417	origin = iocb->ki_pos;
418	retval = p9_client_write(file->private_data, iocb->ki_pos, from, &err);
419	if (retval > 0) {
420		struct inode *inode = file_inode(file);
421		loff_t i_size;
422		unsigned long pg_start, pg_end;
423		pg_start = origin >> PAGE_CACHE_SHIFT;
424		pg_end = (origin + retval - 1) >> PAGE_CACHE_SHIFT;
425		if (inode->i_mapping && inode->i_mapping->nrpages)
426			invalidate_inode_pages2_range(inode->i_mapping,
427						      pg_start, pg_end);
428		iocb->ki_pos += retval;
429		i_size = i_size_read(inode);
430		if (iocb->ki_pos > i_size) {
431			inode_add_bytes(inode, iocb->ki_pos - i_size);
432			i_size_write(inode, iocb->ki_pos);
433		}
434		return retval;
435	}
436	return err;
437}
438
439static int v9fs_file_fsync(struct file *filp, loff_t start, loff_t end,
440			   int datasync)
441{
442	struct p9_fid *fid;
443	struct inode *inode = filp->f_mapping->host;
444	struct p9_wstat wstat;
445	int retval;
446
447	retval = filemap_write_and_wait_range(inode->i_mapping, start, end);
448	if (retval)
449		return retval;
450
451	mutex_lock(&inode->i_mutex);
452	p9_debug(P9_DEBUG_VFS, "filp %p datasync %x\n", filp, datasync);
453
454	fid = filp->private_data;
455	v9fs_blank_wstat(&wstat);
456
457	retval = p9_client_wstat(fid, &wstat);
458	mutex_unlock(&inode->i_mutex);
459
460	return retval;
461}
462
463int v9fs_file_fsync_dotl(struct file *filp, loff_t start, loff_t end,
464			 int datasync)
465{
466	struct p9_fid *fid;
467	struct inode *inode = filp->f_mapping->host;
468	int retval;
469
470	retval = filemap_write_and_wait_range(inode->i_mapping, start, end);
471	if (retval)
472		return retval;
473
474	mutex_lock(&inode->i_mutex);
475	p9_debug(P9_DEBUG_VFS, "filp %p datasync %x\n", filp, datasync);
476
477	fid = filp->private_data;
478
479	retval = p9_client_fsync(fid, datasync);
480	mutex_unlock(&inode->i_mutex);
481
482	return retval;
483}
484
485static int
486v9fs_file_mmap(struct file *filp, struct vm_area_struct *vma)
487{
488	int retval;
489
490
491	retval = generic_file_mmap(filp, vma);
492	if (!retval)
493		vma->vm_ops = &v9fs_file_vm_ops;
494
495	return retval;
496}
497
498static int
499v9fs_mmap_file_mmap(struct file *filp, struct vm_area_struct *vma)
500{
501	int retval;
502	struct inode *inode;
503	struct v9fs_inode *v9inode;
504	struct p9_fid *fid;
505
506	inode = file_inode(filp);
507	v9inode = V9FS_I(inode);
508	mutex_lock(&v9inode->v_mutex);
509	if (!v9inode->writeback_fid &&
510	    (vma->vm_flags & VM_WRITE)) {
511		/*
512		 * clone a fid and add it to writeback_fid
513		 * we do it during mmap instead of
514		 * page dirty time via write_begin/page_mkwrite
515		 * because we want write after unlink usecase
516		 * to work.
517		 */
518		fid = v9fs_writeback_fid(filp->f_path.dentry);
519		if (IS_ERR(fid)) {
520			retval = PTR_ERR(fid);
521			mutex_unlock(&v9inode->v_mutex);
522			return retval;
523		}
524		v9inode->writeback_fid = (void *) fid;
525	}
526	mutex_unlock(&v9inode->v_mutex);
527
528	retval = generic_file_mmap(filp, vma);
529	if (!retval)
530		vma->vm_ops = &v9fs_mmap_file_vm_ops;
531
532	return retval;
533}
534
535static int
536v9fs_vm_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
537{
538	struct v9fs_inode *v9inode;
539	struct page *page = vmf->page;
540	struct file *filp = vma->vm_file;
541	struct inode *inode = file_inode(filp);
542
543
544	p9_debug(P9_DEBUG_VFS, "page %p fid %lx\n",
545		 page, (unsigned long)filp->private_data);
546
547	/* Update file times before taking page lock */
548	file_update_time(filp);
549
550	v9inode = V9FS_I(inode);
551	/* make sure the cache has finished storing the page */
552	v9fs_fscache_wait_on_page_write(inode, page);
553	BUG_ON(!v9inode->writeback_fid);
554	lock_page(page);
555	if (page->mapping != inode->i_mapping)
556		goto out_unlock;
557	wait_for_stable_page(page);
558
559	return VM_FAULT_LOCKED;
560out_unlock:
561	unlock_page(page);
562	return VM_FAULT_NOPAGE;
563}
564
565/**
566 * v9fs_mmap_file_read - read from a file
567 * @filp: file pointer to read
