2 * linux/fs/9p/vfs_file.c
4 * This file contians vfs file ops for 9P2000.
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.
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.
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
26 #include <linux/module.h>
27 #include <linux/errno.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 <net/9p/9p.h>
40 #include <net/9p/client.h>
47 static const struct vm_operations_struct v9fs_file_vm_ops;
48 static const struct vm_operations_struct v9fs_mmap_file_vm_ops;
51 * v9fs_file_open - open a file (or directory)
52 * @inode: inode to be opened
53 * @file: file being opened
57 int v9fs_file_open(struct inode *inode, struct file *file)
60 struct v9fs_inode *v9inode;
61 struct v9fs_session_info *v9ses;
65 p9_debug(P9_DEBUG_VFS, "inode: %p file: %p\n", inode, file);
66 v9inode = V9FS_I(inode);
67 v9ses = v9fs_inode2v9ses(inode);
68 if (v9fs_proto_dotl(v9ses))
69 omode = v9fs_open_to_dotl_flags(file->f_flags);
71 omode = v9fs_uflags2omode(file->f_flags,
72 v9fs_proto_dotu(v9ses));
73 fid = file->private_data;
75 fid = v9fs_fid_clone(file->f_path.dentry);
79 err = p9_client_open(fid, omode);
84 if ((file->f_flags & O_APPEND) &&
85 (!v9fs_proto_dotu(v9ses) && !v9fs_proto_dotl(v9ses)))
86 generic_file_llseek(file, 0, SEEK_END);
89 file->private_data = fid;
90 mutex_lock(&v9inode->v_mutex);
91 if ((v9ses->cache == CACHE_LOOSE || v9ses->cache == CACHE_FSCACHE) &&
92 !v9inode->writeback_fid &&
93 ((file->f_flags & O_ACCMODE) != O_RDONLY)) {
95 * clone a fid and add it to writeback_fid
96 * we do it during open time instead of
97 * page dirty time via write_begin/page_mkwrite
98 * because we want write after unlink usecase
101 fid = v9fs_writeback_fid(file->f_path.dentry);
104 mutex_unlock(&v9inode->v_mutex);
107 v9inode->writeback_fid = (void *) fid;
109 mutex_unlock(&v9inode->v_mutex);
110 if (v9ses->cache == CACHE_LOOSE || v9ses->cache == CACHE_FSCACHE)
111 v9fs_cache_inode_set_cookie(inode, file);
114 p9_client_clunk(file->private_data);
115 file->private_data = NULL;
120 * v9fs_file_lock - lock a file (or directory)
121 * @filp: file to be locked
123 * @fl: file lock structure
125 * Bugs: this looks like a local only lock, we should extend into 9P
126 * by using open exclusive
129 static int v9fs_file_lock(struct file *filp, int cmd, struct file_lock *fl)
132 struct inode *inode = file_inode(filp);
134 p9_debug(P9_DEBUG_VFS, "filp: %p lock: %p\n", filp, fl);
136 /* No mandatory locks */
137 if (__mandatory_lock(inode) && fl->fl_type != F_UNLCK)
140 if ((IS_SETLK(cmd) || IS_SETLKW(cmd)) && fl->fl_type != F_UNLCK) {
141 filemap_write_and_wait(inode->i_mapping);
142 invalidate_mapping_pages(&inode->i_data, 0, -1);
148 static int v9fs_file_do_lock(struct file *filp, int cmd, struct file_lock *fl)
150 struct p9_flock flock;
154 unsigned char fl_type;
156 fid = filp->private_data;
159 if ((fl->fl_flags & FL_POSIX) != FL_POSIX)
162 res = posix_lock_file_wait(filp, fl);
166 /* convert posix lock to p9 tlock args */
167 memset(&flock, 0, sizeof(flock));
168 /* map the lock type */
169 switch (fl->fl_type) {
171 flock.type = P9_LOCK_TYPE_RDLCK;
174 flock.type = P9_LOCK_TYPE_WRLCK;
177 flock.type = P9_LOCK_TYPE_UNLCK;
180 flock.start = fl->fl_start;
181 if (fl->fl_end == OFFSET_MAX)
184 flock.length = fl->fl_end - fl->fl_start + 1;
185 flock.proc_id = fl->fl_pid;
186 flock.client_id = fid->clnt->name;
188 flock.flags = P9_LOCK_FLAGS_BLOCK;
191 * if its a blocked request and we get P9_LOCK_BLOCKED as the status
192 * for lock request, keep on trying
195 res = p9_client_lock_dotl(fid, &flock, &status);
199 if (status != P9_LOCK_BLOCKED)
