~hc/RK356X_SDK_RELEASE.git

..	..	@@ -33,6 +33,7 @@
33	33	#include <linux/mount.h>
34	34	#include <linux/slab.h>
35	35	#include <linux/swap.h>
	36	+#include <linux/mm.h>
36	37	#include <asm/div64.h>
37	38	#include "cifsfs.h"
38	39	#include "cifspdu.h"
..	..	@@ -222,9 +223,6 @@
222	223	if (!buf)
223	224	return -ENOMEM;
224	225
225		- if (backup_cred(cifs_sb))
226		- create_options \|= CREATE_OPEN_BACKUP_INTENT;
227		-
228	226	/* O_SYNC also has bit for O_DSYNC so following check picks up either */
229	227	if (f_flags & O_SYNC)
230	228	create_options \|= CREATE_WRITE_THROUGH;
..	..	@@ -235,7 +233,7 @@
235	233	oparms.tcon = tcon;
236	234	oparms.cifs_sb = cifs_sb;
237	235	oparms.desired_access = desired_access;
238		- oparms.create_options = create_options;
	236	+ oparms.create_options = cifs_create_options(cifs_sb, create_options);
239	237	oparms.disposition = disposition;
240	238	oparms.path = full_path;
241	239	oparms.fid = fid;
..	..	@@ -246,6 +244,7 @@
246	244	if (rc)
247	245	goto out;
248	246
	247	+ /* TODO: Add support for calling posix query info but with passing in fid */
249	248	if (tcon->unix_ext)
250	249	rc = cifs_get_inode_info_unix(&inode, full_path, inode->i_sb,
251	250	xid);
..	..	@@ -288,6 +287,8 @@
288	287	msleep(10);
289	288	}
290	289
	290	+static void cifsFileInfo_put_work(struct work_struct *work);
	291	+
291	292	struct cifsFileInfo *
292	293	cifs_new_fileinfo(struct cifs_fid fid, struct file file,
293	294	struct tcon_link *tlink, __u32 oplock)
..	..	@@ -322,6 +323,7 @@
322	323	cfile->invalidHandle = false;
323	324	cfile->tlink = cifs_get_tlink(tlink);
324	325	INIT_WORK(&cfile->oplock_break, cifs_oplock_break);
	326	+ INIT_WORK(&cfile->put, cifsFileInfo_put_work);
325	327	mutex_init(&cfile->fh_mutex);
326	328	spin_lock_init(&cfile->file_info_lock);
327	329
..	..	@@ -349,6 +351,7 @@
349	351	server->ops->set_fid(cfile, fid, oplock);
350	352
351	353	list_add(&cfile->tlist, &tcon->openFileList);
	354	+ atomic_inc(&tcon->num_local_opens);
352	355
353	356	/* if readable file instance put first in list*/
354	357	spin_lock(&cinode->open_file_lock);
..	..	@@ -375,6 +378,41 @@
375	378	return cifs_file;
376	379	}
377	380
	381	+static void cifsFileInfo_put_final(struct cifsFileInfo *cifs_file)
	382	+{
	383	+ struct inode *inode = d_inode(cifs_file->dentry);
	384	+ struct cifsInodeInfo *cifsi = CIFS_I(inode);
	385	+ struct cifsLockInfo li, tmp;
	386	+ struct super_block *sb = inode->i_sb;
	387	+
	388	+ /*
	389	+ * Delete any outstanding lock records. We'll lose them when the file
	390	+ * is closed anyway.
	391	+ */
	392	+ cifs_down_write(&cifsi->lock_sem);
	393	+ list_for_each_entry_safe(li, tmp, &cifs_file->llist->locks, llist) {
	394	+ list_del(&li->llist);
	395	+ cifs_del_lock_waiters(li);
	396	+ kfree(li);
	397	+ }
	398	+ list_del(&cifs_file->llist->llist);
	399	+ kfree(cifs_file->llist);
	400	+ up_write(&cifsi->lock_sem);
	401	+
	402	+ cifs_put_tlink(cifs_file->tlink);
	403	+ dput(cifs_file->dentry);
	404	+ cifs_sb_deactive(sb);
	405	+ kfree(cifs_file);
	406	+}
	407	+
	408	+static void cifsFileInfo_put_work(struct work_struct *work)
	409	+{
	410	+ struct cifsFileInfo *cifs_file = container_of(work,
	411	+ struct cifsFileInfo, put);
	412	+
	413	+ cifsFileInfo_put_final(cifs_file);
	414	+}
	415	+
378	416	/**
379	417	* cifsFileInfo_put - release a reference of file priv data
380	418	*
..	..	@@ -382,15 +420,15 @@
382	420	*/
383	421	void cifsFileInfo_put(struct cifsFileInfo *cifs_file)
384	422	{
385		- _cifsFileInfo_put(cifs_file, true);
	423	+ _cifsFileInfo_put(cifs_file, true, true);
386	424	}
387	425
388	426	/**
389	427	* _cifsFileInfo_put - release a reference of file priv data
390	428	*
391	429	* This may involve closing the filehandle @cifs_file out on the
392		- * server. Must be called without holding tcon->open_file_lock and
393		- * cifs_file->file_info_lock.
	430	+ * server. Must be called without holding tcon->open_file_lock,
	431	+ * cinode->open_file_lock and cifs_file->file_info_lock.
394	432	*
395	433	* If @wait_for_oplock_handler is true and we are releasing the last
396	434	* reference, wait for any running oplock break handler of the file
..	..	@@ -398,7 +436,8 @@
398	436	* oplock break handler, you need to pass false.
399	437	*
400	438	*/
401		-void _cifsFileInfo_put(struct cifsFileInfo *cifs_file, bool wait_oplock_handler)
	439	+void _cifsFileInfo_put(struct cifsFileInfo *cifs_file,
	440	+ bool wait_oplock_handler, bool offload)
402	441	{
403	442	struct inode *inode = d_inode(cifs_file->dentry);
404	443	struct cifs_tcon *tcon = tlink_tcon(cifs_file->tlink);
..	..	@@ -406,7 +445,6 @@
406	445	struct cifsInodeInfo *cifsi = CIFS_I(inode);
407	446	struct super_block *sb = inode->i_sb;
408	447	struct cifs_sb_info *cifs_sb = CIFS_SB(sb);
409		- struct cifsLockInfo li, tmp;
410	448	struct cifs_fid fid;
411	449	struct cifs_pending_open open;
412	450	bool oplock_break_cancelled;
..	..	@@ -431,6 +469,7 @@
431	469	/* remove it from the lists */
432	470	list_del(&cifs_file->flist);
433	471	list_del(&cifs_file->tlist);
	472	+ atomic_dec(&tcon->num_local_opens);
434	473
435	474	if (list_empty(&cifsi->openFileList)) {
436	475	cifs_dbg(FYI, "closing last open instance for inode %p\n",
..	..	@@ -456,7 +495,9 @@
456	495	unsigned int xid;
457	496
458	497	xid = get_xid();
459		- if (server->ops->close)
	498	+ if (server->ops->close_getattr)
	499	+ server->ops->close_getattr(xid, tcon, cifs_file);
	500	+ else if (server->ops->close)
460	501	server->ops->close(xid, tcon, &cifs_file->fid);
461	502	_free_xid(xid);
462	503	}
..	..	@@ -466,24 +507,10 @@
466	507
467	508	cifs_del_pending_open(&open);
468	509
469		- /*
470		- * Delete any outstanding lock records. We'll lose them when the file
471		- * is closed anyway.
472		- */
473		- cifs_down_write(&cifsi->lock_sem);
474		- list_for_each_entry_safe(li, tmp, &cifs_file->llist->locks, llist) {
475		- list_del(&li->llist);
476		- cifs_del_lock_waiters(li);
477		- kfree(li);
478		- }
479		- list_del(&cifs_file->llist->llist);
480		- kfree(cifs_file->llist);
481		- up_write(&cifsi->lock_sem);
482		-
483		- cifs_put_tlink(cifs_file->tlink);
484		- dput(cifs_file->dentry);
485		- cifs_sb_deactive(sb);
486		- kfree(cifs_file);
	510	+ if (offload)
	511	+ queue_work(fileinfo_put_wq, &cifs_file->put);
	512	+ else
	513	+ cifsFileInfo_put_final(cifs_file);
487	514	}
488	515
489	516	int cifs_open(struct inode inode, struct file file)
..	..	@@ -724,9 +751,6 @@
724	751
725	752	desired_access = cifs_convert_flags(cfile->f_flags);
726	753
727		- if (backup_cred(cifs_sb))
728		- create_options \|= CREATE_OPEN_BACKUP_INTENT;
729		-
730	754	/* O_SYNC also has bit for O_DSYNC so following check picks up either */
731	755	if (cfile->f_flags & O_SYNC)
732	756	create_options \|= CREATE_WRITE_THROUGH;
..	..	@@ -740,7 +764,7 @@
740	764	oparms.tcon = tcon;
741	765	oparms.cifs_sb = cifs_sb;
742	766	oparms.desired_access = desired_access;
743		- oparms.create_options = create_options;
	767	+ oparms.create_options = cifs_create_options(cifs_sb, create_options);
744	768	oparms.disposition = disposition;
745	769	oparms.path = full_path;
746	770	oparms.fid = &cfile->fid;
..	..	@@ -778,7 +802,9 @@
778	802	if (!is_interrupt_error(rc))
779	803	mapping_set_error(inode->i_mapping, rc);
780	804
781		- if (tcon->unix_ext)
	805	+ if (tcon->posix_extensions)
	806	+ rc = smb311_posix_get_inode_info(&inode, full_path, inode->i_sb, xid);
	807	+ else if (tcon->unix_ext)
782	808	rc = cifs_get_inode_info_unix(&inode, full_path,
783	809	inode->i_sb, xid);
784	810	else
..	..	@@ -814,7 +840,7 @@
814	840	int cifs_close(struct inode inode, struct file file)
815	841	{
816	842	if (file->private_data != NULL) {
817		- cifsFileInfo_put(file->private_data);
	843	+ _cifsFileInfo_put(file->private_data, true, false);
818	844	file->private_data = NULL;
819	845	}
820	846
..	..	@@ -835,7 +861,7 @@
835	861
836	862	tcon->need_reopen_files = false;
837	863
838		- cifs_dbg(FYI, "Reopen persistent handles");
	864	+ cifs_dbg(FYI, "Reopen persistent handles\n");
839	865	INIT_LIST_HEAD(&tmp_list);
840	866
841	867	/* list all files open on tree connection, reopen resilient handles */
..	..	@@ -910,7 +936,7 @@
910	936	}
911	937
912	938	static struct cifsLockInfo *
913		-cifs_lock_init(__u64 offset, __u64 length, __u8 type)
	939	+cifs_lock_init(__u64 offset, __u64 length, __u8 type, __u16 flags)
914	940	{
915	941	struct cifsLockInfo *lock =
916	942	kmalloc(sizeof(struct cifsLockInfo), GFP_KERNEL);
..	..	@@ -920,6 +946,7 @@
920	946	lock->length = length;
921	947	lock->type = type;
922	948	lock->pid = current->tgid;
	949	+ lock->flags = flags;
923	950	INIT_LIST_HEAD(&lock->blist);
924	951	init_waitqueue_head(&lock->block_q);
925	952	return lock;
..	..	@@ -942,7 +969,8 @@
942	969	/* @rw_check : 0 - no op, 1 - read, 2 - write */
943	970	static bool
944	971	cifs_find_fid_lock_conflict(struct cifs_fid_locks *fdlocks, __u64 offset,
945		- __u64 length, __u8 type, struct cifsFileInfo *cfile,
	972	+ __u64 length, __u8 type, __u16 flags,
	973	+ struct cifsFileInfo *cfile,
946	974	struct cifsLockInfo **conf_lock, int rw_check)
947	975	{
948	976	struct cifsLockInfo *li;
..	..	@@ -964,6 +992,10 @@
964	992	((server->ops->compare_fids(cfile, cur_cfile) &&
965	993	current->tgid == li->pid) \|\| type == li->type))
966	994	continue;
	995	+ if (rw_check == CIFS_LOCK_OP &&
	996	+ (flags & FL_OFDLCK) && (li->flags & FL_OFDLCK) &&
	997	+ server->ops->compare_fids(cfile, cur_cfile))
	998	+ continue;
967	999	if (conf_lock)
968	1000	*conf_lock = li;
969	1001	return true;
..	..	@@ -973,8 +1005,8 @@
973	1005
974	1006	bool
975	1007	cifs_find_lock_conflict(struct cifsFileInfo *cfile, __u64 offset, __u64 length,
976		- __u8 type, struct cifsLockInfo **conf_lock,
977		- int rw_check)
	1008	+ __u8 type, __u16 flags,
	1009	+ struct cifsLockInfo **conf_lock, int rw_check)
978	1010	{
979	1011	bool rc = false;
980	1012	struct cifs_fid_locks *cur;
..	..	@@ -982,7 +1014,8 @@
982	1014
983	1015	list_for_each_entry(cur, &cinode->llist, llist) {
984	1016	rc = cifs_find_fid_lock_conflict(cur, offset, length, type,
985		- cfile, conf_lock, rw_check);
	1017	+ flags, cfile, conf_lock,
	1018	+ rw_check);
986	1019	if (rc)
987	1020	break;
988	1021	}
..	..	@@ -1010,7 +1043,8 @@
1010	1043	down_read(&cinode->lock_sem);
1011	1044
1012	1045	exist = cifs_find_lock_conflict(cfile, offset, length, type,
1013		- &conf_lock, CIFS_LOCK_OP);
	1046	+ flock->fl_flags, &conf_lock,
	1047	+ CIFS_LOCK_OP);
1014	1048	if (exist) {
1015	1049	flock->fl_start = conf_lock->offset;
1016	1050	flock->fl_end = conf_lock->offset + conf_lock->length - 1;
..	..	@@ -1041,7 +1075,7 @@
1041	1075	* Set the byte-range lock (mandatory style). Returns:
1042	1076	* 1) 0, if we set the lock and don't need to request to the server;
1043	1077	* 2) 1, if no locks prevent us but we need to request to the server;
1044		- * 3) -EACCESS, if there is a lock that prevents us and wait is false.