568 * @data: user data buffer to read data into
569 * @count: size of buffer
570 * @offset: offset at which to read data
571 *
572 */
573static ssize_t
574v9fs_mmap_file_read_iter(struct kiocb *iocb, struct iov_iter *to)
575{
576	/* TODO: Check if there are dirty pages */
577	return v9fs_file_read_iter(iocb, to);
578}
579
580/**
581 * v9fs_mmap_file_write - write to a file
582 * @filp: file pointer to write
583 * @data: data buffer to write data from
584 * @count: size of buffer
585 * @offset: offset at which to write data
586 *
587 */
588static ssize_t
589v9fs_mmap_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
590{
591	/*
592	 * TODO: invalidate mmaps on filp's inode between
593	 * offset and offset+count
594	 */
595	return v9fs_file_write_iter(iocb, from);
596}
597
598static void v9fs_mmap_vm_close(struct vm_area_struct *vma)
599{
600	struct inode *inode;
601
602	struct writeback_control wbc = {
603		.nr_to_write = LONG_MAX,
604		.sync_mode = WB_SYNC_ALL,
605		.range_start = vma->vm_pgoff * PAGE_SIZE,
606		 /* absolute end, byte at end included */
607		.range_end = vma->vm_pgoff * PAGE_SIZE +
608			(vma->vm_end - vma->vm_start - 1),
609	};
610
611
612	p9_debug(P9_DEBUG_VFS, "9p VMA close, %p, flushing", vma);
613
614	inode = file_inode(vma->vm_file);
615
616	if (!mapping_cap_writeback_dirty(inode->i_mapping))
617		wbc.nr_to_write = 0;
618
619	might_sleep();
620	sync_inode(inode, &wbc);
621}
622
623
624static const struct vm_operations_struct v9fs_file_vm_ops = {
625	.fault = filemap_fault,
626	.map_pages = filemap_map_pages,
627	.page_mkwrite = v9fs_vm_page_mkwrite,
628};
629
630static const struct vm_operations_struct v9fs_mmap_file_vm_ops = {
631	.close = v9fs_mmap_vm_close,
632	.fault = filemap_fault,
633	.map_pages = filemap_map_pages,
634	.page_mkwrite = v9fs_vm_page_mkwrite,
635};
636
637
638const struct file_operations v9fs_cached_file_operations = {
639	.llseek = generic_file_llseek,
640	.read_iter = generic_file_read_iter,
641	.write_iter = generic_file_write_iter,
642	.open = v9fs_file_open,
643	.release = v9fs_dir_release,
644	.lock = v9fs_file_lock,
645	.mmap = v9fs_file_mmap,
646	.fsync = v9fs_file_fsync,
647};
648
649const struct file_operations v9fs_cached_file_operations_dotl = {
650	.llseek = generic_file_llseek,
651	.read_iter = generic_file_read_iter,
652	.write_iter = generic_file_write_iter,
653	.open = v9fs_file_open,
654	.release = v9fs_dir_release,
655	.lock = v9fs_file_lock_dotl,
656	.flock = v9fs_file_flock_dotl,
657	.mmap = v9fs_file_mmap,
658	.fsync = v9fs_file_fsync_dotl,
659};
660
661const struct file_operations v9fs_file_operations = {
662	.llseek = generic_file_llseek,
663	.read_iter = v9fs_file_read_iter,
664	.write_iter = v9fs_file_write_iter,
665	.open = v9fs_file_open,
666	.release = v9fs_dir_release,
667	.lock = v9fs_file_lock,
668	.mmap = generic_file_readonly_mmap,
669	.fsync = v9fs_file_fsync,
670};
671
672const struct file_operations v9fs_file_operations_dotl = {
673	.llseek = generic_file_llseek,
674	.read_iter = v9fs_file_read_iter,
675	.write_iter = v9fs_file_write_iter,
676	.open = v9fs_file_open,
677	.release = v9fs_dir_release,
678	.lock = v9fs_file_lock_dotl,
679	.flock = v9fs_file_flock_dotl,
680	.mmap = generic_file_readonly_mmap,
681	.fsync = v9fs_file_fsync_dotl,
682};
683
684const struct file_operations v9fs_mmap_file_operations = {
685	.llseek = generic_file_llseek,
686	.read_iter = v9fs_mmap_file_read_iter,
687	.write_iter = v9fs_mmap_file_write_iter,
688	.open = v9fs_file_open,
689	.release = v9fs_dir_release,
690	.lock = v9fs_file_lock,
691	.mmap = v9fs_mmap_file_mmap,
692	.fsync = v9fs_file_fsync,
693};
694
695const struct file_operations v9fs_mmap_file_operations_dotl = {
696	.llseek = generic_file_llseek,
697	.read_iter = v9fs_mmap_file_read_iter,
698	.write_iter = v9fs_mmap_file_write_iter,
699	.open = v9fs_file_open,
700	.release = v9fs_dir_release,
701	.lock = v9fs_file_lock_dotl,
702	.flock = v9fs_file_flock_dotl,
703	.mmap = v9fs_mmap_file_mmap,
704	.fsync = v9fs_file_fsync_dotl,
705};
706