201 if (status == P9_LOCK_BLOCKED && !IS_SETLKW(cmd))
203 if (schedule_timeout_interruptible(P9_LOCK_TIMEOUT) != 0)
207 /* map 9p status to VFS status */
209 case P9_LOCK_SUCCESS:
212 case P9_LOCK_BLOCKED:
224 * incase server returned error for lock request, revert
227 if (res < 0 && fl->fl_type != F_UNLCK) {
228 fl_type = fl->fl_type;
229 fl->fl_type = F_UNLCK;
230 res = posix_lock_file_wait(filp, fl);
231 fl->fl_type = fl_type;
237 static int v9fs_file_getlock(struct file *filp, struct file_lock *fl)
239 struct p9_getlock glock;
243 fid = filp->private_data;
246 posix_test_lock(filp, fl);
248 * if we have a conflicting lock locally, no need to validate
251 if (fl->fl_type != F_UNLCK)
254 /* convert posix lock to p9 tgetlock args */
255 memset(&glock, 0, sizeof(glock));
256 glock.type = P9_LOCK_TYPE_UNLCK;
257 glock.start = fl->fl_start;
258 if (fl->fl_end == OFFSET_MAX)
261 glock.length = fl->fl_end - fl->fl_start + 1;
262 glock.proc_id = fl->fl_pid;
263 glock.client_id = fid->clnt->name;
265 res = p9_client_getlock_dotl(fid, &glock);
268 /* map 9p lock type to os lock type */
269 switch (glock.type) {
270 case P9_LOCK_TYPE_RDLCK:
271 fl->fl_type = F_RDLCK;
273 case P9_LOCK_TYPE_WRLCK:
274 fl->fl_type = F_WRLCK;
276 case P9_LOCK_TYPE_UNLCK:
277 fl->fl_type = F_UNLCK;
280 if (glock.type != P9_LOCK_TYPE_UNLCK) {
281 fl->fl_start = glock.start;
282 if (glock.length == 0)
283 fl->fl_end = OFFSET_MAX;
285 fl->fl_end = glock.start + glock.length - 1;
286 fl->fl_pid = glock.proc_id;
292 * v9fs_file_lock_dotl - lock a file (or directory)
293 * @filp: file to be locked
295 * @fl: file lock structure
299 static int v9fs_file_lock_dotl(struct file *filp, int cmd, struct file_lock *fl)
301 struct inode *inode = file_inode(filp);
304 p9_debug(P9_DEBUG_VFS, "filp: %p cmd:%d lock: %p name: %s\n",
305 filp, cmd, fl, filp->f_path.dentry->d_name.name);
307 /* No mandatory locks */
308 if (__mandatory_lock(inode) && fl->fl_type != F_UNLCK)
311 if ((IS_SETLK(cmd) || IS_SETLKW(cmd)) && fl->fl_type != F_UNLCK) {
312 filemap_write_and_wait(inode->i_mapping);
313 invalidate_mapping_pages(&inode->i_data, 0, -1);
316 if (IS_SETLK(cmd) || IS_SETLKW(cmd))
317 ret = v9fs_file_do_lock(filp, cmd, fl);
318 else if (IS_GETLK(cmd))
319 ret = v9fs_file_getlock(filp, fl);
327 * v9fs_file_flock_dotl - lock a file
328 * @filp: file to be locked
330 * @fl: file lock structure
334 static int v9fs_file_flock_dotl(struct file *filp, int cmd,
335 struct file_lock *fl)
337 struct inode *inode = file_inode(filp);
340 p9_debug(P9_DEBUG_VFS, "filp: %p cmd:%d lock: %p name: %s\n",
341 filp, cmd, fl, filp->f_path.dentry->d_name.name);
343 /* No mandatory locks */
344 if (__mandatory_lock(inode) && fl->fl_type != F_UNLCK)
347 if (!(fl->fl_flags & FL_FLOCK))
350 if ((IS_SETLK(cmd) || IS_SETLKW(cmd)) && fl->fl_type != F_UNLCK) {
351 filemap_write_and_wait(inode->i_mapping);
352 invalidate_mapping_pages(&inode->i_data, 0, -1);
354 /* Convert flock to posix lock */
355 fl->fl_owner = (fl_owner_t)filp;
357 fl->fl_end = OFFSET_MAX;
358 fl->fl_flags |= FL_POSIX;
359 fl->fl_flags ^= FL_FLOCK;
361 if (IS_SETLK(cmd) | IS_SETLKW(cmd))
362 ret = v9fs_file_do_lock(filp, cmd, fl);
370 * v9fs_fid_readn - read from a fid
372 * @data: data buffer to read data into
373 * @udata: user data buffer to read data into
374 * @count: size of buffer
375 * @offset: offset at which to read data
379 v9fs_fid_readn(struct p9_fid *fid, char *data, char __user *udata, u32 count,
384 p9_debug(P9_DEBUG_VFS, "fid %d offset %llu count %d\n",
385 fid->fid, (long long unsigned)offset, count);
388 size = fid->iounit ? fid->iounit : fid->clnt->msize - P9_IOHDRSZ;
390 n = p9_client_read(fid, data, udata, offset, count);