	1078	+ * 3) -EACCES, if there is a lock that prevents us and wait is false.
1045	1079	*/
1046	1080	static int
1047	1081	cifs_lock_add_if(struct cifsFileInfo cfile, struct cifsLockInfo lock,
..	..	@@ -1057,7 +1091,8 @@
1057	1091	cifs_down_write(&cinode->lock_sem);
1058	1092
1059	1093	exist = cifs_find_lock_conflict(cfile, lock->offset, lock->length,
1060		- lock->type, &conf_lock, CIFS_LOCK_OP);
	1094	+ lock->type, lock->flags, &conf_lock,
	1095	+ CIFS_LOCK_OP);
1061	1096	if (!exist && cinode->can_cache_brlcks) {
1062	1097	list_add_tail(&lock->llist, &cfile->llist->locks);
1063	1098	up_write(&cinode->lock_sem);
..	..	@@ -1115,20 +1150,20 @@
1115	1150
1116	1151	/*
1117	1152	* Set the byte-range lock (posix style). Returns:
1118		- * 1) 0, if we set the lock and don't need to request to the server;
1119		- * 2) 1, if we need to request to the server;
1120		- * 3) <0, if the error occurs while setting the lock.
	1153	+ * 1) <0, if the error occurs while setting the lock;
	1154	+ * 2) 0, if we set the lock and don't need to request to the server;
	1155	+ * 3) FILE_LOCK_DEFERRED, if we will wait for some other file_lock;
	1156	+ * 4) FILE_LOCK_DEFERRED + 1, if we need to request to the server.
1121	1157	*/
1122	1158	static int
1123	1159	cifs_posix_lock_set(struct file file, struct file_lock flock)
1124	1160	{
1125	1161	struct cifsInodeInfo *cinode = CIFS_I(file_inode(file));
1126		- int rc = 1;
	1162	+ int rc = FILE_LOCK_DEFERRED + 1;
1127	1163
1128	1164	if ((flock->fl_flags & FL_POSIX) == 0)
1129	1165	return rc;
1130	1166
1131		-try_again:
1132	1167	cifs_down_write(&cinode->lock_sem);
1133	1168	if (!cinode->can_cache_brlcks) {
1134	1169	up_write(&cinode->lock_sem);
..	..	@@ -1137,12 +1172,6 @@
1137	1172
1138	1173	rc = posix_lock_file(file, flock, NULL);
1139	1174	up_write(&cinode->lock_sem);
1140		- if (rc == FILE_LOCK_DEFERRED) {
1141		- rc = wait_event_interruptible(flock->fl_wait, !flock->fl_next);
1142		- if (!rc)
1143		- goto try_again;
1144		- posix_unblock_lock(flock);
1145		- }
1146	1175	return rc;
1147	1176	}
1148	1177
..	..	@@ -1371,7 +1400,7 @@
1371	1400	cifs_dbg(FYI, "Lease on file - not implemented yet\n");
1372	1401	if (flock->fl_flags &
1373	1402	(~(FL_POSIX \| FL_FLOCK \| FL_SLEEP \|
1374		- FL_ACCESS \| FL_LEASE \| FL_CLOSE)))
	1403	+ FL_ACCESS \| FL_LEASE \| FL_CLOSE \| FL_OFDLCK)))
1375	1404	cifs_dbg(FYI, "Unknown lock flags 0x%x\n", flock->fl_flags);
1376	1405
1377	1406	*type = server->vals->large_lock_type;
..	..	@@ -1617,7 +1646,7 @@
1617	1646	int posix_lock_type;
1618	1647
1619	1648	rc = cifs_posix_lock_set(file, flock);
1620		- if (!rc \|\| rc < 0)
	1649	+ if (rc <= FILE_LOCK_DEFERRED)
1621	1650	return rc;
1622	1651
1623	1652	if (type & server->vals->shared_lock_type)
..	..	@@ -1638,7 +1667,8 @@
1638	1667	if (lock) {
1639	1668	struct cifsLockInfo *lock;
1640	1669
1641		- lock = cifs_lock_init(flock->fl_start, length, type);
	1670	+ lock = cifs_lock_init(flock->fl_start, length, type,
	1671	+ flock->fl_flags);
1642	1672	if (!lock)
1643	1673	return -ENOMEM;
1644	1674
..	..	@@ -1677,7 +1707,7 @@
1677	1707	rc = server->ops->mand_unlock_range(cfile, flock, xid);
1678	1708
1679	1709	out:
1680		- if (flock->fl_flags & FL_POSIX) {
	1710	+ if ((flock->fl_flags & FL_POSIX) \|\| (flock->fl_flags & FL_FLOCK)) {
1681	1711	/*
1682	1712	* If this is a request to remove all locks because we
1683	1713	* are closing the file, it doesn't matter if the
..	..	@@ -1694,6 +1724,55 @@
1694	1724	return rc;
1695	1725	}
1696	1726
	1727	+int cifs_flock(struct file file, int cmd, struct file_lock fl)
	1728	+{
	1729	+ int rc, xid;
	1730	+ int lock = 0, unlock = 0;
	1731	+ bool wait_flag = false;
	1732	+ bool posix_lck = false;
	1733	+ struct cifs_sb_info *cifs_sb;
	1734	+ struct cifs_tcon *tcon;
	1735	+ struct cifsFileInfo *cfile;
	1736	+ __u32 type;
	1737	+
	1738	+ xid = get_xid();
	1739	+
	1740	+ if (!(fl->fl_flags & FL_FLOCK)) {
	1741	+ rc = -ENOLCK;
	1742	+ free_xid(xid);
	1743	+ return rc;
	1744	+ }
	1745	+
	1746	+ cfile = (struct cifsFileInfo *)file->private_data;
	1747	+ tcon = tlink_tcon(cfile->tlink);
	1748	+
	1749	+ cifs_read_flock(fl, &type, &lock, &unlock, &wait_flag,
	1750	+ tcon->ses->server);
	1751	+ cifs_sb = CIFS_FILE_SB(file);
	1752	+
	1753	+ if (cap_unix(tcon->ses) &&
	1754	+ (CIFS_UNIX_FCNTL_CAP & le64_to_cpu(tcon->fsUnixInfo.Capability)) &&
	1755	+ ((cifs_sb->mnt_cifs_flags & CIFS_MOUNT_NOPOSIXBRL) == 0))
	1756	+ posix_lck = true;
	1757	+
	1758	+ if (!lock && !unlock) {
	1759	+ /*
	1760	+ * if no lock or unlock then nothing to do since we do not
	1761	+ * know what it is
	1762	+ */
	1763	+ rc = -EOPNOTSUPP;
	1764	+ free_xid(xid);
	1765	+ return rc;
	1766	+ }
	1767	+
	1768	+ rc = cifs_setlk(file, fl, type, wait_flag, posix_lck, lock, unlock,
	1769	+ xid);
	1770	+ free_xid(xid);
	1771	+ return rc;
	1772	+
	1773	+
	1774	+}
	1775	+
1697	1776	int cifs_lock(struct file file, int cmd, struct file_lock flock)
1698	1777	{
1699	1778	int rc, xid;
..	..	@@ -1702,9 +1781,7 @@
1702	1781	bool posix_lck = false;
1703	1782	struct cifs_sb_info *cifs_sb;
1704	1783	struct cifs_tcon *tcon;
1705		- struct cifsInodeInfo *cinode;
1706	1784	struct cifsFileInfo *cfile;
1707		- __u16 netfid;
1708	1785	__u32 type;
1709	1786
1710	1787	rc = -EACCES;
..	..	@@ -1719,10 +1796,7 @@
1719	1796
1720	1797	cifs_read_flock(flock, &type, &lock, &unlock, &wait_flag,
1721	1798	tcon->ses->server);
1722		-
1723	1799	cifs_sb = CIFS_FILE_SB(file);
1724		- netfid = cfile->fid.netfid;
1725		- cinode = CIFS_I(file_inode(file));
1726	1800
1727	1801	if (cap_unix(tcon->ses) &&
1728	1802	(CIFS_UNIX_FCNTL_CAP & le64_to_cpu(tcon->fsUnixInfo.Capability)) &&
..	..	@@ -1774,15 +1848,12 @@
1774	1848	int rc = 0;
1775	1849	unsigned int bytes_written = 0;
1776	1850	unsigned int total_written;
1777		- struct cifs_sb_info *cifs_sb;
1778	1851	struct cifs_tcon *tcon;
1779	1852	struct TCP_Server_Info *server;
1780	1853	unsigned int xid;
1781	1854	struct dentry *dentry = open_file->dentry;
1782	1855	struct cifsInodeInfo *cifsi = CIFS_I(d_inode(dentry));
1783		- struct cifs_io_parms io_parms;
1784		-
1785		- cifs_sb = CIFS_SB(dentry->d_sb);
	1856	+ struct cifs_io_parms io_parms = {0};
1786	1857
1787	1858	cifs_dbg(FYI, "write %zd bytes to offset %lld of %pd\n",
1788	1859	write_size, *offset, dentry);
..	..	@@ -1886,23 +1957,30 @@
1886	1957	return NULL;
1887	1958	}
1888	1959
1889		-struct cifsFileInfo find_writable_file(struct cifsInodeInfo cifs_inode,
1890		- bool fsuid_only)
	1960	+/* Return -EBADF if no handle is found and general rc otherwise */
	1961	+int
	1962	+cifs_get_writable_file(struct cifsInodeInfo *cifs_inode, int flags,
	1963	+ struct cifsFileInfo **ret_file)
1891	1964	{
1892	1965	struct cifsFileInfo open_file, inv_file = NULL;
1893	1966	struct cifs_sb_info *cifs_sb;
1894	1967	bool any_available = false;
1895		- int rc;
	1968	+ int rc = -EBADF;
1896	1969	unsigned int refind = 0;
	1970	+ bool fsuid_only = flags & FIND_WR_FSUID_ONLY;
	1971	+ bool with_delete = flags & FIND_WR_WITH_DELETE;
	1972	+ *ret_file = NULL;
1897	1973
1898		- /* Having a null inode here (because mapping->host was set to zero by
1899		- the VFS or MM) should not happen but we had reports of on oops (due to
1900		- it being zero) during stress testcases so we need to check for it */
