xref: /openbmc/linux/fs/nfs/file.c (revision a2ad63da)
1457c8996SThomas Gleixner // SPDX-License-Identifier: GPL-2.0-only
21da177e4SLinus Torvalds /*
31da177e4SLinus Torvalds  *  linux/fs/nfs/file.c
41da177e4SLinus Torvalds  *
51da177e4SLinus Torvalds  *  Copyright (C) 1992  Rick Sladkey
61da177e4SLinus Torvalds  *
71da177e4SLinus Torvalds  *  Changes Copyright (C) 1994 by Florian La Roche
81da177e4SLinus Torvalds  *   - Do not copy data too often around in the kernel.
91da177e4SLinus Torvalds  *   - In nfs_file_read the return value of kmalloc wasn't checked.
101da177e4SLinus Torvalds  *   - Put in a better version of read look-ahead buffering. Original idea
111da177e4SLinus Torvalds  *     and implementation by Wai S Kok elekokws@ee.nus.sg.
121da177e4SLinus Torvalds  *
131da177e4SLinus Torvalds  *  Expire cache on write to a file by Wai S Kok (Oct 1994).
141da177e4SLinus Torvalds  *
151da177e4SLinus Torvalds  *  Total rewrite of read side for new NFS buffer cache.. Linus.
161da177e4SLinus Torvalds  *
171da177e4SLinus Torvalds  *  nfs regular file handling functions
181da177e4SLinus Torvalds  */
191da177e4SLinus Torvalds 
20ddda8e0aSBryan Schumaker #include <linux/module.h>
211da177e4SLinus Torvalds #include <linux/time.h>
221da177e4SLinus Torvalds #include <linux/kernel.h>
231da177e4SLinus Torvalds #include <linux/errno.h>
241da177e4SLinus Torvalds #include <linux/fcntl.h>
251da177e4SLinus Torvalds #include <linux/stat.h>
261da177e4SLinus Torvalds #include <linux/nfs_fs.h>
271da177e4SLinus Torvalds #include <linux/nfs_mount.h>
281da177e4SLinus Torvalds #include <linux/mm.h>
291da177e4SLinus Torvalds #include <linux/pagemap.h>
305a0e3ad6STejun Heo #include <linux/gfp.h>
31b608b283STrond Myklebust #include <linux/swap.h>
321da177e4SLinus Torvalds 
337c0f6ba6SLinus Torvalds #include <linux/uaccess.h>
341da177e4SLinus Torvalds 
351da177e4SLinus Torvalds #include "delegation.h"
3694387fb1STrond Myklebust #include "internal.h"
3791d5b470SChuck Lever #include "iostat.h"
38545db45fSDavid Howells #include "fscache.h"
39612aa983SChristoph Hellwig #include "pnfs.h"
401da177e4SLinus Torvalds 
41f4ce1299STrond Myklebust #include "nfstrace.h"
42f4ce1299STrond Myklebust 
431da177e4SLinus Torvalds #define NFSDBG_FACILITY		NFSDBG_FILE
441da177e4SLinus Torvalds 
45f0f37e2fSAlexey Dobriyan static const struct vm_operations_struct nfs_file_vm_ops;
4694387fb1STrond Myklebust 
47ce4ef7c0SBryan Schumaker int nfs_check_flags(int flags)
481da177e4SLinus Torvalds {
491da177e4SLinus Torvalds 	if ((flags & (O_APPEND | O_DIRECT)) == (O_APPEND | O_DIRECT))
501da177e4SLinus Torvalds 		return -EINVAL;
511da177e4SLinus Torvalds 
521da177e4SLinus Torvalds 	return 0;
531da177e4SLinus Torvalds }
5489d77c8fSBryan Schumaker EXPORT_SYMBOL_GPL(nfs_check_flags);
551da177e4SLinus Torvalds 
561da177e4SLinus Torvalds /*
571da177e4SLinus Torvalds  * Open file
581da177e4SLinus Torvalds  */
591da177e4SLinus Torvalds static int
601da177e4SLinus Torvalds nfs_file_open(struct inode *inode, struct file *filp)
611da177e4SLinus Torvalds {
621da177e4SLinus Torvalds 	int res;
631da177e4SLinus Torvalds 
646de1472fSAl Viro 	dprintk("NFS: open file(%pD2)\n", filp);
65cc0dd2d1SChuck Lever 
66c2459dc4SChuck Lever 	nfs_inc_stats(inode, NFSIOS_VFSOPEN);
671da177e4SLinus Torvalds 	res = nfs_check_flags(filp->f_flags);
681da177e4SLinus Torvalds 	if (res)
691da177e4SLinus Torvalds 		return res;
701da177e4SLinus Torvalds 
7146cb650cSTrond Myklebust 	res = nfs_open(inode, filp);
72*a2ad63daSNeilBrown 	if (res == 0)
73*a2ad63daSNeilBrown 		filp->f_mode |= FMODE_CAN_ODIRECT;
741da177e4SLinus Torvalds 	return res;
751da177e4SLinus Torvalds }
761da177e4SLinus Torvalds 
77ce4ef7c0SBryan Schumaker int
781da177e4SLinus Torvalds nfs_file_release(struct inode *inode, struct file *filp)
791da177e4SLinus Torvalds {
806de1472fSAl Viro 	dprintk("NFS: release(%pD2)\n", filp);
816da24bc9SChuck Lever 
8291d5b470SChuck Lever 	nfs_inc_stats(inode, NFSIOS_VFSRELEASE);
83aff8d8dcSAnna Schumaker 	nfs_file_clear_open_context(filp);
84a6b5a28eSDave Wysochanski 	nfs_fscache_release_file(inode, filp);
85aff8d8dcSAnna Schumaker 	return 0;
861da177e4SLinus Torvalds }
8789d77c8fSBryan Schumaker EXPORT_SYMBOL_GPL(nfs_file_release);
881da177e4SLinus Torvalds 
89980802e3STrond Myklebust /**
9037eaeed1STrond Myklebust  * nfs_revalidate_file_size - Revalidate the file size
91302fad7bSTrond Myklebust  * @inode: pointer to inode struct
92302fad7bSTrond Myklebust  * @filp: pointer to struct file
93980802e3STrond Myklebust  *
94980802e3STrond Myklebust  * Revalidates the file length. This is basically a wrapper around
95980802e3STrond Myklebust  * nfs_revalidate_inode() that takes into account the fact that we may
96980802e3STrond Myklebust  * have cached writes (in which case we don't care about the server's
97980802e3STrond Myklebust  * idea of what the file length is), or O_DIRECT (in which case we
98980802e3STrond Myklebust  * shouldn't trust the cache).