402 } while (count > 0 && n == size);
411 * v9fs_file_readn - read from a file
412 * @filp: file pointer to read
413 * @data: data buffer to read data into
414 * @udata: user data buffer to read data into
415 * @count: size of buffer
416 * @offset: offset at which to read data
420 v9fs_file_readn(struct file *filp, char *data, char __user *udata, u32 count,
423 return v9fs_fid_readn(filp->private_data, data, udata, count, offset);
427 * v9fs_file_read - read from a file
428 * @filp: file pointer to read
429 * @udata: user data buffer to read data into
430 * @count: size of buffer
431 * @offset: offset at which to read data
436 v9fs_file_read(struct file *filp, char __user *udata, size_t count,
443 p9_debug(P9_DEBUG_VFS, "count %zu offset %lld\n", count, *offset);
444 fid = filp->private_data;
446 size = fid->iounit ? fid->iounit : fid->clnt->msize - P9_IOHDRSZ;
448 ret = v9fs_file_readn(filp, NULL, udata, count, *offset);
450 ret = p9_client_read(fid, NULL, udata, *offset, count);
459 v9fs_file_write_internal(struct inode *inode, struct p9_fid *fid,
460 const char __user *data, size_t count,
461 loff_t *offset, int invalidate)
466 loff_t origin = *offset;
467 unsigned long pg_start, pg_end;
469 p9_debug(P9_DEBUG_VFS, "data %p count %d offset %x\n",
470 data, (int)count, (int)*offset);
473 n = p9_client_write(fid, NULL, data+total, origin+total, count);
480 if (invalidate && (total > 0)) {
481 pg_start = origin >> PAGE_CACHE_SHIFT;
482 pg_end = (origin + total - 1) >> PAGE_CACHE_SHIFT;
483 if (inode->i_mapping && inode->i_mapping->nrpages)
484 invalidate_inode_pages2_range(inode->i_mapping,
487 i_size = i_size_read(inode);
488 if (*offset > i_size) {
489 inode_add_bytes(inode, *offset - i_size);
490 i_size_write(inode, *offset);
500 * v9fs_file_write - write to a file
501 * @filp: file pointer to write
502 * @data: data buffer to write data from
503 * @count: size of buffer
504 * @offset: offset at which to write data
508 v9fs_file_write(struct file *filp, const char __user * data,
509 size_t count, loff_t *offset)
512 loff_t origin = *offset;
515 retval = generic_write_checks(filp, &origin, &count, 0);
520 if ((ssize_t) count < 0)
526 retval = v9fs_file_write_internal(file_inode(filp),
528 data, count, &origin, 1);
529 /* update offset on successful write */
537 static int v9fs_file_fsync(struct file *filp, loff_t start, loff_t end,
541 struct inode *inode = filp->f_mapping->host;
542 struct p9_wstat wstat;
545 retval = filemap_write_and_wait_range(inode->i_mapping, start, end);
549 mutex_lock(&inode->i_mutex);
550 p9_debug(P9_DEBUG_VFS, "filp %p datasync %x\n", filp, datasync);
552 fid = filp->private_data;
553 v9fs_blank_wstat(&wstat);
555 retval = p9_client_wstat(fid, &wstat);
556 mutex_unlock(&inode->i_mutex);
561 int v9fs_file_fsync_dotl(struct file *filp, loff_t start, loff_t end,
565 struct inode *inode = filp->f_mapping->host;
568 retval = filemap_write_and_wait_range(inode->i_mapping, start, end);
572 mutex_lock(&inode->i_mutex);
573 p9_debug(P9_DEBUG_VFS, "filp %p datasync %x\n", filp, datasync);
575 fid = filp->private_data;
577 retval = p9_client_fsync(fid, datasync);
578 mutex_unlock(&inode->i_mutex);
584 v9fs_file_mmap(struct file *filp, struct vm_area_struct *vma)
589 retval = generic_file_mmap(filp, vma);
591 vma->vm_ops = &v9fs_file_vm_ops;
597 v9fs_mmap_file_mmap(struct file *filp, struct vm_area_struct *vma)
601 struct v9fs_inode *v9inode;
604 inode = file_inode(filp);
605 v9inode = V9FS_I(inode);
606 mutex_lock(&v9inode->v_mutex);
607 if (!v9inode->writeback_fid &&
608 (vma->vm_flags & VM_WRITE)) {
610 * clone a fid and add it to writeback_fid