	1974	+ /*
	1975	+ * Having a null inode here (because mapping->host was set to zero by
	1976	+ * the VFS or MM) should not happen but we had reports of on oops (due
	1977	+ * to it being zero) during stress testcases so we need to check for it
	1978	+ */
1901	1979
1902	1980	if (cifs_inode == NULL) {
1903	1981	cifs_dbg(VFS, "Null inode passed to cifs_writeable_file\n");
1904	1982	dump_stack();
1905		- return NULL;
	1983	+ return rc;
1906	1984	}
1907	1985
1908	1986	cifs_sb = CIFS_SB(cifs_inode->vfs_inode.i_sb);
..	..	@@ -1915,19 +1993,22 @@
1915	1993	refind_writable:
1916	1994	if (refind > MAX_REOPEN_ATT) {
1917	1995	spin_unlock(&cifs_inode->open_file_lock);
1918		- return NULL;
	1996	+ return rc;
1919	1997	}
1920	1998	list_for_each_entry(open_file, &cifs_inode->openFileList, flist) {
1921	1999	if (!any_available && open_file->pid != current->tgid)
1922	2000	continue;
1923	2001	if (fsuid_only && !uid_eq(open_file->uid, current_fsuid()))
1924	2002	continue;
	2003	+ if (with_delete && !(open_file->fid.access & DELETE))
	2004	+ continue;
1925	2005	if (OPEN_FMODE(open_file->f_flags) & FMODE_WRITE) {
1926	2006	if (!open_file->invalidHandle) {
1927	2007	/* found a good writable file */
1928	2008	cifsFileInfo_get(open_file);
1929	2009	spin_unlock(&cifs_inode->open_file_lock);
1930		- return open_file;
	2010	+ *ret_file = open_file;
	2011	+ return 0;
1931	2012	} else {
1932	2013	if (!inv_file)
1933	2014	inv_file = open_file;
..	..	@@ -1949,22 +2030,107 @@
1949	2030
1950	2031	if (inv_file) {
1951	2032	rc = cifs_reopen_file(inv_file, false);
1952		- if (!rc)
1953		- return inv_file;
1954		- else {
1955		- spin_lock(&cifs_inode->open_file_lock);
1956		- list_move_tail(&inv_file->flist,
1957		- &cifs_inode->openFileList);
1958		- spin_unlock(&cifs_inode->open_file_lock);
1959		- cifsFileInfo_put(inv_file);
1960		- ++refind;
1961		- inv_file = NULL;
1962		- spin_lock(&cifs_inode->open_file_lock);
1963		- goto refind_writable;
	2033	+ if (!rc) {
	2034	+ *ret_file = inv_file;
	2035	+ return 0;
1964	2036	}
	2037	+
	2038	+ spin_lock(&cifs_inode->open_file_lock);
	2039	+ list_move_tail(&inv_file->flist, &cifs_inode->openFileList);
	2040	+ spin_unlock(&cifs_inode->open_file_lock);
	2041	+ cifsFileInfo_put(inv_file);
	2042	+ ++refind;
	2043	+ inv_file = NULL;
	2044	+ spin_lock(&cifs_inode->open_file_lock);
	2045	+ goto refind_writable;
1965	2046	}
1966	2047
1967		- return NULL;
	2048	+ return rc;
	2049	+}
	2050	+
	2051	+struct cifsFileInfo *
	2052	+find_writable_file(struct cifsInodeInfo *cifs_inode, int flags)
	2053	+{
	2054	+ struct cifsFileInfo *cfile;
	2055	+ int rc;
	2056	+
	2057	+ rc = cifs_get_writable_file(cifs_inode, flags, &cfile);
	2058	+ if (rc)
	2059	+ cifs_dbg(FYI, "Couldn't find writable handle rc=%d\n", rc);
	2060	+
	2061	+ return cfile;
	2062	+}
	2063	+
	2064	+int
	2065	+cifs_get_writable_path(struct cifs_tcon tcon, const char name,
	2066	+ int flags,
	2067	+ struct cifsFileInfo **ret_file)
	2068	+{
	2069	+ struct list_head *tmp;
	2070	+ struct cifsFileInfo *cfile;
	2071	+ struct cifsInodeInfo *cinode;
	2072	+ char *full_path;
	2073	+
	2074	+ *ret_file = NULL;
	2075	+
	2076	+ spin_lock(&tcon->open_file_lock);
	2077	+ list_for_each(tmp, &tcon->openFileList) {
	2078	+ cfile = list_entry(tmp, struct cifsFileInfo,
	2079	+ tlist);
	2080	+ full_path = build_path_from_dentry(cfile->dentry);
	2081	+ if (full_path == NULL) {
	2082	+ spin_unlock(&tcon->open_file_lock);
	2083	+ return -ENOMEM;
	2084	+ }
	2085	+ if (strcmp(full_path, name)) {
	2086	+ kfree(full_path);
	2087	+ continue;
	2088	+ }
	2089	+
	2090	+ kfree(full_path);
	2091	+ cinode = CIFS_I(d_inode(cfile->dentry));
	2092	+ spin_unlock(&tcon->open_file_lock);
	2093	+ return cifs_get_writable_file(cinode, flags, ret_file);
	2094	+ }
	2095	+
	2096	+ spin_unlock(&tcon->open_file_lock);
	2097	+ return -ENOENT;
	2098	+}
	2099	+
	2100	+int
	2101	+cifs_get_readable_path(struct cifs_tcon tcon, const char name,
	2102	+ struct cifsFileInfo **ret_file)
	2103	+{
	2104	+ struct list_head *tmp;
	2105	+ struct cifsFileInfo *cfile;
	2106	+ struct cifsInodeInfo *cinode;
	2107	+ char *full_path;
	2108	+
	2109	+ *ret_file = NULL;
	2110	+
	2111	+ spin_lock(&tcon->open_file_lock);
	2112	+ list_for_each(tmp, &tcon->openFileList) {
	2113	+ cfile = list_entry(tmp, struct cifsFileInfo,
	2114	+ tlist);
	2115	+ full_path = build_path_from_dentry(cfile->dentry);
	2116	+ if (full_path == NULL) {
	2117	+ spin_unlock(&tcon->open_file_lock);
	2118	+ return -ENOMEM;
	2119	+ }
	2120	+ if (strcmp(full_path, name)) {
	2121	+ kfree(full_path);
	2122	+ continue;
	2123	+ }
	2124	+
	2125	+ kfree(full_path);
	2126	+ cinode = CIFS_I(d_inode(cfile->dentry));
	2127	+ spin_unlock(&tcon->open_file_lock);
	2128	+ *ret_file = find_readable_file(cinode, 0);
	2129	+ return *ret_file ? 0 : -ENOENT;
	2130	+ }
	2131	+
	2132	+ spin_unlock(&tcon->open_file_lock);
	2133	+ return -ENOENT;
1968	2134	}
1969	2135
1970	2136	static int cifs_partialpagewrite(struct page *page, unsigned from, unsigned to)
..	..	@@ -2001,8 +2167,9 @@
2001	2167	if (mapping->host->i_size - offset < (loff_t)to)
2002	2168	to = (unsigned)(mapping->host->i_size - offset);
2003	2169
2004		- open_file = find_writable_file(CIFS_I(mapping->host), false);
2005		- if (open_file) {
	2170	+ rc = cifs_get_writable_file(CIFS_I(mapping->host), FIND_WR_ANY,
	2171	+ &open_file);
	2172	+ if (!rc) {
2006	2173	bytes_written = cifs_write(open_file, open_file->pid,
2007	2174	write_data, to - from, &offset);
2008	2175	cifsFileInfo_put(open_file);
..	..	@@ -2012,9 +2179,12 @@
2012	2179	rc = 0;
2013	2180	else if (bytes_written < 0)
2014	2181	rc = bytes_written;
	2182	+ else
	2183	+ rc = -EFAULT;
2015	2184	} else {
2016		- cifs_dbg(FYI, "No writeable filehandles for inode\n");
2017		- rc = -EIO;
	2185	+ cifs_dbg(FYI, "No writable handle for write page rc=%d\n", rc);
	2186	+ if (!is_retryable_error(rc))
	2187	+ rc = -EIO;
2018	2188	}
2019	2189
2020	2190	kunmap(page);
..	..	@@ -2121,9 +2291,7 @@
2121	2291	wdata_send_pages(struct cifs_writedata *wdata, unsigned int nr_pages,
2122	2292	struct address_space mapping, struct writeback_control wbc)
2123	2293	{
2124		- int rc = 0;
2125		- struct TCP_Server_Info *server;
2126		- unsigned int i;
	2294	+ int rc;
2127	2295
2128	2296	wdata->sync_mode = wbc->sync_mode;
2129	2297	wdata->nr_pages = nr_pages;
..	..	@@ -2133,21 +2301,17 @@
2133	2301	page_offset(wdata->pages[nr_pages - 1]),
2134	2302	(loff_t)PAGE_SIZE);
2135	2303	wdata->bytes = ((nr_pages - 1) * PAGE_SIZE) + wdata->tailsz;
	2304	+ wdata->pid = wdata->cfile->pid;
2136	2305
2137		- if (wdata->cfile != NULL)
2138		- cifsFileInfo_put(wdata->cfile);
2139		- wdata->cfile = find_writable_file(CIFS_I(mapping->host), false);
2140		- if (!wdata->cfile) {
2141		- cifs_dbg(VFS, "No writable handles for inode\n");
2142		- rc = -EBADF;
2143		- } else {
2144		- wdata->pid = wdata->cfile->pid;
2145		- server = tlink_tcon(wdata->cfile->tlink)->ses->server;
2146		- rc = server->ops->async_writev(wdata, cifs_writedata_release);
2147		- }
	2306	+ rc = adjust_credits(wdata->server, &wdata->credits, wdata->bytes);
	2307	+ if (rc)
	2308	+ return rc;
2148	2309
2149		- for (i = 0; i < nr_pages; ++i)
2150		- unlock_page(wdata->pages[i]);
	2310	+ if (wdata->cfile->invalidHandle)
