aboutsummaryrefslogtreecommitdiffstats
path: root/fs
diff options
context:
space:
mode:
Diffstat (limited to 'fs')
-rw-r--r--fs/aio.c26
-rw-r--r--fs/exec.c12
-rw-r--r--fs/locks.c42
-rw-r--r--fs/namei.c95
-rw-r--r--fs/nfs/delegation.c6
-rw-r--r--fs/nfs/delegation.h16
-rw-r--r--fs/nfs/dir.c44
-rw-r--r--fs/nfs/file.c30
-rw-r--r--fs/nfs/inode.c17
-rw-r--r--fs/nfs/nfs3proc.c2
-rw-r--r--fs/nfs/nfs4_fs.h53
-rw-r--r--fs/nfs/nfs4proc.c577
-rw-r--r--fs/nfs/nfs4state.c165
-rw-r--r--fs/nfs/nfs4xdr.c62
-rw-r--r--fs/nfs/proc.c2
-rw-r--r--fs/open.c79
-rw-r--r--fs/proc/base.c12
-rw-r--r--fs/proc/nommu.c1
18 files changed, 824 insertions, 417 deletions
diff --git a/fs/aio.c b/fs/aio.c
index d6b1551342..9fe7216457 100644
--- a/fs/aio.c
+++ b/fs/aio.c
@@ -398,7 +398,7 @@ static struct kiocb fastcall *__aio_get_req(struct kioctx *ctx)
398 if (unlikely(!req)) 398 if (unlikely(!req))
399 return NULL; 399 return NULL;
400 400
401 req->ki_flags = 1 << KIF_LOCKED; 401 req->ki_flags = 0;
402 req->ki_users = 2; 402 req->ki_users = 2;
403 req->ki_key = 0; 403 req->ki_key = 0;
404 req->ki_ctx = ctx; 404 req->ki_ctx = ctx;
@@ -547,25 +547,6 @@ struct kioctx *lookup_ioctx(unsigned long ctx_id)
547 return ioctx; 547 return ioctx;
548} 548}
549 549
550static int lock_kiocb_action(void *param)
551{
552 schedule();
553 return 0;
554}
555
556static inline void lock_kiocb(struct kiocb *iocb)
557{
558 wait_on_bit_lock(&iocb->ki_flags, KIF_LOCKED, lock_kiocb_action,
559 TASK_UNINTERRUPTIBLE);
560}
561
562static inline void unlock_kiocb(struct kiocb *iocb)
563{
564 kiocbClearLocked(iocb);
565 smp_mb__after_clear_bit();
566 wake_up_bit(&iocb->ki_flags, KIF_LOCKED);
567}
568
569/* 550/*
570 * use_mm 551 * use_mm
571 * Makes the calling kernel thread take on the specified 552 * Makes the calling kernel thread take on the specified
@@ -796,9 +777,7 @@ static int __aio_run_iocbs(struct kioctx *ctx)
796 * Hold an extra reference while retrying i/o. 777 * Hold an extra reference while retrying i/o.