611 * we do it during mmap instead of
612 * page dirty time via write_begin/page_mkwrite
613 * because we want write after unlink usecase
616 fid = v9fs_writeback_fid(filp->f_path.dentry);
618 retval = PTR_ERR(fid);
619 mutex_unlock(&v9inode->v_mutex);
622 v9inode->writeback_fid = (void *) fid;
624 mutex_unlock(&v9inode->v_mutex);
626 retval = generic_file_mmap(filp, vma);
628 vma->vm_ops = &v9fs_mmap_file_vm_ops;
634 v9fs_vm_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
636 struct v9fs_inode *v9inode;
637 struct page *page = vmf->page;
638 struct file *filp = vma->vm_file;
639 struct inode *inode = file_inode(filp);
642 p9_debug(P9_DEBUG_VFS, "page %p fid %lx\n",
643 page, (unsigned long)filp->private_data);
645 /* Update file times before taking page lock */
646 file_update_time(filp);
648 v9inode = V9FS_I(inode);
649 /* make sure the cache has finished storing the page */
650 v9fs_fscache_wait_on_page_write(inode, page);
651 BUG_ON(!v9inode->writeback_fid);
653 if (page->mapping != inode->i_mapping)
655 wait_for_stable_page(page);
657 return VM_FAULT_LOCKED;
660 return VM_FAULT_NOPAGE;
664 v9fs_direct_read(struct file *filp, char __user *udata, size_t count,
669 struct address_space *mapping;
672 mapping = filp->f_mapping;
673 inode = mapping->host;
676 size = i_size_read(inode);
678 filemap_write_and_wait_range(mapping, offset,
681 return v9fs_file_read(filp, udata, count, offsetp);
685 * v9fs_cached_file_read - read from a file
686 * @filp: file pointer to read
687 * @udata: user data buffer to read data into
688 * @count: size of buffer
689 * @offset: offset at which to read data
693 v9fs_cached_file_read(struct file *filp, char __user *data, size_t count,
696 if (filp->f_flags & O_DIRECT)
697 return v9fs_direct_read(filp, data, count, offset);
698 return do_sync_read(filp, data, count, offset);
702 * v9fs_mmap_file_read - read from a file
703 * @filp: file pointer to read
704 * @udata: user data buffer to read data into
705 * @count: size of buffer
706 * @offset: offset at which to read data
710 v9fs_mmap_file_read(struct file *filp, char __user *data, size_t count,
713 /* TODO: Check if there are dirty pages */
714 return v9fs_file_read(filp, data, count, offset);
718 v9fs_direct_write(struct file *filp, const char __user * data,
719 size_t count, loff_t *offsetp)
724 struct address_space *mapping;
727 mapping = filp->f_mapping;
728 inode = mapping->host;
732 mutex_lock(&inode->i_mutex);
733 retval = filemap_write_and_wait_range(mapping, offset,
738 * After a write we want buffered reads to be sure to go to disk to get
739 * the new data. We invalidate clean cached page from the region we're
740 * about to write. We do this *before* the write so that if we fail
741 * here we fall back to buffered write
743 if (mapping->nrpages) {
744 pgoff_t pg_start = offset >> PAGE_CACHE_SHIFT;
745 pgoff_t pg_end = (offset + count - 1) >> PAGE_CACHE_SHIFT;
747 retval = invalidate_inode_pages2_range(mapping,
750 * If a page can not be invalidated, fall back
754 if (retval == -EBUSY)
759 retval = v9fs_file_write(filp, data, count, offsetp);
761 mutex_unlock(&inode->i_mutex);
765 mutex_unlock(&inode->i_mutex);
766 return do_sync_write(filp, data, count, offsetp);
770 * v9fs_cached_file_write - write to a file
771 * @filp: file pointer to write
772 * @data: data buffer to write data from
773 * @count: size of buffer
774 * @offset: offset at which to write data
778 v9fs_cached_file_write(struct file *filp, const char __user * data,
779 size_t count, loff_t *offset)
782 if (filp->f_flags & O_DIRECT)
783 return v9fs_direct_write(filp, data, count, offset);
784 return do_sync_write(filp, data, count, offset);