	2311	+ rc = -EAGAIN;
	2312	+ else
	2313	+ rc = wdata->server->ops->async_writev(wdata,
	2314	+ cifs_writedata_release);
2151	2315
2152	2316	return rc;
2153	2317	}
..	..	@@ -2155,13 +2319,16 @@
2155	2319	static int cifs_writepages(struct address_space *mapping,
2156	2320	struct writeback_control *wbc)
2157	2321	{
2158		- struct cifs_sb_info *cifs_sb = CIFS_SB(mapping->host->i_sb);
	2322	+ struct inode *inode = mapping->host;
	2323	+ struct cifs_sb_info *cifs_sb = CIFS_SB(inode->i_sb);
2159	2324	struct TCP_Server_Info *server;
2160	2325	bool done = false, scanned = false, range_whole = false;
2161	2326	pgoff_t end, index;
2162	2327	struct cifs_writedata *wdata;
	2328	+ struct cifsFileInfo *cfile = NULL;
2163	2329	int rc = 0;
2164	2330	int saved_rc = 0;
	2331	+ unsigned int xid;
2165	2332
2166	2333	/*
2167	2334	* If wsize is smaller than the page cache size, default to writing
..	..	@@ -2170,6 +2337,7 @@
2170	2337	if (cifs_sb->wsize < PAGE_SIZE)
2171	2338	return generic_writepages(mapping, wbc);
2172	2339
	2340	+ xid = get_xid();
2173	2341	if (wbc->range_cyclic) {
2174	2342	index = mapping->writeback_index; /* Start from prev offset */
2175	2343	end = -1;
..	..	@@ -2180,14 +2348,27 @@
2180	2348	range_whole = true;
2181	2349	scanned = true;
2182	2350	}
2183		- server = cifs_sb_master_tcon(cifs_sb)->ses->server;
	2351	+ server = cifs_pick_channel(cifs_sb_master_tcon(cifs_sb)->ses);
	2352	+
2184	2353	retry:
2185	2354	while (!done && index <= end) {
2186		- unsigned int i, nr_pages, found_pages, wsize, credits;
	2355	+ unsigned int i, nr_pages, found_pages, wsize;
2187	2356	pgoff_t next = 0, tofind, saved_index = index;
	2357	+ struct cifs_credits credits_on_stack;
	2358	+ struct cifs_credits *credits = &credits_on_stack;
	2359	+ int get_file_rc = 0;
	2360	+
	2361	+ if (cfile)
	2362	+ cifsFileInfo_put(cfile);
	2363	+
	2364	+ rc = cifs_get_writable_file(CIFS_I(inode), FIND_WR_ANY, &cfile);
	2365	+
	2366	+ /* in case of an error store it to return later */
	2367	+ if (rc)
	2368	+ get_file_rc = rc;
2188	2369
2189	2370	rc = server->ops->wait_mtu_credits(server, cifs_sb->wsize,
2190		- &wsize, &credits);
	2371	+ &wsize, credits);
2191	2372	if (rc != 0) {
2192	2373	done = true;
2193	2374	break;
..	..	@@ -2220,13 +2401,27 @@
2220	2401	continue;
2221	2402	}
2222	2403
2223		- wdata->credits = credits;
	2404	+ wdata->credits = credits_on_stack;
	2405	+ wdata->cfile = cfile;
	2406	+ wdata->server = server;
	2407	+ cfile = NULL;
2224	2408
2225		- rc = wdata_send_pages(wdata, nr_pages, mapping, wbc);
	2409	+ if (!wdata->cfile) {
	2410	+ cifs_dbg(VFS, "No writable handle in writepages rc=%d\n",
	2411	+ get_file_rc);
	2412	+ if (is_retryable_error(get_file_rc))
	2413	+ rc = get_file_rc;
	2414	+ else
	2415	+ rc = -EBADF;
	2416	+ } else
	2417	+ rc = wdata_send_pages(wdata, nr_pages, mapping, wbc);
	2418	+
	2419	+ for (i = 0; i < nr_pages; ++i)
	2420	+ unlock_page(wdata->pages[i]);
2226	2421
2227	2422	/* send failure -- clean up the mess */
2228	2423	if (rc != 0) {
2229		- add_credits_and_wake_if(server, wdata->credits, 0);
	2424	+ add_credits_and_wake_if(server, &wdata->credits, 0);
2230	2425	for (i = 0; i < nr_pages; ++i) {
2231	2426	if (is_retryable_error(rc))
2232	2427	redirty_page_for_writepage(wbc,
..	..	@@ -2278,6 +2473,9 @@
2278	2473	if (wbc->range_cyclic \|\| (range_whole && wbc->nr_to_write > 0))
2279	2474	mapping->writeback_index = index;
2280	2475
	2476	+ if (cfile)
	2477	+ cifsFileInfo_put(cfile);
	2478	+ free_xid(xid);
2281	2479	return rc;
2282	2480	}
2283	2481
..	..	@@ -2402,9 +2600,10 @@
2402	2600	struct cifs_sb_info *cifs_sb = CIFS_SB(inode->i_sb);
2403	2601
2404	2602	rc = file_write_and_wait_range(file, start, end);
2405		- if (rc)
	2603	+ if (rc) {
	2604	+ trace_cifs_fsync_err(inode->i_ino, rc);
2406	2605	return rc;
2407		- inode_lock(inode);
	2606	+ }
2408	2607
2409	2608	xid = get_xid();
2410	2609
..	..	@@ -2422,14 +2621,24 @@
2422	2621	tcon = tlink_tcon(smbfile->tlink);
2423	2622	if (!(cifs_sb->mnt_cifs_flags & CIFS_MOUNT_NOSSYNC)) {
2424	2623	server = tcon->ses->server;
2425		- if (server->ops->flush)
2426		- rc = server->ops->flush(xid, tcon, &smbfile->fid);
2427		- else
	2624	+ if (server->ops->flush == NULL) {
2428	2625	rc = -ENOSYS;
	2626	+ goto strict_fsync_exit;
	2627	+ }
	2628	+
	2629	+ if ((OPEN_FMODE(smbfile->f_flags) & FMODE_WRITE) == 0) {
	2630	+ smbfile = find_writable_file(CIFS_I(inode), FIND_WR_ANY);
	2631	+ if (smbfile) {
	2632	+ rc = server->ops->flush(xid, tcon, &smbfile->fid);
	2633	+ cifsFileInfo_put(smbfile);
	2634	+ } else
	2635	+ cifs_dbg(FYI, "ignore fsync for file not open for write\n");
	2636	+ } else
	2637	+ rc = server->ops->flush(xid, tcon, &smbfile->fid);
2429	2638	}
2430	2639
	2640	+strict_fsync_exit:
2431	2641	free_xid(xid);
2432		- inode_unlock(inode);
2433	2642	return rc;
2434	2643	}
2435	2644
..	..	@@ -2440,13 +2649,14 @@
2440	2649	struct cifs_tcon *tcon;
2441	2650	struct TCP_Server_Info *server;
2442	2651	struct cifsFileInfo *smbfile = file->private_data;
	2652	+ struct inode *inode = file_inode(file);
2443	2653	struct cifs_sb_info *cifs_sb = CIFS_FILE_SB(file);
2444		- struct inode *inode = file->f_mapping->host;
2445	2654
2446	2655	rc = file_write_and_wait_range(file, start, end);
2447		- if (rc)
	2656	+ if (rc) {
	2657	+ trace_cifs_fsync_err(file_inode(file)->i_ino, rc);
2448	2658	return rc;
2449		- inode_lock(inode);
	2659	+ }
2450	2660
2451	2661	xid = get_xid();
2452	2662
..	..	@@ -2456,14 +2666,24 @@
2456	2666	tcon = tlink_tcon(smbfile->tlink);
2457	2667	if (!(cifs_sb->mnt_cifs_flags & CIFS_MOUNT_NOSSYNC)) {
2458	2668	server = tcon->ses->server;
2459		- if (server->ops->flush)
2460		- rc = server->ops->flush(xid, tcon, &smbfile->fid);
2461		- else
	2669	+ if (server->ops->flush == NULL) {
2462	2670	rc = -ENOSYS;
	2671	+ goto fsync_exit;
	2672	+ }
	2673	+
	2674	+ if ((OPEN_FMODE(smbfile->f_flags) & FMODE_WRITE) == 0) {
	2675	+ smbfile = find_writable_file(CIFS_I(inode), FIND_WR_ANY);
	2676	+ if (smbfile) {
	2677	+ rc = server->ops->flush(xid, tcon, &smbfile->fid);
	2678	+ cifsFileInfo_put(smbfile);
	2679	+ } else
	2680	+ cifs_dbg(FYI, "ignore fsync for file not open for write\n");
	2681	+ } else
	2682	+ rc = server->ops->flush(xid, tcon, &smbfile->fid);
2463	2683	}
2464	2684
	2685	+fsync_exit:
2465	2686	free_xid(xid);
2466		- inode_unlock(inode);
2467	2687	return rc;
2468	2688	}
2469	2689
..	..	@@ -2480,7 +2700,8 @@
2480	2700	rc = filemap_write_and_wait(inode->i_mapping);
2481	2701
2482	2702	cifs_dbg(FYI, "Flush inode %p file %p rc %d\n", inode, file, rc);
2483		-
	2703	+ if (rc)
	2704	+ trace_cifs_flush_err(inode->i_ino, rc);
2484	2705	return rc;
2485	2706	}
2486	2707
..	..	@@ -2603,6 +2824,76 @@
2603	2824	}
2604	2825
2605	2826	static int
	2827	+cifs_resend_wdata(struct cifs_writedata wdata, struct list_head wdata_list,
	2828	+ struct cifs_aio_ctx *ctx)
	2829	+{
	2830	+ unsigned int wsize;
	2831	+ struct cifs_credits credits;
	2832	+ int rc;
	2833	+ struct TCP_Server_Info *server = wdata->server;
	2834	+
	2835	+ do {
	2836	+ if (wdata->cfile->invalidHandle) {
	2837	+ rc = cifs_reopen_file(wdata->cfile, false);
	2838	+ if (rc == -EAGAIN)
	2839	+ continue;
	2840	+ else if (rc)
	2841	+ break;
	2842	+ }
	2843	+
	2844	+
	2845	+ /*
	2846	+ * Wait for credits to resend this wdata.