797 */ 778 */
798 iocb->ki_users++; /* grab extra reference */ 779 iocb->ki_users++; /* grab extra reference */
799 lock_kiocb(iocb);
800 aio_run_iocb(iocb); 780 aio_run_iocb(iocb);
801 unlock_kiocb(iocb);
802 if (__aio_put_req(ctx, iocb)) /* drop extra ref */ 781 if (__aio_put_req(ctx, iocb)) /* drop extra ref */
803 put_ioctx(ctx); 782 put_ioctx(ctx);
804 } 783 }
@@ -1542,7 +1521,6 @@ int fastcall io_submit_one(struct kioctx *ctx, struct iocb __user *user_iocb,
1542 1521
1543 spin_lock_irq(&ctx->ctx_lock); 1522 spin_lock_irq(&ctx->ctx_lock);
1544 aio_run_iocb(req); 1523 aio_run_iocb(req);
1545 unlock_kiocb(req);
1546 if (!list_empty(&ctx->run_list)) { 1524 if (!list_empty(&ctx->run_list)) {
1547 /* drain the run list */ 1525 /* drain the run list */
1548 while (__aio_run_iocbs(ctx)) 1526 while (__aio_run_iocbs(ctx))
@@ -1674,7 +1652,6 @@ asmlinkage long sys_io_cancel(aio_context_t ctx_id, struct iocb __user *iocb,
1674 if (NULL != cancel) { 1652 if (NULL != cancel) {
1675 struct io_event tmp; 1653 struct io_event tmp;
1676 pr_debug("calling cancel\n"); 1654 pr_debug("calling cancel\n");
1677 lock_kiocb(kiocb);
1678 memset(&tmp, 0, sizeof(tmp)); 1655 memset(&tmp, 0, sizeof(tmp));
1679 tmp.obj = (u64)(unsigned long)kiocb->ki_obj.user; 1656 tmp.obj = (u64)(unsigned long)kiocb->ki_obj.user;
1680 tmp.data = kiocb->ki_user_data; 1657 tmp.data = kiocb->ki_user_data;
@@ -1686,7 +1663,6 @@ asmlinkage long sys_io_cancel(aio_context_t ctx_id, struct iocb __user *iocb,
1686 if (copy_to_user(result, &tmp, sizeof(tmp))) 1663 if (copy_to_user(result, &tmp, sizeof(tmp)))
1687 ret = -EFAULT; 1664 ret = -EFAULT;
1688 } 1665 }
1689 unlock_kiocb(kiocb);
1690 } else 1666 } else
1691 ret = -EINVAL; 1667 ret = -EINVAL;
1692 1668
diff --git a/fs/exec.c b/fs/exec.c
index a04a575ad4..d2208f7c87 100644
--- a/fs/exec.c
+++ b/fs/exec.c
@@ -126,8 +126,7 @@ asmlinkage long sys_uselib(const char __user * library)
126 struct nameidata nd; 126 struct nameidata nd;
127 int error; 127 int error;
128 128
129 nd.intent.open.flags = FMODE_READ; 129 error = __user_path_lookup_open(library, LOOKUP_FOLLOW, &nd, FMODE_READ);
130 error = __user_walk(library, LOOKUP_FOLLOW|LOOKUP_OPEN, &nd);
131 if (error) 130 if (error)
132 goto out; 131 goto out;
133 132
@@ -139,7 +138,7 @@ asmlinkage long sys_uselib(const char __user * library)
139 if (error) 138 if (error)
140 goto exit; 139 goto exit;
141 140
142 file = dentry_open(nd.dentry, nd.mnt, O_RDONLY); 141 file = nameidata_to_filp(&nd, O_RDONLY);
143 error = PTR_ERR(file); 142 error = PTR_ERR(file);
144 if (IS_ERR(file)) 143 if (IS_ERR(file))
145 goto out; 144 goto out;
@@ -167,6 +166,7 @@ asmlinkage long sys_uselib(const char __user * library)
167out: 166out:
168 return error; 167 return error;
169exit: 168exit:
169 release_open_intent(&nd);
170 path_release(&nd); 170 path_release(&nd);
171 goto out; 171 goto out;
172} 172}
@@ -490,8 +490,7 @@ struct file *open_exec(const char *name)
490 int err; 490 int err;
491 struct file *file; 491 struct file *file;
492 492
493 nd.intent.open.flags = FMODE_READ; 493 err = path_lookup_open(name, LOOKUP_FOLLOW, &nd, FMODE_READ);
494 err = path_lookup(name, LOOKUP_FOLLOW|LOOKUP_OPEN, &nd);
495 file = ERR_PTR(err); 494 file = ERR_PTR(err);
496 495
497 if (!err) { 496 if (!err) {
@@ -504,7 +503,7 @@ struct file *open_exec(const char *name)
504 err = -EACCES; 503 err = -EACCES;
505 file = ERR_PTR(err); 504 file = ERR_PTR(err);
506 if (!err) { 505 if (!err) {
507 file = dentry_open(nd.dentry, nd.mnt, O_RDONLY); 506 file = nameidata_to_filp(&nd, O_RDONLY);
508 if (!IS_ERR(file)) { 507 if (!IS_ERR(file)) {