789 * v9fs_mmap_file_write - write to a file
790 * @filp: file pointer to write
791 * @data: data buffer to write data from
792 * @count: size of buffer
793 * @offset: offset at which to write data
797 v9fs_mmap_file_write(struct file *filp, const char __user *data,
798 size_t count, loff_t *offset)
801 * TODO: invalidate mmaps on filp's inode between
802 * offset and offset+count
804 return v9fs_file_write(filp, data, count, offset);
807 static void v9fs_mmap_vm_close(struct vm_area_struct *vma)
811 struct writeback_control wbc = {
812 .nr_to_write = LONG_MAX,
813 .sync_mode = WB_SYNC_ALL,
814 .range_start = vma->vm_pgoff * PAGE_SIZE,
815 /* absolute end, byte at end included */
816 .range_end = vma->vm_pgoff * PAGE_SIZE +
817 (vma->vm_end - vma->vm_start - 1),
821 p9_debug(P9_DEBUG_VFS, "9p VMA close, %p, flushing", vma);
823 inode = file_inode(vma->vm_file);
825 if (!mapping_cap_writeback_dirty(inode->i_mapping))
829 sync_inode(inode, &wbc);
833 static const struct vm_operations_struct v9fs_file_vm_ops = {
834 .fault = filemap_fault,
835 .map_pages = filemap_map_pages,
836 .page_mkwrite = v9fs_vm_page_mkwrite,
837 .remap_pages = generic_file_remap_pages,
840 static const struct vm_operations_struct v9fs_mmap_file_vm_ops = {
841 .close = v9fs_mmap_vm_close,
842 .fault = filemap_fault,
843 .map_pages = filemap_map_pages,
844 .page_mkwrite = v9fs_vm_page_mkwrite,
845 .remap_pages = generic_file_remap_pages,
849 const struct file_operations v9fs_cached_file_operations = {
850 .llseek = generic_file_llseek,
851 .read = v9fs_cached_file_read,
852 .write = v9fs_cached_file_write,
853 .aio_read = generic_file_aio_read,
854 .aio_write = generic_file_aio_write,
855 .open = v9fs_file_open,
856 .release = v9fs_dir_release,
857 .lock = v9fs_file_lock,
858 .mmap = v9fs_file_mmap,
859 .fsync = v9fs_file_fsync,
862 const struct file_operations v9fs_cached_file_operations_dotl = {
863 .llseek = generic_file_llseek,
864 .read = v9fs_cached_file_read,
865 .write = v9fs_cached_file_write,
866 .aio_read = generic_file_aio_read,
867 .aio_write = generic_file_aio_write,
868 .open = v9fs_file_open,
869 .release = v9fs_dir_release,
870 .lock = v9fs_file_lock_dotl,
871 .flock = v9fs_file_flock_dotl,
872 .mmap = v9fs_file_mmap,
873 .fsync = v9fs_file_fsync_dotl,
876 const struct file_operations v9fs_file_operations = {
877 .llseek = generic_file_llseek,
878 .read = v9fs_file_read,
879 .write = v9fs_file_write,
880 .open = v9fs_file_open,
881 .release = v9fs_dir_release,
882 .lock = v9fs_file_lock,
883 .mmap = generic_file_readonly_mmap,
884 .fsync = v9fs_file_fsync,
887 const struct file_operations v9fs_file_operations_dotl = {
888 .llseek = generic_file_llseek,
889 .read = v9fs_file_read,
890 .write = v9fs_file_write,
891 .open = v9fs_file_open,
892 .release = v9fs_dir_release,
893 .lock = v9fs_file_lock_dotl,
894 .flock = v9fs_file_flock_dotl,
895 .mmap = generic_file_readonly_mmap,
896 .fsync = v9fs_file_fsync_dotl,
899 const struct file_operations v9fs_mmap_file_operations = {
900 .llseek = generic_file_llseek,
901 .read = v9fs_mmap_file_read,
902 .write = v9fs_mmap_file_write,
903 .open = v9fs_file_open,
904 .release = v9fs_dir_release,
905 .lock = v9fs_file_lock,
906 .mmap = v9fs_mmap_file_mmap,
907 .fsync = v9fs_file_fsync,
910 const struct file_operations v9fs_mmap_file_operations_dotl = {
911 .llseek = generic_file_llseek,
912 .read = v9fs_mmap_file_read,
913 .write = v9fs_mmap_file_write,
914 .open = v9fs_file_open,
915 .release = v9fs_dir_release,
916 .lock = v9fs_file_lock_dotl,
917 .flock = v9fs_file_flock_dotl,
918 .mmap = v9fs_mmap_file_mmap,
919 .fsync = v9fs_file_fsync_dotl,