	2847	+ * Note: we are attempting to resend the whole wdata not in
	2848	+ * segments
	2849	+ */
	2850	+ do {
	2851	+ rc = server->ops->wait_mtu_credits(server, wdata->bytes,
	2852	+ &wsize, &credits);
	2853	+ if (rc)
	2854	+ goto fail;
	2855	+
	2856	+ if (wsize < wdata->bytes) {
	2857	+ add_credits_and_wake_if(server, &credits, 0);
	2858	+ msleep(1000);
	2859	+ }
	2860	+ } while (wsize < wdata->bytes);
	2861	+ wdata->credits = credits;
	2862	+
	2863	+ rc = adjust_credits(server, &wdata->credits, wdata->bytes);
	2864	+
	2865	+ if (!rc) {
	2866	+ if (wdata->cfile->invalidHandle)
	2867	+ rc = -EAGAIN;
	2868	+ else {
	2869	+#ifdef CONFIG_CIFS_SMB_DIRECT
	2870	+ if (wdata->mr) {
	2871	+ wdata->mr->need_invalidate = true;
	2872	+ smbd_deregister_mr(wdata->mr);
	2873	+ wdata->mr = NULL;
	2874	+ }
	2875	+#endif
	2876	+ rc = server->ops->async_writev(wdata,
	2877	+ cifs_uncached_writedata_release);
	2878	+ }
	2879	+ }
	2880	+
	2881	+ /* If the write was successfully sent, we are done */
	2882	+ if (!rc) {
	2883	+ list_add_tail(&wdata->list, wdata_list);
	2884	+ return 0;
	2885	+ }
	2886	+
	2887	+ /* Roll back credits and retry if needed */
	2888	+ add_credits_and_wake_if(server, &wdata->credits, 0);
	2889	+ } while (rc == -EAGAIN);
	2890	+
	2891	+fail:
	2892	+ kref_put(&wdata->refcount, cifs_uncached_writedata_release);
	2893	+ return rc;
	2894	+}
	2895	+
	2896	+static int
2606	2897	cifs_write_from_iter(loff_t offset, size_t len, struct iov_iter *from,
2607	2898	struct cifsFileInfo *open_file,
2608	2899	struct cifs_sb_info cifs_sb, struct list_head wdata_list,
..	..	@@ -2616,73 +2907,139 @@
2616	2907	loff_t saved_offset = offset;
2617	2908	pid_t pid;
2618	2909	struct TCP_Server_Info *server;
	2910	+ struct page **pagevec;
	2911	+ size_t start;
	2912	+ unsigned int xid;
2619	2913
2620	2914	if (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_RWPIDFORWARD)
2621	2915	pid = open_file->pid;
2622	2916	else
2623	2917	pid = current->tgid;
2624	2918
2625		- server = tlink_tcon(open_file->tlink)->ses->server;
	2919	+ server = cifs_pick_channel(tlink_tcon(open_file->tlink)->ses);
	2920	+ xid = get_xid();
2626	2921
2627	2922	do {
2628		- unsigned int wsize, credits;
	2923	+ unsigned int wsize;
	2924	+ struct cifs_credits credits_on_stack;
	2925	+ struct cifs_credits *credits = &credits_on_stack;
	2926	+
	2927	+ if (open_file->invalidHandle) {
	2928	+ rc = cifs_reopen_file(open_file, false);
	2929	+ if (rc == -EAGAIN)
	2930	+ continue;
	2931	+ else if (rc)
	2932	+ break;
	2933	+ }
2629	2934
2630	2935	rc = server->ops->wait_mtu_credits(server, cifs_sb->wsize,
2631		- &wsize, &credits);
	2936	+ &wsize, credits);
2632	2937	if (rc)
2633	2938	break;
2634	2939
2635		- nr_pages = get_numpages(wsize, len, &cur_len);
2636		- wdata = cifs_writedata_alloc(nr_pages,
	2940	+ cur_len = min_t(const size_t, len, wsize);
	2941	+
	2942	+ if (ctx->direct_io) {
	2943	+ ssize_t result;
	2944	+
	2945	+ result = iov_iter_get_pages_alloc(
	2946	+ from, &pagevec, cur_len, &start);
	2947	+ if (result < 0) {
	2948	+ cifs_dbg(VFS,
	2949	+ "direct_writev couldn't get user pages (rc=%zd) iter type %d iov_offset %zd count %zd\n",
	2950	+ result, iov_iter_type(from),
	2951	+ from->iov_offset, from->count);
	2952	+ dump_stack();
	2953	+
	2954	+ rc = result;
	2955	+ add_credits_and_wake_if(server, credits, 0);
	2956	+ break;
	2957	+ }
	2958	+ cur_len = (size_t)result;
	2959	+ iov_iter_advance(from, cur_len);
	2960	+
	2961	+ nr_pages =
	2962	+ (cur_len + start + PAGE_SIZE - 1) / PAGE_SIZE;
	2963	+
	2964	+ wdata = cifs_writedata_direct_alloc(pagevec,
2637	2965	cifs_uncached_writev_complete);
2638		- if (!wdata) {
2639		- rc = -ENOMEM;
2640		- add_credits_and_wake_if(server, credits, 0);
2641		- break;
2642		- }
	2966	+ if (!wdata) {
	2967	+ rc = -ENOMEM;
	2968	+ add_credits_and_wake_if(server, credits, 0);
	2969	+ break;
	2970	+ }
2643	2971
2644		- rc = cifs_write_allocate_pages(wdata->pages, nr_pages);
2645		- if (rc) {
2646		- kfree(wdata);
2647		- add_credits_and_wake_if(server, credits, 0);
2648		- break;
2649		- }
2650	2972
2651		- num_pages = nr_pages;
2652		- rc = wdata_fill_from_iovec(wdata, from, &cur_len, &num_pages);
2653		- if (rc) {
2654		- for (i = 0; i < nr_pages; i++)
2655		- put_page(wdata->pages[i]);
2656		- kfree(wdata);
2657		- add_credits_and_wake_if(server, credits, 0);
2658		- break;
2659		- }
	2973	+ wdata->page_offset = start;
	2974	+ wdata->tailsz =
	2975	+ nr_pages > 1 ?
	2976	+ cur_len - (PAGE_SIZE - start) -
	2977	+ (nr_pages - 2) * PAGE_SIZE :
	2978	+ cur_len;
	2979	+ } else {
	2980	+ nr_pages = get_numpages(wsize, len, &cur_len);
	2981	+ wdata = cifs_writedata_alloc(nr_pages,
	2982	+ cifs_uncached_writev_complete);
	2983	+ if (!wdata) {
	2984	+ rc = -ENOMEM;
	2985	+ add_credits_and_wake_if(server, credits, 0);
	2986	+ break;
	2987	+ }
2660	2988
2661		- /*
2662		- * Bring nr_pages down to the number of pages we actually used,
2663		- * and free any pages that we didn't use.
2664		- */
2665		- for ( ; nr_pages > num_pages; nr_pages--)
2666		- put_page(wdata->pages[nr_pages - 1]);
	2989	+ rc = cifs_write_allocate_pages(wdata->pages, nr_pages);
	2990	+ if (rc) {
	2991	+ kvfree(wdata->pages);
	2992	+ kfree(wdata);
	2993	+ add_credits_and_wake_if(server, credits, 0);
	2994	+ break;
	2995	+ }
	2996	+
	2997	+ num_pages = nr_pages;
	2998	+ rc = wdata_fill_from_iovec(
	2999	+ wdata, from, &cur_len, &num_pages);
	3000	+ if (rc) {
	3001	+ for (i = 0; i < nr_pages; i++)
	3002	+ put_page(wdata->pages[i]);
	3003	+ kvfree(wdata->pages);
	3004	+ kfree(wdata);
	3005	+ add_credits_and_wake_if(server, credits, 0);
	3006	+ break;
	3007	+ }
	3008	+
	3009	+ /*
	3010	+ * Bring nr_pages down to the number of pages we
	3011	+ * actually used, and free any pages that we didn't use.
	3012	+ */
	3013	+ for ( ; nr_pages > num_pages; nr_pages--)
	3014	+ put_page(wdata->pages[nr_pages - 1]);
	3015	+
	3016	+ wdata->tailsz = cur_len - ((nr_pages - 1) * PAGE_SIZE);
	3017	+ }
2667	3018
2668	3019	wdata->sync_mode = WB_SYNC_ALL;
2669	3020	wdata->nr_pages = nr_pages;
2670	3021	wdata->offset = (__u64)offset;
2671	3022	wdata->cfile = cifsFileInfo_get(open_file);
	3023	+ wdata->server = server;
2672	3024	wdata->pid = pid;
2673	3025	wdata->bytes = cur_len;
2674	3026	wdata->pagesz = PAGE_SIZE;
2675		- wdata->tailsz = cur_len - ((nr_pages - 1) * PAGE_SIZE);
2676		- wdata->credits = credits;
	3027	+ wdata->credits = credits_on_stack;
2677	3028	wdata->ctx = ctx;
2678	3029	kref_get(&ctx->refcount);
2679	3030
2680		- if (!wdata->cfile->invalidHandle \|\|
2681		- !(rc = cifs_reopen_file(wdata->cfile, false)))
2682		- rc = server->ops->async_writev(wdata,
	3031	+ rc = adjust_credits(server, &wdata->credits, wdata->bytes);
	3032	+
	3033	+ if (!rc) {
	3034	+ if (wdata->cfile->invalidHandle)
	3035	+ rc = -EAGAIN;
	3036	+ else
	3037	+ rc = server->ops->async_writev(wdata,
2683	3038	cifs_uncached_writedata_release);
	3039	+ }
	3040	+
2684	3041	if (rc) {
2685		- add_credits_and_wake_if(server, wdata->credits, 0);
	3042	+ add_credits_and_wake_if(server, &wdata->credits, 0);
2686	3043	kref_put(&wdata->refcount,
2687	3044	cifs_uncached_writedata_release);
2688	3045	if (rc == -EAGAIN) {
..	..	@@ -2698,6 +3055,7 @@
2698	3055	len -= cur_len;
2699	3056	} while (len > 0);
2700	3057
	3058	+ free_xid(xid);
2701	3059	return rc;
2702	3060	}
2703	3061
..	..	@@ -2707,8 +3065,7 @@
2707	3065	struct cifs_tcon *tcon;
2708	3066	struct cifs_sb_info *cifs_sb;
2709	3067	struct dentry *dentry = ctx->cfile->dentry;
2710		- unsigned int i;
2711		- int rc;
	3068	+ ssize_t rc;
2712	3069
2713	3070	tcon = tlink_tcon(ctx->cfile->tlink);
2714	3071	cifs_sb = CIFS_SB(dentry->d_sb);
..	..	@@ -2747,27 +3104,29 @@
2747	3104	INIT_LIST_HEAD(&tmp_list);
2748	3105	list_del_init(&wdata->list);
2749	3106
2750		- iov_iter_advance(&tmp_from,
	3107	+ if (ctx->direct_io)
	3108	+ rc = cifs_resend_wdata(
	3109	+ wdata, &tmp_list, ctx);
	3110	+ else {
	3111	+ iov_iter_advance(&tmp_from,
2751	3112	wdata->offset - ctx->pos);
2752	3113
2753		- rc = cifs_write_from_iter(wdata->offset,
	3114	+ rc = cifs_write_from_iter(wdata->offset,
2754	3115	wdata->bytes, &tmp_from,
2755	3116	ctx->cfile, cifs_sb, &tmp_list,
2756	3117	ctx);
2757	3118
2758		- list_splice(&tmp_list, &ctx->list);
	3119	+ kref_put(&wdata->refcount,
	3120	+ cifs_uncached_writedata_release);
	3121	+ }
2759	3122
2760		- kref_put(&wdata->refcount,
2761		- cifs_uncached_writedata_release);
	3123	+ list_splice(&tmp_list, &ctx->list);
2762	3124	goto restart_loop;
2763	3125	}
2764	3126	}
2765	3127	list_del_init(&wdata->list);
2766	3128	kref_put(&wdata->refcount, cifs_uncached_writedata_release);
2767	3129	}
2768		-
2769		- for (i = 0; i < ctx->npages; i++)
2770		- put_page(ctx->bv[i].bv_page);
2771	3130
2772	3131	cifs_stats_bytes_written(tcon, ctx->total_len);
2773	3132	set_bit(CIFS_INO_INVALID_MAPPING, &CIFS_I(dentry->d_inode)->flags);
..	..	@@ -2782,7 +3141,8 @@
2782	3141	complete(&ctx->done);
2783	3142	}
2784	3143
2785		-ssize_t cifs_user_writev(struct kiocb iocb, struct iov_iter from)
	3144	+static ssize_t __cifs_writev(
	3145	+ struct kiocb iocb, struct iov_iter from, bool direct)
2786	3146	{
2787	3147	struct file *file = iocb->ki_filp;
2788	3148	ssize_t total_written = 0;
..	..	@@ -2791,13 +3151,18 @@
2791	3151	struct cifs_sb_info *cifs_sb;
2792	3152	struct cifs_aio_ctx *ctx;
2793	3153	struct iov_iter saved_from = *from;
	3154	+ size_t len = iov_iter_count(from);
2794	3155	int rc;
2795	3156
2796	3157	/*
2797		- * BB - optimize the way when signing is disabled. We can drop this
2798		- * extra memory-to-memory copying and use iovec buffers for constructing
2799		- * write request.
	3158	+ * iov_iter_get_pages_alloc doesn't work with ITER_KVEC.
	3159	+ * In this case, fall back to non-direct write function.