99980802e3STrond Myklebust  */
100980802e3STrond Myklebust static int nfs_revalidate_file_size(struct inode *inode, struct file *filp)
101980802e3STrond Myklebust {
102980802e3STrond Myklebust 	struct nfs_server *server = NFS_SERVER(inode);
103d7cf8dd0STrond Myklebust 
104980802e3STrond Myklebust 	if (filp->f_flags & O_DIRECT)
105980802e3STrond Myklebust 		goto force_reval;
10613c0b082STrond Myklebust 	if (nfs_check_cache_invalid(inode, NFS_INO_INVALID_SIZE))
107d7cf8dd0STrond Myklebust 		goto force_reval;
108fe51beecSTrond Myklebust 	return 0;
109980802e3STrond Myklebust force_reval:
110980802e3STrond Myklebust 	return __nfs_revalidate_inode(server, inode);
111980802e3STrond Myklebust }
112980802e3STrond Myklebust 
113965c8e59SAndrew Morton loff_t nfs_file_llseek(struct file *filp, loff_t offset, int whence)
114980802e3STrond Myklebust {
1156de1472fSAl Viro 	dprintk("NFS: llseek file(%pD2, %lld, %d)\n",
1166de1472fSAl Viro 			filp, offset, whence);
117b84e06c5SChuck Lever 
11806222e49SJosef Bacik 	/*
119965c8e59SAndrew Morton 	 * whence == SEEK_END || SEEK_DATA || SEEK_HOLE => we must revalidate
12006222e49SJosef Bacik 	 * the cached file length
12106222e49SJosef Bacik 	 */
122965c8e59SAndrew Morton 	if (whence != SEEK_SET && whence != SEEK_CUR) {
123980802e3STrond Myklebust 		struct inode *inode = filp->f_mapping->host;
124d5e66348STrond Myklebust 
125980802e3STrond Myklebust 		int retval = nfs_revalidate_file_size(inode, filp);
126980802e3STrond Myklebust 		if (retval < 0)
127980802e3STrond Myklebust 			return (loff_t)retval;
12879835a71SAndi Kleen 	}
129d5e66348STrond Myklebust 
130965c8e59SAndrew Morton 	return generic_file_llseek(filp, offset, whence);
131980802e3STrond Myklebust }
13289d77c8fSBryan Schumaker EXPORT_SYMBOL_GPL(nfs_file_llseek);
133980802e3STrond Myklebust 
1341da177e4SLinus Torvalds /*
1351da177e4SLinus Torvalds  * Flush all dirty pages, and check for write errors.
1361da177e4SLinus Torvalds  */
1375445b1fbSTrond Myklebust static int
13875e1fcc0SMiklos Szeredi nfs_file_flush(struct file *file, fl_owner_t id)
1391da177e4SLinus Torvalds {
1406de1472fSAl Viro 	struct inode	*inode = file_inode(file);
14167dd23f9SScott Mayhew 	errseq_t since;
1421da177e4SLinus Torvalds 
1436de1472fSAl Viro 	dprintk("NFS: flush(%pD2)\n", file);
1441da177e4SLinus Torvalds 
145c2459dc4SChuck Lever 	nfs_inc_stats(inode, NFSIOS_VFSFLUSH);
1461da177e4SLinus Torvalds 	if ((file->f_mode & FMODE_WRITE) == 0)
1471da177e4SLinus Torvalds 		return 0;
1487b159fc1STrond Myklebust 
1497fe5c398STrond Myklebust 	/* Flush writes to the server and return any errors */
15067dd23f9SScott Mayhew 	since = filemap_sample_wb_err(file->f_mapping);
15167dd23f9SScott Mayhew 	nfs_wb_all(inode);
15267dd23f9SScott Mayhew 	return filemap_check_wb_err(file->f_mapping, since);
1531da177e4SLinus Torvalds }
1541da177e4SLinus Torvalds 
155ce4ef7c0SBryan Schumaker ssize_t
1563aa2d199SAl Viro nfs_file_read(struct kiocb *iocb, struct iov_iter *to)
1571da177e4SLinus Torvalds {
1586de1472fSAl Viro 	struct inode *inode = file_inode(iocb->ki_filp);
1591da177e4SLinus Torvalds 	ssize_t result;
1601da177e4SLinus Torvalds 
1612ba48ce5SAl Viro 	if (iocb->ki_flags & IOCB_DIRECT)
16264158668SNeilBrown 		return nfs_file_direct_read(iocb, to, false);
1631da177e4SLinus Torvalds 
164619d30b4SAl Viro 	dprintk("NFS: read(%pD2, %zu@%lu)\n",
1656de1472fSAl Viro 		iocb->ki_filp,
1663aa2d199SAl Viro 		iov_iter_count(to), (unsigned long) iocb->ki_pos);
1671da177e4SLinus Torvalds 
168a5864c99STrond Myklebust 	nfs_start_io_read(inode);
169a5864c99STrond Myklebust 	result = nfs_revalidate_mapping(inode, iocb->ki_filp->f_mapping);
1704184dcf2SChuck Lever 	if (!result) {
1713aa2d199SAl Viro 		result = generic_file_read_iter(iocb, to);
1724184dcf2SChuck Lever 		if (result > 0)
1734184dcf2SChuck Lever 			nfs_add_stats(inode, NFSIOS_NORMALREADBYTES, result);
1744184dcf2SChuck Lever 	}
175a5864c99STrond Myklebust 	nfs_end_io_read(inode);
1761da177e4SLinus Torvalds 	return result;
1771da177e4SLinus Torvalds }
17889d77c8fSBryan Schumaker EXPORT_SYMBOL_GPL(nfs_file_read);
1791da177e4SLinus Torvalds 
180ce4ef7c0SBryan Schumaker int
1811da177e4SLinus Torvalds nfs_file_mmap(struct file * file, struct vm_area_struct * vma)
1821da177e4SLinus Torvalds {
1836de1472fSAl Viro 	struct inode *inode = file_inode(file);
1841da177e4SLinus Torvalds 	int	status;
1851da177e4SLinus Torvalds 
1866de1472fSAl Viro 	dprintk("NFS: mmap(%pD2)\n", file);
1871da177e4SLinus Torvalds 
188e1ebfd33STrond Myklebust 	/* Note: generic_file_mmap() returns ENOSYS on nommu systems
189e1ebfd33STrond Myklebust 	 *       so we call that before revalidating the mapping
190e1ebfd33STrond Myklebust 	 */
191e1ebfd33STrond Myklebust 	status = generic_file_mmap(file, vma);
19294387fb1STrond Myklebust 	if (!status) {
19394387fb1STrond Myklebust 		vma->vm_ops = &nfs_file_vm_ops;
194e1ebfd33STrond Myklebust 		status = nfs_revalidate_mapping(inode, file->f_mapping);
19594387fb1STrond Myklebust 	}
1961da177e4SLinus Torvalds 	return status;
1971da177e4SLinus Torvalds }
19889d77c8fSBryan Schumaker EXPORT_SYMBOL_GPL(nfs_file_mmap);
1991da177e4SLinus Torvalds 
2001da177e4SLinus Torvalds /*
2011da177e4SLinus Torvalds  * Flush any dirty pages for this process, and check for write errors.
2021da177e4SLinus Torvalds  * The return status from this call provides a reliable indication of
2031da177e4SLinus Torvalds  * whether any write errors occurred for this process.