509 err = deny_write_access(file); 508 err = deny_write_access(file);
510 if (err) { 509 if (err) {
@@ -516,6 +515,7 @@ out:
516 return file; 515 return file;
517 } 516 }
518 } 517 }
518 release_open_intent(&nd);
519 path_release(&nd); 519 path_release(&nd);
520 } 520 }
521 goto out; 521 goto out;
diff --git a/fs/locks.c b/fs/locks.c
index 7eb1d77b92..a1e8b22480 100644
--- a/fs/locks.c
+++ b/fs/locks.c
@@ -316,21 +316,22 @@ static int flock_to_posix_lock(struct file *filp, struct file_lock *fl,
316 /* POSIX-1996 leaves the case l->l_len < 0 undefined; 316 /* POSIX-1996 leaves the case l->l_len < 0 undefined;
317 POSIX-2001 defines it. */ 317 POSIX-2001 defines it. */
318 start += l->l_start; 318 start += l->l_start;
319 end = start + l->l_len - 1; 319 if (start < 0)
320 if (l->l_len < 0) { 320 return -EINVAL;
321 fl->fl_end = OFFSET_MAX;
322 if (l->l_len > 0) {
323 end = start + l->l_len - 1;
324 fl->fl_end = end;
325 } else if (l->l_len < 0) {
321 end = start - 1; 326 end = start - 1;
327 fl->fl_end = end;
322 start += l->l_len; 328 start += l->l_len;
329 if (start < 0)
330 return -EINVAL;
323 } 331 }
324
325 if (start < 0)
326 return -EINVAL;
327 if (l->l_len > 0 && end < 0)
328 return -EOVERFLOW;
329
330 fl->fl_start = start; /* we record the absolute position */ 332 fl->fl_start = start; /* we record the absolute position */
331 fl->fl_end = end; 333 if (fl->fl_end < fl->fl_start)
332 if (l->l_len == 0) 334 return -EOVERFLOW;
333 fl->fl_end = OFFSET_MAX;
334 335
335 fl->fl_owner = current->files; 336 fl->fl_owner = current->files;
336 fl->fl_pid = current->tgid; 337 fl->fl_pid = current->tgid;
@@ -362,14 +363,21 @@ static int flock64_to_posix_lock(struct file *filp, struct file_lock *fl,
362 return -EINVAL; 363 return -EINVAL;
363 } 364 }
364 365
365 if (((start += l->l_start) < 0) || (l->l_len < 0)) 366 start += l->l_start;
367 if (start < 0)
366 return -EINVAL; 368 return -EINVAL;
367 fl->fl_end = start + l->l_len - 1; 369 fl->fl_end = OFFSET_MAX;
368 if (l->l_len > 0 && fl->fl_end < 0) 370 if (l->l_len > 0) {
369 return -EOVERFLOW; 371 fl->fl_end = start + l->l_len - 1;
372 } else if (l->l_len < 0) {
373 fl->fl_end = start - 1;
374 start += l->l_len;
375 if (start < 0)
376 return -EINVAL;
377 }
370 fl->fl_start = start; /* we record the absolute position */ 378 fl->fl_start = start; /* we record the absolute position */
371 if (l->l_len == 0) 379 if (fl->fl_end < fl->fl_start)
372 fl->fl_end = OFFSET_MAX; 380 return -EOVERFLOW;
373 381
374 fl->fl_owner = current->files; 382 fl->fl_owner = current->files;
375 fl->fl_pid = current->tgid; 383 fl->fl_pid = current->tgid;
diff --git a/fs/namei.c b/fs/namei.c
index aa62dbda93..aaaa810362 100644
--- a/fs/namei.c
+++ b/fs/namei.c
@@ -28,6 +28,7 @@
28#include <linux/syscalls.h> 28#include <linux/syscalls.h>
29#include <linux/mount.h> 29#include <linux/mount.h>
30#include <linux/audit.h> 30#include <linux/audit.h>
31#include <linux/file.h>
31#include <asm/namei.h> 32#include <asm/namei.h>
32#include <asm/uaccess.h> 33#include <asm/uaccess.h>
33 34
@@ -317,6 +318,18 @@ void path_release_on_umount(struct nameidata *nd)
317 mntput_no_expire(nd->mnt); 318 mntput_no_expire(nd->mnt);
318} 319}
319 320
321/**
322 * release_open_intent - free up open intent resources
323 * @nd: pointer to nameidata
324 */
325void release_open_intent(struct nameidata *nd)
326{
327 if (nd->intent.open.file->f_dentry == NULL)
328 put_filp(nd->intent.open.file);
329 else
330 fput(nd->intent.open.file);
331}
332
320/* 333/*
321 * Internal lookup() using the new generic dcache. 334 * Internal lookup() using the new generic dcache.