	3160	+ * this could be improved by getting pages directly in ITER_KVEC
2800	3161	*/
	3162	+ if (direct && iov_iter_is_kvec(from)) {
	3163	+ cifs_dbg(FYI, "use non-direct cifs_writev for kvec I/O\n");
	3164	+ direct = false;
	3165	+ }
2801	3166
2802	3167	rc = generic_write_checks(iocb, from);
2803	3168	if (rc <= 0)
..	..	@@ -2821,10 +3186,16 @@
2821	3186
2822	3187	ctx->pos = iocb->ki_pos;
2823	3188
2824		- rc = setup_aio_ctx_iter(ctx, from, WRITE);
2825		- if (rc) {
2826		- kref_put(&ctx->refcount, cifs_aio_ctx_release);
2827		- return rc;
	3189	+ if (direct) {
	3190	+ ctx->direct_io = true;
	3191	+ ctx->iter = *from;
	3192	+ ctx->len = len;
	3193	+ } else {
	3194	+ rc = setup_aio_ctx_iter(ctx, from, WRITE);
	3195	+ if (rc) {
	3196	+ kref_put(&ctx->refcount, cifs_aio_ctx_release);
	3197	+ return rc;
	3198	+ }
2828	3199	}
2829	3200
2830	3201	/* grab a lock here due to read response handlers can access ctx */
..	..	@@ -2874,6 +3245,19 @@
2874	3245	return total_written;
2875	3246	}
2876	3247
	3248	+ssize_t cifs_direct_writev(struct kiocb iocb, struct iov_iter from)
	3249	+{
	3250	+ struct file *file = iocb->ki_filp;
	3251	+
	3252	+ cifs_revalidate_mapping(file->f_inode);
	3253	+ return __cifs_writev(iocb, from, true);
	3254	+}
	3255	+
	3256	+ssize_t cifs_user_writev(struct kiocb iocb, struct iov_iter from)
	3257	+{
	3258	+ return __cifs_writev(iocb, from, false);
	3259	+}
	3260	+
2877	3261	static ssize_t
2878	3262	cifs_writev(struct kiocb iocb, struct iov_iter from)
2879	3263	{
..	..	@@ -2896,8 +3280,8 @@
2896	3280	goto out;
2897	3281
2898	3282	if (!cifs_find_lock_conflict(cfile, iocb->ki_pos, iov_iter_count(from),
2899		- server->vals->exclusive_lock_type, NULL,
2900		- CIFS_WRITE_OP))
	3283	+ server->vals->exclusive_lock_type, 0,
	3284	+ NULL, CIFS_WRITE_OP))
2901	3285	rc = __generic_file_write_iter(iocb, from);
2902	3286	else
2903	3287	rc = -EACCES;
..	..	@@ -3048,7 +3432,6 @@
3048	3432	kref_put(&rdata->ctx->refcount, cifs_aio_ctx_release);
3049	3433	for (i = 0; i < rdata->nr_pages; i++) {
3050	3434	put_page(rdata->pages[i]);
3051		- rdata->pages[i] = NULL;
3052	3435	}
3053	3436	cifs_readdata_release(refcount);
3054	3437	}
..	..	@@ -3073,7 +3456,7 @@
3073	3456	size_t copy = min_t(size_t, remaining, PAGE_SIZE);
3074	3457	size_t written;
3075	3458
3076		- if (unlikely(iter->type & ITER_PIPE)) {
	3459	+ if (unlikely(iov_iter_is_pipe(iter))) {
3077	3460	void *addr = kmap_atomic(page);
3078	3461
3079	3462	written = copy_to_iter(addr, copy, iter);
..	..	@@ -3156,7 +3539,7 @@
3156	3539	rdata->got_bytes += result;
3157	3540	}
3158	3541
3159		- return rdata->got_bytes > 0 && result != -ECONNABORTED ?
	3542	+ return result != -ECONNABORTED && rdata->got_bytes > 0 ?
3160	3543	rdata->got_bytes : result;
3161	3544	}
3162	3545
..	..	@@ -3175,70 +3558,209 @@
3175	3558	return uncached_fill_pages(server, rdata, iter, iter->count);
3176	3559	}
3177	3560
	3561	+static int cifs_resend_rdata(struct cifs_readdata *rdata,
	3562	+ struct list_head *rdata_list,
	3563	+ struct cifs_aio_ctx *ctx)
	3564	+{
	3565	+ unsigned int rsize;
	3566	+ struct cifs_credits credits;
	3567	+ int rc;
	3568	+ struct TCP_Server_Info *server;
	3569	+
	3570	+ /* XXX: should we pick a new channel here? */
	3571	+ server = rdata->server;
	3572	+
	3573	+ do {
	3574	+ if (rdata->cfile->invalidHandle) {
	3575	+ rc = cifs_reopen_file(rdata->cfile, true);
	3576	+ if (rc == -EAGAIN)
	3577	+ continue;
	3578	+ else if (rc)
	3579	+ break;
	3580	+ }
	3581	+
	3582	+ /*
	3583	+ * Wait for credits to resend this rdata.
	3584	+ * Note: we are attempting to resend the whole rdata not in
	3585	+ * segments
	3586	+ */
	3587	+ do {
	3588	+ rc = server->ops->wait_mtu_credits(server, rdata->bytes,
	3589	+ &rsize, &credits);
	3590	+
	3591	+ if (rc)
	3592	+ goto fail;
	3593	+
	3594	+ if (rsize < rdata->bytes) {
	3595	+ add_credits_and_wake_if(server, &credits, 0);
	3596	+ msleep(1000);
	3597	+ }
	3598	+ } while (rsize < rdata->bytes);
	3599	+ rdata->credits = credits;
	3600	+
	3601	+ rc = adjust_credits(server, &rdata->credits, rdata->bytes);
	3602	+ if (!rc) {
	3603	+ if (rdata->cfile->invalidHandle)
	3604	+ rc = -EAGAIN;
	3605	+ else {
	3606	+#ifdef CONFIG_CIFS_SMB_DIRECT
	3607	+ if (rdata->mr) {
	3608	+ rdata->mr->need_invalidate = true;
	3609	+ smbd_deregister_mr(rdata->mr);
	3610	+ rdata->mr = NULL;
	3611	+ }
	3612	+#endif
	3613	+ rc = server->ops->async_readv(rdata);
	3614	+ }
	3615	+ }
	3616	+
	3617	+ /* If the read was successfully sent, we are done */
	3618	+ if (!rc) {
	3619	+ /* Add to aio pending list */
	3620	+ list_add_tail(&rdata->list, rdata_list);
	3621	+ return 0;
	3622	+ }
	3623	+
	3624	+ /* Roll back credits and retry if needed */
	3625	+ add_credits_and_wake_if(server, &rdata->credits, 0);
	3626	+ } while (rc == -EAGAIN);
	3627	+
	3628	+fail:
	3629	+ kref_put(&rdata->refcount, cifs_uncached_readdata_release);
	3630	+ return rc;
	3631	+}
	3632	+
3178	3633	static int
3179	3634	cifs_send_async_read(loff_t offset, size_t len, struct cifsFileInfo *open_file,
3180	3635	struct cifs_sb_info cifs_sb, struct list_head rdata_list,
3181	3636	struct cifs_aio_ctx *ctx)
3182	3637	{
3183	3638	struct cifs_readdata *rdata;
3184		- unsigned int npages, rsize, credits;
	3639	+ unsigned int npages, rsize;
	3640	+ struct cifs_credits credits_on_stack;
	3641	+ struct cifs_credits *credits = &credits_on_stack;
3185	3642	size_t cur_len;
3186	3643	int rc;
3187	3644	pid_t pid;
3188	3645	struct TCP_Server_Info *server;
	3646	+ struct page **pagevec;
	3647	+ size_t start;
	3648	+ struct iov_iter direct_iov = ctx->iter;
3189	3649
3190		- server = tlink_tcon(open_file->tlink)->ses->server;
	3650	+ server = cifs_pick_channel(tlink_tcon(open_file->tlink)->ses);
3191	3651
3192	3652	if (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_RWPIDFORWARD)
3193	3653	pid = open_file->pid;
3194	3654	else
3195	3655	pid = current->tgid;
3196	3656
	3657	+ if (ctx->direct_io)
	3658	+ iov_iter_advance(&direct_iov, offset - ctx->pos);
	3659	+
3197	3660	do {
	3661	+ if (open_file->invalidHandle) {
	3662	+ rc = cifs_reopen_file(open_file, true);
	3663	+ if (rc == -EAGAIN)
	3664	+ continue;
	3665	+ else if (rc)
	3666	+ break;
	3667	+ }
	3668	+
3198	3669	rc = server->ops->wait_mtu_credits(server, cifs_sb->rsize,
3199		- &rsize, &credits);
	3670	+ &rsize, credits);
3200	3671	if (rc)
3201	3672	break;
3202	3673
3203	3674	cur_len = min_t(const size_t, len, rsize);
3204		- npages = DIV_ROUND_UP(cur_len, PAGE_SIZE);
3205	3675
3206		- /* allocate a readdata struct */
3207		- rdata = cifs_readdata_alloc(npages,
	3676	+ if (ctx->direct_io) {
	3677	+ ssize_t result;
	3678	+
	3679	+ result = iov_iter_get_pages_alloc(
	3680	+ &direct_iov, &pagevec,
	3681	+ cur_len, &start);
	3682	+ if (result < 0) {
	3683	+ cifs_dbg(VFS,
	3684	+ "Couldn't get user pages (rc=%zd) iter type %d iov_offset %zd count %zd\n",
	3685	+ result, iov_iter_type(&direct_iov),
	3686	+ direct_iov.iov_offset,
	3687	+ direct_iov.count);
	3688	+ dump_stack();
	3689	+
	3690	+ rc = result;
	3691	+ add_credits_and_wake_if(server, credits, 0);
	3692	+ break;
	3693	+ }
	3694	+ cur_len = (size_t)result;
	3695	+ iov_iter_advance(&direct_iov, cur_len);
	3696	+
	3697	+ rdata = cifs_readdata_direct_alloc(
	3698	+ pagevec, cifs_uncached_readv_complete);
	3699	+ if (!rdata) {
	3700	+ add_credits_and_wake_if(server, credits, 0);
	3701	+ rc = -ENOMEM;
	3702	+ break;
	3703	+ }
	3704	+
	3705	+ npages = (cur_len + start + PAGE_SIZE-1) / PAGE_SIZE;
	3706	+ rdata->page_offset = start;
	3707	+ rdata->tailsz = npages > 1 ?