2041da177e4SLinus Torvalds  */
2054ff79bc7SChristoph Hellwig static int
206bf4b4905SNeilBrown nfs_file_fsync_commit(struct file *file, int datasync)
2071da177e4SLinus Torvalds {
2086de1472fSAl Viro 	struct inode *inode = file_inode(file);
2092197e9b0STrond Myklebust 	int ret;
210af7fa165STrond Myklebust 
2116de1472fSAl Viro 	dprintk("NFS: fsync file(%pD2) datasync %d\n", file, datasync);
2121da177e4SLinus Torvalds 
21391d5b470SChuck Lever 	nfs_inc_stats(inode, NFSIOS_VFSFSYNC);
2142197e9b0STrond Myklebust 	ret = nfs_commit_inode(inode, FLUSH_SYNC);
2152197e9b0STrond Myklebust 	if (ret < 0)
216a5c58892SBryan Schumaker 		return ret;
2172197e9b0STrond Myklebust 	return file_check_and_advance_wb_err(file);
218a5c58892SBryan Schumaker }
219a5c58892SBryan Schumaker 
2204ff79bc7SChristoph Hellwig int
221a5c58892SBryan Schumaker nfs_file_fsync(struct file *file, loff_t start, loff_t end, int datasync)
222a5c58892SBryan Schumaker {
2232197e9b0STrond Myklebust 	struct nfs_open_context *ctx = nfs_file_open_context(file);
224496ad9aaSAl Viro 	struct inode *inode = file_inode(file);
2252197e9b0STrond Myklebust 	int ret;
226a5c58892SBryan Schumaker 
227f4ce1299STrond Myklebust 	trace_nfs_fsync_enter(inode);
228f4ce1299STrond Myklebust 
2292197e9b0STrond Myklebust 	for (;;) {
2306fbda89bSTrond Myklebust 		ret = file_write_and_wait_range(file, start, end);
2317b281ee0STrond Myklebust 		if (ret != 0)
23205990d1bSTrond Myklebust 			break;
233bf4b4905SNeilBrown 		ret = nfs_file_fsync_commit(file, datasync);
2342197e9b0STrond Myklebust 		if (ret != 0)
2352197e9b0STrond Myklebust 			break;
2364ff79bc7SChristoph Hellwig 		ret = pnfs_sync_inode(inode, !!datasync);
2372197e9b0STrond Myklebust 		if (ret != 0)
2382197e9b0STrond Myklebust 			break;
2392197e9b0STrond Myklebust 		if (!test_and_clear_bit(NFS_CONTEXT_RESEND_WRITES, &ctx->flags))
2402197e9b0STrond Myklebust 			break;
241dcfc4f25STrond Myklebust 		/*
242dcfc4f25STrond Myklebust 		 * If nfs_file_fsync_commit detected a server reboot, then
243dcfc4f25STrond Myklebust 		 * resend all dirty pages that might have been covered by
244dcfc4f25STrond Myklebust 		 * the NFS_CONTEXT_RESEND_WRITES flag
245dcfc4f25STrond Myklebust 		 */
246dcfc4f25STrond Myklebust 		start = 0;
247dcfc4f25STrond Myklebust 		end = LLONG_MAX;
2482197e9b0STrond Myklebust 	}
24905990d1bSTrond Myklebust 
250f4ce1299STrond Myklebust 	trace_nfs_fsync_exit(inode, ret);
251af7fa165STrond Myklebust 	return ret;
2521da177e4SLinus Torvalds }
2534ff79bc7SChristoph Hellwig EXPORT_SYMBOL_GPL(nfs_file_fsync);
2541da177e4SLinus Torvalds 
2551da177e4SLinus Torvalds /*
25638c73044SPeter Staubach  * Decide whether a read/modify/write cycle may be more efficient
25738c73044SPeter Staubach  * then a modify/write/read cycle when writing to a page in the
25838c73044SPeter Staubach  * page cache.
25938c73044SPeter Staubach  *
2602cde04e9SKazuo Ito  * Some pNFS layout drivers can only read/write at a certain block
2612cde04e9SKazuo Ito  * granularity like all block devices and therefore we must perform
2622cde04e9SKazuo Ito  * read/modify/write whenever a page hasn't read yet and the data
2632cde04e9SKazuo Ito  * to be written there is not aligned to a block boundary and/or
2642cde04e9SKazuo Ito  * smaller than the block size.
2652cde04e9SKazuo Ito  *
26638c73044SPeter Staubach  * The modify/write/read cycle may occur if a page is read before
26738c73044SPeter Staubach  * being completely filled by the writer.  In this situation, the
26838c73044SPeter Staubach  * page must be completely written to stable storage on the server
26938c73044SPeter Staubach  * before it can be refilled by reading in the page from the server.
27038c73044SPeter Staubach  * This can lead to expensive, small, FILE_SYNC mode writes being
27138c73044SPeter Staubach  * done.
27238c73044SPeter Staubach  *
27338c73044SPeter Staubach  * It may be more efficient to read the page first if the file is
27438c73044SPeter Staubach  * open for reading in addition to writing, the page is not marked
27538c73044SPeter Staubach  * as Uptodate, it is not dirty or waiting to be committed,
27638c73044SPeter Staubach  * indicating that it was previously allocated and then modified,
27738c73044SPeter Staubach  * that there were valid bytes of data in that range of the file,
27838c73044SPeter Staubach  * and that the new data won't completely replace the old data in
27938c73044SPeter Staubach  * that range of the file.
28038c73044SPeter Staubach  */
2812cde04e9SKazuo Ito static bool nfs_full_page_write(struct page *page, loff_t pos, unsigned int len)
28238c73044SPeter Staubach {
28338c73044SPeter Staubach 	unsigned int pglen = nfs_page_length(page);
28409cbfeafSKirill A. Shutemov 	unsigned int offset = pos & (PAGE_SIZE - 1);
28538c73044SPeter Staubach 	unsigned int end = offset + len;
28638c73044SPeter Staubach 
2872cde04e9SKazuo Ito 	return !pglen || (end >= pglen && !offset);
288612aa983SChristoph Hellwig }
289612aa983SChristoph Hellwig 
2902cde04e9SKazuo Ito static bool nfs_want_read_modify_write(struct file *file, struct page *page,
2912cde04e9SKazuo Ito 			loff_t pos, unsigned int len)
2922cde04e9SKazuo Ito {
2932cde04e9SKazuo Ito 	/*
2942cde04e9SKazuo Ito 	 * Up-to-date pages, those with ongoing or full-page write
2952cde04e9SKazuo Ito 	 * don't need read/modify/write
2962cde04e9SKazuo Ito 	 */
2972cde04e9SKazuo Ito 	if (PageUptodate(page) || PagePrivate(page) ||
2982cde04e9SKazuo Ito 	    nfs_full_page_write(page, pos, len))
2992cde04e9SKazuo Ito 		return false;
3002cde04e9SKazuo Ito 
3012cde04e9SKazuo Ito 	if (pnfs_ld_read_whole_page(file->f_mapping->host))
3022cde04e9SKazuo Ito 		return true;
3032cde04e9SKazuo Ito 	/* Open for reading too? */
3042cde04e9SKazuo Ito 	if (file->f_mode & FMODE_READ)
3052cde04e9SKazuo Ito 		return true;
3062cde04e9SKazuo Ito 	return false;
30738c73044SPeter Staubach }
30838c73044SPeter Staubach 
30938c73044SPeter Staubach /*
3104899f9c8SNick Piggin  * This does the "real" work of the write. We must allocate and lock the
3114899f9c8SNick Piggin  * page to be sent back to the generic routine, which then copies the
3124899f9c8SNick Piggin  * data from user space.
3131da177e4SLinus Torvalds  *
3141da177e4SLinus Torvalds  * If the writer ends up delaying the write, the writer needs to
3151da177e4SLinus Torvalds  * increment the page use counts until he is done with the page.