322 * SMP-safe 335 * SMP-safe
@@ -750,6 +763,7 @@ static fastcall int __link_path_walk(const char * name, struct nameidata *nd)
750 struct qstr this; 763 struct qstr this;
751 unsigned int c; 764 unsigned int c;
752 765
766 nd->flags |= LOOKUP_CONTINUE;
753 err = exec_permission_lite(inode, nd); 767 err = exec_permission_lite(inode, nd);
754 if (err == -EAGAIN) { 768 if (err == -EAGAIN) {
755 err = permission(inode, MAY_EXEC, nd); 769 err = permission(inode, MAY_EXEC, nd);
@@ -802,7 +816,6 @@ static fastcall int __link_path_walk(const char * name, struct nameidata *nd)
802 if (err < 0) 816 if (err < 0)
803 break; 817 break;
804 } 818 }
805 nd->flags |= LOOKUP_CONTINUE;
806 /* This does the actual lookups.. */ 819 /* This does the actual lookups.. */
807 err = do_lookup(nd, &this, &next); 820 err = do_lookup(nd, &this, &next);
808 if (err) 821 if (err)
@@ -1052,6 +1065,70 @@ out:
1052 return retval; 1065 return retval;
1053} 1066}
1054 1067
1068static int __path_lookup_intent_open(const char *name, unsigned int lookup_flags,
1069 struct nameidata *nd, int open_flags, int create_mode)
1070{
1071 struct file *filp = get_empty_filp();
1072 int err;
1073
1074 if (filp == NULL)
1075 return -ENFILE;
1076 nd->intent.open.file = filp;
1077 nd->intent.open.flags = open_flags;
1078 nd->intent.open.create_mode = create_mode;
1079 err = path_lookup(name, lookup_flags|LOOKUP_OPEN, nd);
1080 if (IS_ERR(nd->intent.open.file)) {
1081 if (err == 0) {
1082 err = PTR_ERR(nd->intent.open.file);
1083 path_release(nd);
1084 }
1085 } else if (err != 0)
1086 release_open_intent(nd);
1087 return err;
1088}
1089
1090/**
1091 * path_lookup_open - lookup a file path with open intent
1092 * @name: pointer to file name
1093 * @lookup_flags: lookup intent flags
1094 * @nd: pointer to nameidata
1095 * @open_flags: open intent flags
1096 */
1097int path_lookup_open(const char *name, unsigned int lookup_flags,
1098 struct nameidata *nd, int open_flags)
1099{
1100 return __path_lookup_intent_open(name, lookup_flags, nd,
1101 open_flags, 0);
1102}
1103
1104/**
1105 * path_lookup_create - lookup a file path with open + create intent
1106 * @name: pointer to file name
1107 * @lookup_flags: lookup intent flags
1108 * @nd: pointer to nameidata
1109 * @open_flags: open intent flags
1110 * @create_mode: create intent flags
1111 */
1112int path_lookup_create(const char *name, unsigned int lookup_flags,
1113 struct nameidata *nd, int open_flags, int create_mode)
1114{
1115 return __path_lookup_intent_open(name, lookup_flags|LOOKUP_CREATE, nd,
1116 open_flags, create_mode);
1117}
1118
1119int __user_path_lookup_open(const char __user *name, unsigned int lookup_flags,
1120 struct nameidata *nd, int open_flags)
1121{
1122 char *tmp = getname(name);
1123 int err = PTR_ERR(tmp);
1124
1125 if (!IS_ERR(tmp)) {
1126 err = __path_lookup_intent_open(tmp, lookup_flags, nd, open_flags, 0);
1127 putname(tmp);
1128 }
1129 return err;
1130}
1131
1055/* 1132/*
1056 * Restricted form of lookup. Doesn't follow links, single-component only, 1133 * Restricted form of lookup. Doesn't follow links, single-component only,
1057 * needs parent already locked. Doesn't follow mounts. 1134 * needs parent already locked. Doesn't follow mounts.