	3708	+ cur_len-(PAGE_SIZE-start)-(npages-2)*PAGE_SIZE :
	3709	+ cur_len;
	3710	+
	3711	+ } else {
	3712	+
	3713	+ npages = DIV_ROUND_UP(cur_len, PAGE_SIZE);
	3714	+ /* allocate a readdata struct */
	3715	+ rdata = cifs_readdata_alloc(npages,
3208	3716	cifs_uncached_readv_complete);
3209		- if (!rdata) {
3210		- add_credits_and_wake_if(server, credits, 0);
3211		- rc = -ENOMEM;
3212		- break;
	3717	+ if (!rdata) {
	3718	+ add_credits_and_wake_if(server, credits, 0);
	3719	+ rc = -ENOMEM;
	3720	+ break;
	3721	+ }
	3722	+
	3723	+ rc = cifs_read_allocate_pages(rdata, npages);
	3724	+ if (rc) {
	3725	+ kvfree(rdata->pages);
	3726	+ kfree(rdata);
	3727	+ add_credits_and_wake_if(server, credits, 0);
	3728	+ break;
	3729	+ }
	3730	+
	3731	+ rdata->tailsz = PAGE_SIZE;
3213	3732	}
3214	3733
3215		- rc = cifs_read_allocate_pages(rdata, npages);
3216		- if (rc)
3217		- goto error;
3218		-
	3734	+ rdata->server = server;
3219	3735	rdata->cfile = cifsFileInfo_get(open_file);
3220	3736	rdata->nr_pages = npages;
3221	3737	rdata->offset = offset;
3222	3738	rdata->bytes = cur_len;
3223	3739	rdata->pid = pid;
3224	3740	rdata->pagesz = PAGE_SIZE;
3225		- rdata->tailsz = PAGE_SIZE;
3226	3741	rdata->read_into_pages = cifs_uncached_read_into_pages;
3227	3742	rdata->copy_into_pages = cifs_uncached_copy_into_pages;
3228		- rdata->credits = credits;
	3743	+ rdata->credits = credits_on_stack;
3229	3744	rdata->ctx = ctx;
3230	3745	kref_get(&ctx->refcount);
3231	3746
3232		- if (!rdata->cfile->invalidHandle \|\|
3233		- !(rc = cifs_reopen_file(rdata->cfile, true)))
3234		- rc = server->ops->async_readv(rdata);
3235		-error:
	3747	+ rc = adjust_credits(server, &rdata->credits, rdata->bytes);
	3748	+
	3749	+ if (!rc) {
	3750	+ if (rdata->cfile->invalidHandle)
	3751	+ rc = -EAGAIN;
	3752	+ else
	3753	+ rc = server->ops->async_readv(rdata);
	3754	+ }
	3755	+
3236	3756	if (rc) {
3237		- add_credits_and_wake_if(server, rdata->credits, 0);
	3757	+ add_credits_and_wake_if(server, &rdata->credits, 0);
3238	3758	kref_put(&rdata->refcount,
3239		- cifs_uncached_readdata_release);
3240		- if (rc == -EAGAIN)
	3759	+ cifs_uncached_readdata_release);
	3760	+ if (rc == -EAGAIN) {
	3761	+ iov_iter_revert(&direct_iov, cur_len);
3241	3762	continue;
	3763	+ }
3242	3764	break;
3243	3765	}
3244	3766
..	..	@@ -3256,11 +3778,8 @@
3256	3778	struct cifs_readdata rdata, tmp;
3257	3779	struct iov_iter *to = &ctx->iter;
3258	3780	struct cifs_sb_info *cifs_sb;
3259		- struct cifs_tcon *tcon;
3260		- unsigned int i;
3261	3781	int rc;
3262	3782
3263		- tcon = tlink_tcon(ctx->cfile->tlink);
3264	3783	cifs_sb = CIFS_SB(ctx->cfile->dentry->d_sb);
3265	3784
3266	3785	mutex_lock(&ctx->aio_mutex);
..	..	@@ -3294,47 +3813,55 @@
3294	3813	* reading.
3295	3814	*/
3296	3815	if (got_bytes && got_bytes < rdata->bytes) {
3297		- rc = cifs_readdata_to_iov(rdata, to);
	3816	+ rc = 0;
	3817	+ if (!ctx->direct_io)
	3818	+ rc = cifs_readdata_to_iov(rdata, to);
3298	3819	if (rc) {
3299	3820	kref_put(&rdata->refcount,
3300		- cifs_uncached_readdata_release);
	3821	+ cifs_uncached_readdata_release);
3301	3822	continue;
3302	3823	}
3303	3824	}
3304	3825
3305		- rc = cifs_send_async_read(
	3826	+ if (ctx->direct_io) {
	3827	+ /*
	3828	+ * Re-use rdata as this is a
	3829	+ * direct I/O
	3830	+ */
	3831	+ rc = cifs_resend_rdata(
	3832	+ rdata,
	3833	+ &tmp_list, ctx);
	3834	+ } else {
	3835	+ rc = cifs_send_async_read(
3306	3836	rdata->offset + got_bytes,
3307	3837	rdata->bytes - got_bytes,
3308	3838	rdata->cfile, cifs_sb,
3309	3839	&tmp_list, ctx);
3310	3840
	3841	+ kref_put(&rdata->refcount,
	3842	+ cifs_uncached_readdata_release);
	3843	+ }
	3844	+
3311	3845	list_splice(&tmp_list, &ctx->list);
3312	3846
3313		- kref_put(&rdata->refcount,
3314		- cifs_uncached_readdata_release);
3315	3847	goto again;
3316	3848	} else if (rdata->result)
3317	3849	rc = rdata->result;
3318		- else
	3850	+ else if (!ctx->direct_io)
3319	3851	rc = cifs_readdata_to_iov(rdata, to);
3320	3852
3321	3853	/* if there was a short read -- discard anything left */
3322	3854	if (rdata->got_bytes && rdata->got_bytes < rdata->bytes)
3323	3855	rc = -ENODATA;
	3856	+
	3857	+ ctx->total_len += rdata->got_bytes;
3324	3858	}
3325	3859	list_del_init(&rdata->list);
3326	3860	kref_put(&rdata->refcount, cifs_uncached_readdata_release);
3327	3861	}
3328	3862
3329		- for (i = 0; i < ctx->npages; i++) {
3330		- if (ctx->should_dirty)
3331		- set_page_dirty(ctx->bv[i].bv_page);
3332		- put_page(ctx->bv[i].bv_page);
3333		- }
3334		-
3335		- ctx->total_len = ctx->len - iov_iter_count(to);
3336		-
3337		- cifs_stats_bytes_read(tcon, ctx->total_len);
	3863	+ if (!ctx->direct_io)
	3864	+ ctx->total_len = ctx->len - iov_iter_count(to);
3338	3865
3339	3866	/* mask nodata case */
3340	3867	if (rc == -ENODATA)
..	..	@@ -3350,17 +3877,27 @@
3350	3877	complete(&ctx->done);
3351	3878	}
3352	3879
3353		-ssize_t cifs_user_readv(struct kiocb iocb, struct iov_iter to)
	3880	+static ssize_t __cifs_readv(
	3881	+ struct kiocb iocb, struct iov_iter to, bool direct)
3354	3882	{
3355		- struct file *file = iocb->ki_filp;
3356		- ssize_t rc;
3357	3883	size_t len;
3358		- ssize_t total_read = 0;
3359		- loff_t offset = iocb->ki_pos;
	3884	+ struct file *file = iocb->ki_filp;
3360	3885	struct cifs_sb_info *cifs_sb;
3361		- struct cifs_tcon *tcon;
3362	3886	struct cifsFileInfo *cfile;
	3887	+ struct cifs_tcon *tcon;
	3888	+ ssize_t rc, total_read = 0;
	3889	+ loff_t offset = iocb->ki_pos;
3363	3890	struct cifs_aio_ctx *ctx;
	3891	+
	3892	+ /*
	3893	+ * iov_iter_get_pages_alloc() doesn't work with ITER_KVEC,
	3894	+ * fall back to data copy read path
	3895	+ * this could be improved by getting pages directly in ITER_KVEC
	3896	+ */
	3897	+ if (direct && iov_iter_is_kvec(to)) {
	3898	+ cifs_dbg(FYI, "use non-direct cifs_user_readv for kvec I/O\n");
	3899	+ direct = false;
	3900	+ }
3364	3901
3365	3902	len = iov_iter_count(to);
3366	3903	if (!len)
..	..	@@ -3385,16 +3922,31 @@
3385	3922	if (!is_sync_kiocb(iocb))
3386	3923	ctx->iocb = iocb;
3387	3924
3388		- if (to->type == ITER_IOVEC)
	3925	+ if (iter_is_iovec(to))
3389	3926	ctx->should_dirty = true;
3390	3927
3391		- rc = setup_aio_ctx_iter(ctx, to, READ);
3392		- if (rc) {
3393		- kref_put(&ctx->refcount, cifs_aio_ctx_release);
3394		- return rc;
	3928	+ if (direct) {
	3929	+ ctx->pos = offset;
	3930	+ ctx->direct_io = true;
	3931	+ ctx->iter = *to;
	3932	+ ctx->len = len;
	3933	+ } else {
	3934	+ rc = setup_aio_ctx_iter(ctx, to, READ);
	3935	+ if (rc) {
	3936	+ kref_put(&ctx->refcount, cifs_aio_ctx_release);
	3937	+ return rc;
	3938	+ }
	3939	+ len = ctx->len;
3395	3940	}
3396	3941
3397		- len = ctx->len;
	3942	+ if (direct) {
	3943	+ rc = filemap_write_and_wait_range(file->f_inode->i_mapping,
	3944	+ offset, offset + len - 1);
	3945	+ if (rc) {
	3946	+ kref_put(&ctx->refcount, cifs_aio_ctx_release);
	3947	+ return -EAGAIN;
	3948	+ }
	3949	+ }
3398	3950
3399	3951	/* grab a lock here due to read response handlers can access ctx */
3400	3952	mutex_lock(&ctx->aio_mutex);
..	..	@@ -3437,6 +3989,16 @@
3437	3989	return rc;
3438	3990	}
3439	3991
	3992	+ssize_t cifs_direct_readv(struct kiocb iocb, struct iov_iter to)
	3993	+{
	3994	+ return __cifs_readv(iocb, to, true);
	3995	+}
	3996	+
	3997	+ssize_t cifs_user_readv(struct kiocb iocb, struct iov_iter to)
	3998	+{
	3999	+ return __cifs_readv(iocb, to, false);
	4000	+}
	4001	+
3440	4002	ssize_t
3441	4003	cifs_strict_readv(struct kiocb iocb, struct iov_iter to)
3442	4004	{
..	..	@@ -3471,7 +4033,7 @@
3471	4033	down_read(&cinode->lock_sem);
3472	4034	if (!cifs_find_lock_conflict(cfile, iocb->ki_pos, iov_iter_count(to),
3473	4035	tcon->ses->server->vals->shared_lock_type,
3474		- NULL, CIFS_READ_OP))
	4036	+ 0, NULL, CIFS_READ_OP))
3475	4037	rc = generic_file_read_iter(iocb, to);
3476	4038	up_read(&cinode->lock_sem);
3477	4039	return rc;
..	..	@@ -3491,7 +4053,7 @@
3491	4053	unsigned int xid;
3492	4054	char *cur_offset;
3493	4055	struct cifsFileInfo *open_file;
3494		- struct cifs_io_parms io_parms;
	4056	+ struct cifs_io_parms io_parms = {0};
3495	4057	int buf_type = CIFS_NO_BUFFER;
3496	4058	__u32 pid;
3497	4059
..	..	@@ -3508,7 +4070,7 @@
3508	4070	}
3509	4071	open_file = file->private_data;
3510	4072	tcon = tlink_tcon(open_file->tlink);
3511		- server = tcon->ses->server;
	4073	+ server = cifs_pick_channel(tcon->ses);
3512	4074
3513	4075	if (!server->ops->sync_read) {
3514	4076	free_xid(xid);
..	..	@@ -3547,6 +4109,7 @@
3547	4109	io_parms.tcon = tcon;
3548	4110	io_parms.offset = *offset;
3549	4111	io_parms.length = current_read_size;
	4112	+ io_parms.server = server;
3550	4113	rc = server->ops->sync_read(xid, &open_file->fid, &io_parms,
3551	4114	&bytes_read, &cur_offset,
3552	4115	&buf_type);
..	..	@@ -3635,7 +4198,7 @@
3635	4198	for (i = 0; i < rdata->nr_pages; i++) {
3636	4199	struct page *page = rdata->pages[i];
3637	4200
3638		- lru_cache_add_file(page);
	4201	+ lru_cache_add(page);
3639	4202
3640	4203	if (rdata->result == 0 \|\|
3641	4204	(rdata->result == -EAGAIN && got_bytes)) {
..	..	@@ -3705,7 +4268,7 @@
3705	4268	* fill them until the writes are flushed.