3161da177e4SLinus Torvalds  */
3174899f9c8SNick Piggin static int nfs_write_begin(struct file *file, struct address_space *mapping,
3184899f9c8SNick Piggin 			loff_t pos, unsigned len, unsigned flags,
3194899f9c8SNick Piggin 			struct page **pagep, void **fsdata)
3201da177e4SLinus Torvalds {
3214899f9c8SNick Piggin 	int ret;
32209cbfeafSKirill A. Shutemov 	pgoff_t index = pos >> PAGE_SHIFT;
3234899f9c8SNick Piggin 	struct page *page;
32438c73044SPeter Staubach 	int once_thru = 0;
3254899f9c8SNick Piggin 
3261e8968c5SNiels de Vos 	dfprintk(PAGECACHE, "NFS: write_begin(%pD2(%lu), %u@%lld)\n",
3276de1472fSAl Viro 		file, mapping->host->i_ino, len, (long long) pos);
328b7eaefaaSChuck Lever 
32938c73044SPeter Staubach start:
33054566b2cSNick Piggin 	page = grab_cache_page_write_begin(mapping, index, flags);
3314899f9c8SNick Piggin 	if (!page)
3324899f9c8SNick Piggin 		return -ENOMEM;
3334899f9c8SNick Piggin 	*pagep = page;
3344899f9c8SNick Piggin 
3354899f9c8SNick Piggin 	ret = nfs_flush_incompatible(file, page);
3364899f9c8SNick Piggin 	if (ret) {
3374899f9c8SNick Piggin 		unlock_page(page);
33809cbfeafSKirill A. Shutemov 		put_page(page);
33938c73044SPeter Staubach 	} else if (!once_thru &&
34038c73044SPeter Staubach 		   nfs_want_read_modify_write(file, page, pos, len)) {
34138c73044SPeter Staubach 		once_thru = 1;
34238c73044SPeter Staubach 		ret = nfs_readpage(file, page);
34309cbfeafSKirill A. Shutemov 		put_page(page);
34438c73044SPeter Staubach 		if (!ret)
34538c73044SPeter Staubach 			goto start;
3464899f9c8SNick Piggin 	}
3474899f9c8SNick Piggin 	return ret;
3481da177e4SLinus Torvalds }
3491da177e4SLinus Torvalds 
3504899f9c8SNick Piggin static int nfs_write_end(struct file *file, struct address_space *mapping,
3514899f9c8SNick Piggin 			loff_t pos, unsigned len, unsigned copied,
3524899f9c8SNick Piggin 			struct page *page, void *fsdata)
3531da177e4SLinus Torvalds {
35409cbfeafSKirill A. Shutemov 	unsigned offset = pos & (PAGE_SIZE - 1);
355dc24826bSAndy Adamson 	struct nfs_open_context *ctx = nfs_file_open_context(file);
3564899f9c8SNick Piggin 	int status;
3571da177e4SLinus Torvalds 
3581e8968c5SNiels de Vos 	dfprintk(PAGECACHE, "NFS: write_end(%pD2(%lu), %u@%lld)\n",
3596de1472fSAl Viro 		file, mapping->host->i_ino, len, (long long) pos);
360b7eaefaaSChuck Lever 
361efc91ed0STrond Myklebust 	/*
362efc91ed0STrond Myklebust 	 * Zero any uninitialised parts of the page, and then mark the page
363efc91ed0STrond Myklebust 	 * as up to date if it turns out that we're extending the file.
364efc91ed0STrond Myklebust 	 */
365efc91ed0STrond Myklebust 	if (!PageUptodate(page)) {
366efc91ed0STrond Myklebust 		unsigned pglen = nfs_page_length(page);
367c0cf3ef5SAl Viro 		unsigned end = offset + copied;
368efc91ed0STrond Myklebust 
369efc91ed0STrond Myklebust 		if (pglen == 0) {
370efc91ed0STrond Myklebust 			zero_user_segments(page, 0, offset,
37109cbfeafSKirill A. Shutemov 					end, PAGE_SIZE);
372efc91ed0STrond Myklebust 			SetPageUptodate(page);
373efc91ed0STrond Myklebust 		} else if (end >= pglen) {
37409cbfeafSKirill A. Shutemov 			zero_user_segment(page, end, PAGE_SIZE);
375efc91ed0STrond Myklebust 			if (offset == 0)
376efc91ed0STrond Myklebust 				SetPageUptodate(page);
377efc91ed0STrond Myklebust 		} else
37809cbfeafSKirill A. Shutemov 			zero_user_segment(page, pglen, PAGE_SIZE);
379efc91ed0STrond Myklebust 	}
380efc91ed0STrond Myklebust 
3814899f9c8SNick Piggin 	status = nfs_updatepage(file, page, offset, copied);
3824899f9c8SNick Piggin 
3834899f9c8SNick Piggin 	unlock_page(page);
38409cbfeafSKirill A. Shutemov 	put_page(page);
3854899f9c8SNick Piggin 
3863d509e54SChuck Lever 	if (status < 0)
3873d509e54SChuck Lever 		return status;
3882701d086SAndy Adamson 	NFS_I(mapping->host)->write_io += copied;
389dc24826bSAndy Adamson 
390ce52914eSScott Mayhew 	if (nfs_ctx_key_to_expire(ctx, mapping->host)) {
391dc24826bSAndy Adamson 		status = nfs_wb_all(mapping->host);
392dc24826bSAndy Adamson 		if (status < 0)
393dc24826bSAndy Adamson 			return status;
394dc24826bSAndy Adamson 	}
395dc24826bSAndy Adamson 
3963d509e54SChuck Lever 	return copied;
3971da177e4SLinus Torvalds }
3981da177e4SLinus Torvalds 
3996b9b3514SDavid Howells /*
4006b9b3514SDavid Howells  * Partially or wholly invalidate a page
4016b9b3514SDavid Howells  * - Release the private state associated with a page if undergoing complete
4026b9b3514SDavid Howells  *   page invalidation
403545db45fSDavid Howells  * - Called if either PG_private or PG_fscache is set on the page
4046b9b3514SDavid Howells  * - Caller holds page lock
4056b9b3514SDavid Howells  */
4066d740c76SMatthew Wilcox (Oracle) static void nfs_invalidate_folio(struct folio *folio, size_t offset,
4076d740c76SMatthew Wilcox (Oracle) 				size_t length)
408cd52ed35STrond Myklebust {
4096d740c76SMatthew Wilcox (Oracle) 	dfprintk(PAGECACHE, "NFS: invalidate_folio(%lu, %zu, %zu)\n",
4106d740c76SMatthew Wilcox (Oracle) 		 folio->index, offset, length);
411b7eaefaaSChuck Lever 
4126d740c76SMatthew Wilcox (Oracle) 	if (offset != 0 || length < folio_size(folio))
4131c75950bSTrond Myklebust 		return;
414d2ccddf0STrond Myklebust 	/* Cancel any unstarted writes on this page */
4156d740c76SMatthew Wilcox (Oracle) 	nfs_wb_folio_cancel(folio->mapping->host, folio);
4166d740c76SMatthew Wilcox (Oracle) 	folio_wait_fscache(folio);
417cd52ed35STrond Myklebust }
418cd52ed35STrond Myklebust 
4196b9b3514SDavid Howells /*
4206b9b3514SDavid Howells  * Attempt to release the private state associated with a page
421545db45fSDavid Howells  * - Called if either PG_private or PG_fscache is set on the page
4226b9b3514SDavid Howells  * - Caller holds page lock
4236b9b3514SDavid Howells  * - Return true (may release page) or false (may not)
4246b9b3514SDavid Howells  */
425cd52ed35STrond Myklebust static int nfs_release_page(struct page *page, gfp_t gfp)
426cd52ed35STrond Myklebust {
427b7eaefaaSChuck Lever 	dfprintk(PAGECACHE, "NFS: release_page(%p)\n", page);
428b7eaefaaSChuck Lever 
429e3db7691STrond Myklebust 	/* If PagePrivate() is set, then the page is not freeable */
430545db45fSDavid Howells 	if (PagePrivate(page))
431ddeff520SNikita Danilov 		return 0;
432545db45fSDavid Howells 	return nfs_fscache_release_page(page, gfp);
433e3db7691STrond Myklebust }
434e3db7691STrond Myklebust 
435f919b196SMel Gorman static void nfs_check_dirty_writeback(struct page *page,
436f919b196SMel Gorman 				bool *dirty, bool *writeback)
437f919b196SMel Gorman {
438f919b196SMel Gorman 	struct nfs_inode *nfsi;
439f919b196SMel Gorman 	struct address_space *mapping = page_file_mapping(page);
440f919b196SMel Gorman 
441f919b196SMel Gorman 	if (!mapping || PageSwapCache(page))
442f919b196SMel Gorman 		return;
443f919b196SMel Gorman 
444f919b196SMel Gorman 	/*
445f919b196SMel Gorman 	 * Check if an unstable page is currently being committed and
446f919b196SMel Gorman 	 * if so, have the VM treat it as if the page is under writeback
447f919b196SMel Gorman 	 * so it will not block due to pages that will shortly be freeable.
448f919b196SMel Gorman 	 */
449f919b196SMel Gorman 	nfsi = NFS_I(mapping->host);
450af7cf057STrond Myklebust 	if (atomic_read(&nfsi->commit_info.rpcs_out)) {
451f919b196SMel Gorman 		*writeback = true;
452f919b196SMel Gorman 		return;
453f919b196SMel Gorman 	}
454f919b196SMel Gorman 
455f919b196SMel Gorman 	/*
456f919b196SMel Gorman 	 * If PagePrivate() is set, then the page is not freeable and as the
457f919b196SMel Gorman 	 * inode is not being committed, it's not going to be cleaned in the
458f919b196SMel Gorman 	 * near future so treat it as dirty
459f919b196SMel Gorman 	 */
460f919b196SMel Gorman 	if (PagePrivate(page))
461f919b196SMel Gorman 		*dirty = true;
462f919b196SMel Gorman }
463f919b196SMel Gorman 
4646b9b3514SDavid Howells /*
4656b9b3514SDavid Howells  * Attempt to clear the private state associated with a page when an error
4666b9b3514SDavid Howells  * occurs that requires the cached contents of an inode to be written back or
4676b9b3514SDavid Howells  * destroyed
468545db45fSDavid Howells  * - Called if either PG_private or fscache is set on the page
4696b9b3514SDavid Howells  * - Caller holds page lock
4706b9b3514SDavid Howells  * - Return 0 if successful, -error otherwise
4716b9b3514SDavid Howells  */
47215a30ab2SMatthew Wilcox (Oracle) static int nfs_launder_folio(struct folio *folio)
473e3db7691STrond Myklebust {
47415a30ab2SMatthew Wilcox (Oracle) 	struct inode *inode = folio->mapping->host;
475b7eaefaaSChuck Lever 
47615a30ab2SMatthew Wilcox (Oracle) 	dfprintk(PAGECACHE, "NFS: launder_folio(%ld, %llu)\n",
47715a30ab2SMatthew Wilcox (Oracle) 		inode->i_ino, folio_pos(folio));
478b7eaefaaSChuck Lever 
47915a30ab2SMatthew Wilcox (Oracle) 	folio_wait_fscache(folio);
48015a30ab2SMatthew Wilcox (Oracle) 	return nfs_wb_page(inode, &folio->page);
481cd52ed35STrond Myklebust }
482cd52ed35STrond Myklebust 
483a564b8f0SMel Gorman static int nfs_swap_activate(struct swap_info_struct *sis, struct file *file,
484a564b8f0SMel Gorman 						sector_t *span)
485a564b8f0SMel Gorman {
486bd89bc67SMurphy Zhou 	unsigned long blocks;
487bd89bc67SMurphy Zhou 	long long isize;
4884b60c0ffSNeilBrown 	int ret;
4894dc73c67SNeilBrown 	struct inode *inode = file_inode(file);
4904dc73c67SNeilBrown 	struct rpc_clnt *clnt = NFS_CLIENT(inode);
4914dc73c67SNeilBrown 	struct nfs_client *cl = NFS_SERVER(inode)->nfs_client;
492bd89bc67SMurphy Zhou 
493bd89bc67SMurphy Zhou 	spin_lock(&inode->i_lock);
494bd89bc67SMurphy Zhou 	blocks = inode->i_blocks;
495bd89bc67SMurphy Zhou 	isize = inode->i_size;
496bd89bc67SMurphy Zhou 	spin_unlock(&inode->i_lock);
497bd89bc67SMurphy Zhou 	if (blocks*512 < isize) {
498bd89bc67SMurphy Zhou 		pr_warn("swap activate: swapfile has holes\n");
499bd89bc67SMurphy Zhou 		return -EINVAL;
500bd89bc67SMurphy Zhou 	}
501dad2b015SJeff Layton 
5024b60c0ffSNeilBrown 	ret = rpc_clnt_swap_activate(clnt);
5034b60c0ffSNeilBrown 	if (ret)
5044b60c0ffSNeilBrown 		return ret;
5054b60c0ffSNeilBrown 	ret = add_swap_extent(sis, 0, sis->max, 0);
5064b60c0ffSNeilBrown 	if (ret < 0) {
5074b60c0ffSNeilBrown 		rpc_clnt_swap_deactivate(clnt);
5084b60c0ffSNeilBrown 		return ret;
5094b60c0ffSNeilBrown 	}
510dad2b015SJeff Layton 
5114b60c0ffSNeilBrown 	*span = sis->pages;
5124dc73c67SNeilBrown 
5134dc73c67SNeilBrown 	if (cl->rpc_ops->enable_swap)
5144dc73c67SNeilBrown 		cl->rpc_ops->enable_swap(inode);
5154dc73c67SNeilBrown 
5164b60c0ffSNeilBrown 	sis->flags |= SWP_FS_OPS;
5174b60c0ffSNeilBrown 	return ret;
518a564b8f0SMel Gorman }
519a564b8f0SMel Gorman 
520a564b8f0SMel Gorman static void nfs_swap_deactivate(struct file *file)
521a564b8f0SMel Gorman {
5224dc73c67SNeilBrown 	struct inode *inode = file_inode(file);
5234dc73c67SNeilBrown 	struct rpc_clnt *clnt = NFS_CLIENT(inode);
5244dc73c67SNeilBrown 	struct nfs_client *cl = NFS_SERVER(inode)->nfs_client;
525dad2b015SJeff Layton 
5263c87ef6eSJeff Layton 	rpc_clnt_swap_deactivate(clnt);
5274dc73c67SNeilBrown 	if (cl->rpc_ops->disable_swap)
5284dc73c67SNeilBrown 		cl->rpc_ops->disable_swap(file_inode(file));
529a564b8f0SMel Gorman }
530a564b8f0SMel Gorman 
531f5e54d6eSChristoph Hellwig const struct address_space_operations nfs_file_aops = {
5321da177e4SLinus Torvalds 	.readpage = nfs_readpage,
5338786fde8SMatthew Wilcox (Oracle) 	.readahead = nfs_readahead,
534187c82cbSMatthew Wilcox (Oracle) 	.dirty_folio = filemap_dirty_folio,
5351da177e4SLinus Torvalds 	.writepage = nfs_writepage,
5361da177e4SLinus Torvalds 	.writepages = nfs_writepages,
5374899f9c8SNick Piggin 	.write_begin = nfs_write_begin,
5384899f9c8SNick Piggin 	.write_end = nfs_write_end,
5396d740c76SMatthew Wilcox (Oracle) 	.invalidate_folio = nfs_invalidate_folio,
540cd52ed35STrond Myklebust 	.releasepage = nfs_release_page,
541f844cd0dSChao Yu #ifdef CONFIG_MIGRATION
542074cc1deSTrond Myklebust 	.migratepage = nfs_migrate_page,
543f844cd0dSChao Yu #endif
54415a30ab2SMatthew Wilcox (Oracle) 	.launder_folio = nfs_launder_folio,
545f919b196SMel Gorman 	.is_dirty_writeback = nfs_check_dirty_writeback,
546f590f333SAndi Kleen 	.error_remove_page = generic_error_remove_page,
547a564b8f0SMel Gorman 	.swap_activate = nfs_swap_activate,