3706	4269	*/
3707	4270	zero_user(page, 0, PAGE_SIZE);
3708		- lru_cache_add_file(page);
	4271	+ lru_cache_add(page);
3709	4272	flush_dcache_page(page);
3710	4273	SetPageUptodate(page);
3711	4274	unlock_page(page);
..	..	@@ -3715,7 +4278,7 @@
3715	4278	continue;
3716	4279	} else {
3717	4280	/* no need to hold page hostage */
3718		- lru_cache_add_file(page);
	4281	+ lru_cache_add(page);
3719	4282	unlock_page(page);
3720	4283	put_page(page);
3721	4284	rdata->pages[i] = NULL;
..	..	@@ -3739,7 +4302,7 @@
3739	4302	rdata->got_bytes += result;
3740	4303	}
3741	4304
3742		- return rdata->got_bytes > 0 && result != -ECONNABORTED ?
	4305	+ return result != -ECONNABORTED && rdata->got_bytes > 0 ?
3743	4306	rdata->got_bytes : result;
3744	4307	}
3745	4308
..	..	@@ -3770,7 +4333,7 @@
3770	4333
3771	4334	INIT_LIST_HEAD(tmplist);
3772	4335
3773		- page = list_entry(page_list->prev, struct page, lru);
	4336	+ page = lru_to_page(page_list);
3774	4337
3775	4338	/*
3776	4339	* Lock the page and put it in the cache. Since no one else
..	..	@@ -3828,7 +4391,9 @@
3828	4391	struct cifs_sb_info *cifs_sb = CIFS_FILE_SB(file);
3829	4392	struct TCP_Server_Info *server;
3830	4393	pid_t pid;
	4394	+ unsigned int xid;
3831	4395
	4396	+ xid = get_xid();
3832	4397	/*
3833	4398	* Reads as many pages as possible from fscache. Returns -ENOBUFS
3834	4399	* immediately if the cookie is negative
..	..	@@ -3838,8 +4403,10 @@
3838	4403	*/
3839	4404	rc = cifs_readpages_from_fscache(mapping->host, mapping, page_list,
3840	4405	&num_pages);
3841		- if (rc == 0)
	4406	+ if (rc == 0) {
	4407	+ free_xid(xid);
3842	4408	return rc;
	4409	+ }
3843	4410
3844	4411	if (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_RWPIDFORWARD)
3845	4412	pid = open_file->pid;
..	..	@@ -3847,7 +4414,7 @@
3847	4414	pid = current->tgid;
3848	4415
3849	4416	rc = 0;
3850		- server = tlink_tcon(open_file->tlink)->ses->server;
	4417	+ server = cifs_pick_channel(tlink_tcon(open_file->tlink)->ses);
3851	4418
3852	4419	cifs_dbg(FYI, "%s: file=%p mapping=%p num_pages=%u\n",
3853	4420	__func__, file, mapping, num_pages);
..	..	@@ -3868,10 +4435,19 @@
3868	4435	loff_t offset;
3869	4436	struct page page, tpage;
3870	4437	struct cifs_readdata *rdata;
3871		- unsigned credits;
	4438	+ struct cifs_credits credits_on_stack;
	4439	+ struct cifs_credits *credits = &credits_on_stack;
	4440	+
	4441	+ if (open_file->invalidHandle) {
	4442	+ rc = cifs_reopen_file(open_file, true);
	4443	+ if (rc == -EAGAIN)
	4444	+ continue;
	4445	+ else if (rc)
	4446	+ break;
	4447	+ }
3872	4448
3873	4449	rc = server->ops->wait_mtu_credits(server, cifs_sb->rsize,
3874		- &rsize, &credits);
	4450	+ &rsize, credits);
3875	4451	if (rc)
3876	4452	break;
3877	4453
..	..	@@ -3883,6 +4459,7 @@
3883	4459	*/
3884	4460	if (unlikely(rsize < PAGE_SIZE)) {
3885	4461	add_credits_and_wake_if(server, credits, 0);
	4462	+ free_xid(xid);
3886	4463	return 0;
3887	4464	}
3888	4465
..	..	@@ -3899,7 +4476,7 @@
3899	4476	/* best to give up if we're out of mem */
3900	4477	list_for_each_entry_safe(page, tpage, &tmplist, lru) {
3901	4478	list_del(&page->lru);
3902		- lru_cache_add_file(page);
	4479	+ lru_cache_add(page);
3903	4480	unlock_page(page);
3904	4481	put_page(page);
3905	4482	}
..	..	@@ -3909,6 +4486,7 @@
3909	4486	}
3910	4487
3911	4488	rdata->cfile = cifsFileInfo_get(open_file);
	4489	+ rdata->server = server;
3912	4490	rdata->mapping = mapping;
3913	4491	rdata->offset = offset;
3914	4492	rdata->bytes = bytes;
..	..	@@ -3917,21 +4495,27 @@
3917	4495	rdata->tailsz = PAGE_SIZE;
3918	4496	rdata->read_into_pages = cifs_readpages_read_into_pages;
3919	4497	rdata->copy_into_pages = cifs_readpages_copy_into_pages;
3920		- rdata->credits = credits;
	4498	+ rdata->credits = credits_on_stack;
3921	4499
3922	4500	list_for_each_entry_safe(page, tpage, &tmplist, lru) {
3923	4501	list_del(&page->lru);
3924	4502	rdata->pages[rdata->nr_pages++] = page;
3925	4503	}
3926	4504
3927		- if (!rdata->cfile->invalidHandle \|\|
3928		- !(rc = cifs_reopen_file(rdata->cfile, true)))
3929		- rc = server->ops->async_readv(rdata);
	4505	+ rc = adjust_credits(server, &rdata->credits, rdata->bytes);
	4506	+
	4507	+ if (!rc) {
	4508	+ if (rdata->cfile->invalidHandle)
	4509	+ rc = -EAGAIN;
	4510	+ else
	4511	+ rc = server->ops->async_readv(rdata);
	4512	+ }
	4513	+
3930	4514	if (rc) {
3931		- add_credits_and_wake_if(server, rdata->credits, 0);
	4515	+ add_credits_and_wake_if(server, &rdata->credits, 0);
3932	4516	for (i = 0; i < rdata->nr_pages; i++) {
3933	4517	page = rdata->pages[i];
3934		- lru_cache_add_file(page);
	4518	+ lru_cache_add(page);
3935	4519	unlock_page(page);
3936	4520	put_page(page);
3937	4521	}
..	..	@@ -3948,6 +4532,7 @@
3948	4532	* allocator.
3949	4533	*/
3950	4534	cifs_fscache_readpages_cancel(mapping->host, page_list);
	4535	+ free_xid(xid);
3951	4536	return rc;
3952	4537	}
3953	4538
..	..	@@ -3975,8 +4560,12 @@
3975	4560	else
3976	4561	cifs_dbg(FYI, "Bytes read %d\n", rc);
3977	4562
3978		- file_inode(file)->i_atime =
3979		- current_time(file_inode(file));
	4563	+ /* we do not want atime to be less than mtime, it broke some apps */
	4564	+ file_inode(file)->i_atime = current_time(file_inode(file));
	4565	+ if (timespec64_compare(&(file_inode(file)->i_atime), &(file_inode(file)->i_mtime)))
	4566	+ file_inode(file)->i_atime = file_inode(file)->i_mtime;
	4567	+ else
	4568	+ file_inode(file)->i_atime = current_time(file_inode(file));
3980	4569
3981	4570	if (PAGE_SIZE > rc)
3982	4571	memset(read_data + rc, 0, PAGE_SIZE - rc);
..	..	@@ -3991,15 +4580,15 @@
3991	4580
3992	4581	io_error:
3993	4582	kunmap(page);
3994		- unlock_page(page);
3995	4583
3996	4584	read_complete:
	4585	+ unlock_page(page);
3997	4586	return rc;
3998	4587	}
3999	4588
4000	4589	static int cifs_readpage(struct file file, struct page page)
4001	4590	{
4002		- loff_t offset = (loff_t)page->index << PAGE_SHIFT;
	4591	+ loff_t offset = page_file_offset(page);
4003	4592	int rc = -EACCES;
4004	4593	unsigned int xid;
4005	4594
..	..	@@ -4236,7 +4825,7 @@
4236	4825	cinode);
4237	4826	cifs_dbg(FYI, "Oplock release rc = %d\n", rc);
4238	4827	}
4239		- _cifsFileInfo_put(cfile, false /* do not wait for ourself */);
	4828	+ _cifsFileInfo_put(cfile, false /* do not wait for ourself */, false);
4240	4829	cifs_done_oplock_break(cinode);
4241	4830	}
4242	4831
..	..	@@ -4259,6 +4848,60 @@
4259	4848	return -EINVAL;
4260	4849	}
4261	4850
	4851	+static int cifs_swap_activate(struct swap_info_struct *sis,
	4852	+ struct file swap_file, sector_t span)
	4853	+{
	4854	+ struct cifsFileInfo *cfile = swap_file->private_data;
	4855	+ struct inode *inode = swap_file->f_mapping->host;
	4856	+ unsigned long blocks;
	4857	+ long long isize;
	4858	+
	4859	+ cifs_dbg(FYI, "swap activate\n");
	4860	+
	4861	+ spin_lock(&inode->i_lock);
	4862	+ blocks = inode->i_blocks;
	4863	+ isize = inode->i_size;
	4864	+ spin_unlock(&inode->i_lock);
	4865	+ if (blocks*512 < isize) {
	4866	+ pr_warn("swap activate: swapfile has holes\n");
	4867	+ return -EINVAL;
	4868	+ }
	4869	+ *span = sis->pages;
	4870	+
	4871	+ pr_warn_once("Swap support over SMB3 is experimental\n");
	4872	+
	4873	+ /*
	4874	+ * TODO: consider adding ACL (or documenting how) to prevent other
	4875	+ * users (on this or other systems) from reading it
	4876	+ */
	4877	+
	4878	+
	4879	+ /* TODO: add sk_set_memalloc(inet) or similar */
	4880	+
	4881	+ if (cfile)
	4882	+ cfile->swapfile = true;
	4883	+ /*
	4884	+ * TODO: Since file already open, we can't open with DENY_ALL here
	4885	+ * but we could add call to grab a byte range lock to prevent others
	4886	+ * from reading or writing the file
	4887	+ */
	4888	+
	4889	+ return 0;
	4890	+}
	4891	+
	4892	+static void cifs_swap_deactivate(struct file *file)
	4893	+{
	4894	+ struct cifsFileInfo *cfile = file->private_data;
	4895	+
	4896	+ cifs_dbg(FYI, "swap deactivate\n");
	4897	+
	4898	+ /* TODO: undo sk_set_memalloc(inet) will eventually be needed */
	4899	+
	4900	+ if (cfile)
	4901	+ cfile->swapfile = false;
	4902	+
	4903	+ /* do we need to unpin (or unlock) the file */
	4904	+}
4262	4905
4263	4906	const struct address_space_operations cifs_addr_ops = {
4264	4907	.readpage = cifs_readpage,
..	..	@@ -4272,6 +4915,13 @@
4272	4915	.direct_IO = cifs_direct_io,
4273	4916	.invalidatepage = cifs_invalidate_page,
4274	4917	.launder_page = cifs_launder_page,
	4918	+ /*
	4919	+ * TODO: investigate and if useful we could add an cifs_migratePage
	4920	+ * helper (under an CONFIG_MIGRATION) in the future, and also
	4921	+ * investigate and add an is_dirty_writeback helper if needed
	4922	+ */
	4923	+ .swap_activate = cifs_swap_activate,
	4924	+ .swap_deactivate = cifs_swap_deactivate,
4275	4925	};
4276	4926
4277	4927	/*