548a564b8f0SMel Gorman 	.swap_deactivate = nfs_swap_deactivate,
549eb79f3afSNeilBrown 	.swap_rw = nfs_swap_rw,
5501da177e4SLinus Torvalds };
5511da177e4SLinus Torvalds 
5526b9b3514SDavid Howells /*
5536b9b3514SDavid Howells  * Notification that a PTE pointing to an NFS page is about to be made
5546b9b3514SDavid Howells  * writable, implying that someone is about to modify the page through a
5556b9b3514SDavid Howells  * shared-writable mapping
5566b9b3514SDavid Howells  */
55701a36844SSouptick Joarder static vm_fault_t nfs_vm_page_mkwrite(struct vm_fault *vmf)
55894387fb1STrond Myklebust {
559c2ec175cSNick Piggin 	struct page *page = vmf->page;
56011bac800SDave Jiang 	struct file *filp = vmf->vma->vm_file;
5616de1472fSAl Viro 	struct inode *inode = file_inode(filp);
56294387fb1STrond Myklebust 	unsigned pagelen;
56301a36844SSouptick Joarder 	vm_fault_t ret = VM_FAULT_NOPAGE;
5644899f9c8SNick Piggin 	struct address_space *mapping;
56594387fb1STrond Myklebust 
5661e8968c5SNiels de Vos 	dfprintk(PAGECACHE, "NFS: vm_page_mkwrite(%pD2(%lu), offset %lld)\n",
5676de1472fSAl Viro 		filp, filp->f_mapping->host->i_ino,
568b7eaefaaSChuck Lever 		(long long)page_offset(page));
569b7eaefaaSChuck Lever 
5709a773e7cSTrond Myklebust 	sb_start_pagefault(inode->i_sb);
5719a773e7cSTrond Myklebust 
572545db45fSDavid Howells 	/* make sure the cache has finished storing the page */
573a6b5a28eSDave Wysochanski 	if (PageFsCache(page) &&
574a6b5a28eSDave Wysochanski 	    wait_on_page_fscache_killable(vmf->page) < 0) {
575a6b5a28eSDave Wysochanski 		ret = VM_FAULT_RETRY;
576a6b5a28eSDave Wysochanski 		goto out;
577a6b5a28eSDave Wysochanski 	}
578545db45fSDavid Howells 
579ef070dcbSTrond Myklebust 	wait_on_bit_action(&NFS_I(inode)->flags, NFS_INO_INVALIDATING,
580ef070dcbSTrond Myklebust 			nfs_wait_bit_killable, TASK_KILLABLE);
581ef070dcbSTrond Myklebust 
58294387fb1STrond Myklebust 	lock_page(page);
583d56b4ddfSMel Gorman 	mapping = page_file_mapping(page);
5846de1472fSAl Viro 	if (mapping != inode->i_mapping)
5858b1f9ee5STrond Myklebust 		goto out_unlock;
5868b1f9ee5STrond Myklebust 
5872aeb98f4STrond Myklebust 	wait_on_page_writeback(page);
5882aeb98f4STrond Myklebust 
5894899f9c8SNick Piggin 	pagelen = nfs_page_length(page);
5908b1f9ee5STrond Myklebust 	if (pagelen == 0)
5918b1f9ee5STrond Myklebust 		goto out_unlock;
5928b1f9ee5STrond Myklebust 
593bc4866b6STrond Myklebust 	ret = VM_FAULT_LOCKED;
594bc4866b6STrond Myklebust 	if (nfs_flush_incompatible(filp, page) == 0 &&
595bc4866b6STrond Myklebust 	    nfs_updatepage(filp, page, 0, pagelen) == 0)
596bc4866b6STrond Myklebust 		goto out;
5978b1f9ee5STrond Myklebust 
598bc4866b6STrond Myklebust 	ret = VM_FAULT_SIGBUS;
5998b1f9ee5STrond Myklebust out_unlock:
6004899f9c8SNick Piggin 	unlock_page(page);
601bc4866b6STrond Myklebust out:
6029a773e7cSTrond Myklebust 	sb_end_pagefault(inode->i_sb);
603bc4866b6STrond Myklebust 	return ret;
60494387fb1STrond Myklebust }
60594387fb1STrond Myklebust 
606f0f37e2fSAlexey Dobriyan static const struct vm_operations_struct nfs_file_vm_ops = {
60794387fb1STrond Myklebust 	.fault = filemap_fault,
608f1820361SKirill A. Shutemov 	.map_pages = filemap_map_pages,
60994387fb1STrond Myklebust 	.page_mkwrite = nfs_vm_page_mkwrite,
61094387fb1STrond Myklebust };
61194387fb1STrond Myklebust 
612ce368536SScott Mayhew static int nfs_need_check_write(struct file *filp, struct inode *inode,
613ce368536SScott Mayhew 				int error)
6147b159fc1STrond Myklebust {
6157b159fc1STrond Myklebust 	struct nfs_open_context *ctx;
6167b159fc1STrond Myklebust 
617cd3758e3STrond Myklebust 	ctx = nfs_file_open_context(filp);
618ce368536SScott Mayhew 	if (nfs_error_is_fatal_on_server(error) ||
619ce368536SScott Mayhew 	    nfs_ctx_key_to_expire(ctx, inode))
6207b159fc1STrond Myklebust 		return 1;
6217b159fc1STrond Myklebust 	return 0;
6227b159fc1STrond Myklebust }
6237b159fc1STrond Myklebust 
624edaf4369SAl Viro ssize_t nfs_file_write(struct kiocb *iocb, struct iov_iter *from)
6251da177e4SLinus Torvalds {
6266de1472fSAl Viro 	struct file *file = iocb->ki_filp;
6276de1472fSAl Viro 	struct inode *inode = file_inode(file);
628ed7bcdb3STrond Myklebust 	unsigned int mntflags = NFS_SERVER(inode)->flags;
629ed7bcdb3STrond Myklebust 	ssize_t result, written;
630ce368536SScott Mayhew 	errseq_t since;
631ce368536SScott Mayhew 	int error;
6321da177e4SLinus Torvalds 
6336de1472fSAl Viro 	result = nfs_key_timeout_notify(file, inode);
634dc24826bSAndy Adamson 	if (result)
635dc24826bSAndy Adamson 		return result;
636dc24826bSAndy Adamson 
63789698b24STrond Myklebust 	if (iocb->ki_flags & IOCB_DIRECT)
63864158668SNeilBrown 		return nfs_file_direct_write(iocb, from, false);
6391da177e4SLinus Torvalds 
640619d30b4SAl Viro 	dprintk("NFS: write(%pD2, %zu@%Ld)\n",
64118290650STrond Myklebust 		file, iov_iter_count(from), (long long) iocb->ki_pos);
6421da177e4SLinus Torvalds 
6431da177e4SLinus Torvalds 	if (IS_SWAPFILE(inode))
6441da177e4SLinus Torvalds 		goto out_swapfile;
6457d52e862STrond Myklebust 	/*
6467d52e862STrond Myklebust 	 * O_APPEND implies that we must revalidate the file length.
6477d52e862STrond Myklebust 	 */
648fc9dc401STrond Myklebust 	if (iocb->ki_flags & IOCB_APPEND || iocb->ki_pos > i_size_read(inode)) {
6496de1472fSAl Viro 		result = nfs_revalidate_file_size(inode, file);
6501da177e4SLinus Torvalds 		if (result)
6511da177e4SLinus Torvalds 			goto out;
652fe51beecSTrond Myklebust 	}
6531da177e4SLinus Torvalds 
65428aa2f9eSTrond Myklebust 	nfs_clear_invalid_mapping(file->f_mapping);
65528aa2f9eSTrond Myklebust 
656ce368536SScott Mayhew 	since = filemap_sample_wb_err(file->f_mapping);
657a5864c99STrond Myklebust 	nfs_start_io_write(inode);
65818290650STrond Myklebust 	result = generic_write_checks(iocb, from);
65918290650STrond Myklebust 	if (result > 0) {
66018290650STrond Myklebust 		current->backing_dev_info = inode_to_bdi(inode);
661800ba295SMatthew Wilcox (Oracle) 		result = generic_perform_write(iocb, from);
66218290650STrond Myklebust 		current->backing_dev_info = NULL;
66318290650STrond Myklebust 	}
664a5864c99STrond Myklebust 	nfs_end_io_write(inode);
66518290650STrond Myklebust 	if (result <= 0)
6661da177e4SLinus Torvalds 		goto out;
6671da177e4SLinus Torvalds 
668c49edecdSTrond Myklebust 	written = result;
66918290650STrond Myklebust 	iocb->ki_pos += written;
670ed7bcdb3STrond Myklebust 
671ed7bcdb3STrond Myklebust 	if (mntflags & NFS_MOUNT_WRITE_EAGER) {
672ed7bcdb3STrond Myklebust 		result = filemap_fdatawrite_range(file->f_mapping,
673ed7bcdb3STrond Myklebust 						  iocb->ki_pos - written,
674ed7bcdb3STrond Myklebust 						  iocb->ki_pos - 1);
675ed7bcdb3STrond Myklebust 		if (result < 0)
676ed7bcdb3STrond Myklebust 			goto out;
677ed7bcdb3STrond Myklebust 	}
678ed7bcdb3STrond Myklebust 	if (mntflags & NFS_MOUNT_WRITE_WAIT) {
679ed7bcdb3STrond Myklebust 		result = filemap_fdatawait_range(file->f_mapping,
680ed7bcdb3STrond Myklebust 						 iocb->ki_pos - written,
681ed7bcdb3STrond Myklebust 						 iocb->ki_pos - 1);
682ed7bcdb3STrond Myklebust 		if (result < 0)
683ed7bcdb3STrond Myklebust 			goto out;
684ed7bcdb3STrond Myklebust 	}
685e973b1a5Starangg@amazon.com 	result = generic_write_sync(iocb, written);
686e973b1a5Starangg@amazon.com 	if (result < 0)
687e973b1a5Starangg@amazon.com 		goto out;
6887e381172SChuck Lever 
6897e94d6c4STrond Myklebust 	/* Return error values */
690ce368536SScott Mayhew 	error = filemap_check_wb_err(file->f_mapping, since);
691ce368536SScott Mayhew 	if (nfs_need_check_write(file, inode, error)) {
692aded8d7bSTrond Myklebust 		int err = nfs_wb_all(inode);
693200baa21STrond Myklebust 		if (err < 0)
694200baa21STrond Myklebust 			result = err;
695200baa21STrond Myklebust 	}
6967e381172SChuck Lever 	nfs_add_stats(inode, NFSIOS_NORMALWRITTENBYTES, written);
6971da177e4SLinus Torvalds out:
6981da177e4SLinus Torvalds 	return result;
6991da177e4SLinus Torvalds 
7001da177e4SLinus Torvalds out_swapfile:
7011da177e4SLinus Torvalds 	printk(KERN_INFO "NFS: attempt to write to active swap file!\n");
70289658c4dSAnna Schumaker 	return -ETXTBSY;
7031da177e4SLinus Torvalds }
70489d77c8fSBryan Schumaker EXPORT_SYMBOL_GPL(nfs_file_write);
7051da177e4SLinus Torvalds 
7065eebde23SSuresh Jayaraman static int
7075eebde23SSuresh Jayaraman do_getlk(struct file *filp, int cmd, struct file_lock *fl, int is_local)
7081da177e4SLinus Torvalds {
7091da177e4SLinus Torvalds 	struct inode *inode = filp->f_mapping->host;
7101da177e4SLinus Torvalds 	int status = 0;
71121ac19d4SSergey Vlasov 	unsigned int saved_type = fl->fl_type;
7121da177e4SLinus Torvalds 
713039c4d7aSTrond Myklebust 	/* Try local locking first */
7146d34ac19SJ. Bruce Fields 	posix_test_lock(filp, fl);
7156d34ac19SJ. Bruce Fields 	if (fl->fl_type != F_UNLCK) {
7166d34ac19SJ. Bruce Fields 		/* found a conflict */
717039c4d7aSTrond Myklebust 		goto out;
7181da177e4SLinus Torvalds 	}
71921ac19d4SSergey Vlasov 	fl->fl_type = saved_type;
720039c4d7aSTrond Myklebust 
721011e2a7fSBryan Schumaker 	if (NFS_PROTO(inode)->have_delegation(inode, FMODE_READ))
722039c4d7aSTrond Myklebust 		goto out_noconflict;
723039c4d7aSTrond Myklebust 
7245eebde23SSuresh Jayaraman 	if (is_local)
725039c4d7aSTrond Myklebust 		goto out_noconflict;
726039c4d7aSTrond Myklebust 
727039c4d7aSTrond Myklebust 	status = NFS_PROTO(inode)->lock(filp, cmd, fl);
728039c4d7aSTrond Myklebust out:
7291da177e4SLinus Torvalds 	return status;
730039c4d7aSTrond Myklebust out_noconflict:
731039c4d7aSTrond Myklebust 	fl->fl_type = F_UNLCK;
732039c4d7aSTrond Myklebust 	goto out;
7331da177e4SLinus Torvalds }
7341da177e4SLinus Torvalds 
7355eebde23SSuresh Jayaraman static int
7365eebde23SSuresh Jayaraman do_unlk(struct file *filp, int cmd, struct file_lock *fl, int is_local)
7371da177e4SLinus Torvalds {
7381da177e4SLinus Torvalds 	struct inode *inode = filp->f_mapping->host;
7397a8203d8STrond Myklebust 	struct nfs_lock_context *l_ctx;
7401da177e4SLinus Torvalds 	int status;
7411da177e4SLinus Torvalds 
7421da177e4SLinus Torvalds 	/*
7431da177e4SLinus Torvalds 	 * Flush all pending writes before doing anything
7441da177e4SLinus Torvalds 	 * with locks..
7451da177e4SLinus Torvalds 	 */
746aded8d7bSTrond Myklebust 	nfs_wb_all(inode);
7471da177e4SLinus Torvalds 
7487a8203d8STrond Myklebust 	l_ctx = nfs_get_lock_context(nfs_file_open_context(filp));
7497a8203d8STrond Myklebust 	if (!IS_ERR(l_ctx)) {
750210c7c17SBenjamin Coddington 		status = nfs_iocounter_wait(l_ctx);
7517a8203d8STrond Myklebust 		nfs_put_lock_context(l_ctx);
7521da177e4SLinus Torvalds 		/*  NOTE: special case
7531da177e4SLinus Torvalds 		 * 	If we're signalled while cleaning up locks on process exit, we
7541da177e4SLinus Torvalds 		 * 	still need to complete the unlock.
7551da177e4SLinus Torvalds 		 */
756f30cb757SBenjamin Coddington 		if (status < 0 && !(fl->fl_flags & FL_CLOSE))
757f30cb757SBenjamin Coddington 			return status;
758f30cb757SBenjamin Coddington 	}
759f30cb757SBenjamin Coddington 
7605eebde23SSuresh Jayaraman 	/*
7615eebde23SSuresh Jayaraman 	 * Use local locking if mounted with "-onolock" or with appropriate
7625eebde23SSuresh Jayaraman 	 * "-olocal_lock="
7635eebde23SSuresh Jayaraman 	 */
7645eebde23SSuresh Jayaraman 	if (!is_local)
7651da177e4SLinus Torvalds 		status = NFS_PROTO(inode)->lock(filp, cmd, fl);
7661da177e4SLinus Torvalds 	else
76775575ddfSJeff Layton 		status = locks_lock_file_wait(filp, fl);
7681da177e4SLinus Torvalds 	return status;
7691da177e4SLinus Torvalds }
7701da177e4SLinus Torvalds 
7715eebde23SSuresh Jayaraman static int
7725eebde23SSuresh Jayaraman do_setlk(struct file *filp, int cmd, struct file_lock *fl, int is_local)
7731da177e4SLinus Torvalds {
7741da177e4SLinus Torvalds 	struct inode *inode = filp->f_mapping->host;
7751da177e4SLinus Torvalds 	int status;
7761da177e4SLinus Torvalds 
7771da177e4SLinus Torvalds 	/*
7781da177e4SLinus Torvalds 	 * Flush all pending writes before doing anything
7791da177e4SLinus Torvalds 	 * with locks..
7801da177e4SLinus Torvalds 	 */
78129884df0STrond Myklebust 	status = nfs_sync_mapping(filp->f_mapping);
78229884df0STrond Myklebust 	if (status != 0)
7831da177e4SLinus Torvalds 		goto out;
7841da177e4SLinus Torvalds 
7855eebde23SSuresh Jayaraman 	/*
7865eebde23SSuresh Jayaraman 	 * Use local locking if mounted with "-onolock" or with appropriate
7875eebde23SSuresh Jayaraman 	 * "-olocal_lock="
7885eebde23SSuresh Jayaraman 	 */
7895eebde23SSuresh Jayaraman 	if (!is_local)
7901da177e4SLinus Torvalds 		status = NFS_PROTO(inode)->lock(filp, cmd, fl);
791c4d7c402STrond Myklebust 	else
79275575ddfSJeff Layton 		status = locks_lock_file_wait(filp, fl);
7931da177e4SLinus Torvalds 	if (status < 0)
7941da177e4SLinus Torvalds 		goto out;
7956b96724eSRicardo Labiaga 
7961da177e4SLinus Torvalds 	/*
797779eafabSNeilBrown 	 * Invalidate cache to prevent missing any changes.  If
798779eafabSNeilBrown 	 * the file is mapped, clear the page cache as well so
799779eafabSNeilBrown 	 * those mappings will be loaded.
8006b96724eSRicardo Labiaga 	 *
8011da177e4SLinus Torvalds 	 * This makes locking act as a cache coherency point.
8021da177e4SLinus Torvalds 	 */
80329884df0STrond Myklebust 	nfs_sync_mapping(filp->f_mapping);
804779eafabSNeilBrown 	if (!NFS_PROTO(inode)->have_delegation(inode, FMODE_READ)) {
805442ce049SNeilBrown 		nfs_zap_caches(inode);
806779eafabSNeilBrown 		if (mapping_mapped(filp->f_mapping))
807779eafabSNeilBrown 			nfs_revalidate_mapping(inode, filp->f_mapping);
808779eafabSNeilBrown 	}
8091da177e4SLinus Torvalds out:
8101da177e4SLinus Torvalds 	return status;
8111da177e4SLinus Torvalds }
8121da177e4SLinus Torvalds 
8131da177e4SLinus Torvalds /*
8141da177e4SLinus Torvalds  * Lock a (portion of) a file
8151da177e4SLinus Torvalds  */
816ce4ef7c0SBryan Schumaker int nfs_lock(struct file *filp, int cmd, struct file_lock *fl)
8171da177e4SLinus Torvalds {
8181da177e4SLinus Torvalds 	struct inode *inode = filp->f_mapping->host;
8192116271aSTrond Myklebust 	int ret = -ENOLCK;
8205eebde23SSuresh Jayaraman 	int is_local = 0;
8211da177e4SLinus Torvalds 
8226de1472fSAl Viro 	dprintk("NFS: lock(%pD2, t=%x, fl=%x, r=%lld:%lld)\n",
8236de1472fSAl Viro 			filp, fl->fl_type, fl->fl_flags,
8241da177e4SLinus Torvalds 			(long long)fl->fl_start, (long long)fl->fl_end);
8256da24bc9SChuck Lever 
82691d5b470SChuck Lever 	nfs_inc_stats(inode, NFSIOS_VFSLOCK);
8271da177e4SLinus Torvalds 
828bb0a55bbSJ. Bruce Fields 	if (fl->fl_flags & FL_RECLAIM)
829bb0a55bbSJ. Bruce Fields 		return -ENOGRACE;
830bb0a55bbSJ. Bruce Fields 
8315eebde23SSuresh Jayaraman 	if (NFS_SERVER(inode)->flags & NFS_MOUNT_LOCAL_FCNTL)
8325eebde23SSuresh Jayaraman 		is_local = 1;
8335eebde23SSuresh Jayaraman 
8342116271aSTrond Myklebust 	if (NFS_PROTO(inode)->lock_check_bounds != NULL) {
8352116271aSTrond Myklebust 		ret = NFS_PROTO(inode)->lock_check_bounds(fl);
8362116271aSTrond Myklebust 		if (ret < 0)
8372116271aSTrond Myklebust 			goto out_err;
8382116271aSTrond Myklebust 	}
8391da177e4SLinus Torvalds 
8401da177e4SLinus Torvalds 	if (IS_GETLK(cmd))
8415eebde23SSuresh Jayaraman 		ret = do_getlk(filp, cmd, fl, is_local);
8422116271aSTrond Myklebust 	else if (fl->fl_type == F_UNLCK)
8435eebde23SSuresh Jayaraman 		ret = do_unlk(filp, cmd, fl, is_local);
8442116271aSTrond Myklebust 	else
8455eebde23SSuresh Jayaraman 		ret = do_setlk(filp, cmd, fl, is_local);
8462116271aSTrond Myklebust out_err:
8472116271aSTrond Myklebust 	return ret;
8481da177e4SLinus Torvalds }
84989d77c8fSBryan Schumaker EXPORT_SYMBOL_GPL(nfs_lock);
8501da177e4SLinus Torvalds 
8511da177e4SLinus Torvalds /*
8521da177e4SLinus Torvalds  * Lock a (portion of) a file
8531da177e4SLinus Torvalds  */
854ce4ef7c0SBryan Schumaker int nfs_flock(struct file *filp, int cmd, struct file_lock *fl)
8551da177e4SLinus Torvalds {
8565eebde23SSuresh Jayaraman 	struct inode *inode = filp->f_mapping->host;
8575eebde23SSuresh Jayaraman 	int is_local = 0;
8585eebde23SSuresh Jayaraman 
8596de1472fSAl Viro 	dprintk("NFS: flock(%pD2, t=%x, fl=%x)\n",
8606de1472fSAl Viro 			filp, fl->fl_type, fl->fl_flags);
8611da177e4SLinus Torvalds 
8621da177e4SLinus Torvalds 	if (!(fl->fl_flags & FL_FLOCK))
8631da177e4SLinus Torvalds 		return -ENOLCK;
8641da177e4SLinus Torvalds 
8655eebde23SSuresh Jayaraman 	if (NFS_SERVER(inode)->flags & NFS_MOUNT_LOCAL_FLOCK)
8665eebde23SSuresh Jayaraman 		is_local = 1;
8675eebde23SSuresh Jayaraman 
868fcfa4470SBenjamin Coddington 	/* We're simulating flock() locks using posix locks on the server */
869fcfa4470SBenjamin Coddington 	if (fl->fl_type == F_UNLCK)
8705eebde23SSuresh Jayaraman 		return do_unlk(filp, cmd, fl, is_local);
8715eebde23SSuresh Jayaraman 	return do_setlk(filp, cmd, fl, is_local);
8721da177e4SLinus Torvalds }
87389d77c8fSBryan Schumaker EXPORT_SYMBOL_GPL(nfs_flock);
874370f6599SJ. Bruce Fields 
8750486958fSJeff Layton const struct file_operations nfs_file_operations = {
8760486958fSJeff Layton 	.llseek		= nfs_file_llseek,
8773aa2d199SAl Viro 	.read_iter	= nfs_file_read,
878edaf4369SAl Viro 	.write_iter	= nfs_file_write,
8790486958fSJeff Layton 	.mmap		= nfs_file_mmap,
8800486958fSJeff Layton 	.open		= nfs_file_open,
8810486958fSJeff Layton 	.flush		= nfs_file_flush,
8820486958fSJeff Layton 	.release	= nfs_file_release,
8830486958fSJeff Layton 	.fsync		= nfs_file_fsync,
8840486958fSJeff Layton 	.lock		= nfs_lock,
8850486958fSJeff Layton 	.flock		= nfs_flock,
88682c156f8SAl Viro 	.splice_read	= generic_file_splice_read,
8874da54c21SAl Viro 	.splice_write	= iter_file_splice_write,
8880486958fSJeff Layton 	.check_flags	= nfs_check_flags,
8891c994a09SJeff Layton 	.setlease	= simple_nosetlease,
8900486958fSJeff Layton };
891ddda8e0aSBryan Schumaker EXPORT_SYMBOL_GPL(nfs_file_operations);
892