@@ -1416,27 +1493,27 @@ int may_open(struct nameidata *nd, int acc_mode, int flag)
1416 */ 1493 */
1417int open_namei(const char * pathname, int flag, int mode, struct nameidata *nd) 1494int open_namei(const char * pathname, int flag, int mode, struct nameidata *nd)
1418{ 1495{
1419 int acc_mode, error = 0; 1496 int acc_mode, error;
1420 struct path path; 1497 struct path path;
1421 struct dentry *dir; 1498 struct dentry *dir;
1422 int count = 0; 1499 int count = 0;
1423 1500
1424 acc_mode = ACC_MODE(flag); 1501 acc_mode = ACC_MODE(flag);
1425 1502
1503 /* O_TRUNC implies we need access checks for write permissions */
1504 if (flag & O_TRUNC)
1505 acc_mode |= MAY_WRITE;
1506
1426 /* Allow the LSM permission hook to distinguish append 1507 /* Allow the LSM permission hook to distinguish append
1427 access from general write access. */ 1508 access from general write access. */
1428 if (flag & O_APPEND) 1509 if (flag & O_APPEND)
1429 acc_mode |= MAY_APPEND; 1510 acc_mode |= MAY_APPEND;
1430 1511
1431 /* Fill in the open() intent data */
1432 nd->intent.open.flags = flag;
1433 nd->intent.open.create_mode = mode;
1434
1435 /* 1512 /*
1436 * The simplest case - just a plain lookup. 1513 * The simplest case - just a plain lookup.
1437 */ 1514 */
1438 if (!(flag & O_CREAT)) { 1515 if (!(flag & O_CREAT)) {
1439 error = path_lookup(pathname, lookup_flags(flag)|LOOKUP_OPEN, nd); 1516 error = path_lookup_open(pathname, lookup_flags(flag), nd, flag);
1440 if (error) 1517 if (error)
1441 return error; 1518 return error;
1442 goto ok; 1519 goto ok;
@@ -1445,7 +1522,7 @@ int open_namei(const char * pathname, int flag, int mode, struct nameidata *nd)
1445 /* 1522 /*
1446 * Create - we need to know the parent. 1523 * Create - we need to know the parent.
1447 */ 1524 */
1448 error = path_lookup(pathname, LOOKUP_PARENT|LOOKUP_OPEN|LOOKUP_CREATE, nd); 1525 error = path_lookup_create(pathname, LOOKUP_PARENT, nd, flag, mode);
1449 if (error) 1526 if (error)
1450 return error; 1527 return error;
1451 1528
@@ -1520,6 +1597,8 @@ ok:
1520exit_dput: 1597exit_dput:
1521 dput_path(&path, nd); 1598 dput_path(&path, nd);
1522exit: 1599exit:
1600 if (!IS_ERR(nd->intent.open.file))
1601 release_open_intent(nd);
1523 path_release(nd); 1602 path_release(nd);
1524 return error; 1603 return error;
1525 1604
diff --git a/fs/nfs/delegation.c b/fs/nfs/delegation.c
index d7f7eb669d..44135af989 100644
--- a/fs/nfs/delegation.c
+++ b/fs/nfs/delegation.c
@@ -85,6 +85,10 @@ int nfs_inode_set_delegation(struct inode *inode, struct rpc_cred *cred, struct
85 struct nfs_delegation *delegation; 85 struct nfs_delegation *delegation;
86 int status = 0; 86 int status = 0;
87 87
88 /* Ensure we first revalidate the attributes and page cache! */
89 if ((nfsi->cache_validity & (NFS_INO_REVAL_PAGECACHE|NFS_INO_INVALID_ATTR)))
90 __nfs_revalidate_inode(NFS_SERVER(inode), inode);
91
88 delegation = nfs_alloc_delegation(); 92 delegation = nfs_alloc_delegation();
89 if (delegation == NULL) 93 if (delegation == NULL)
90 return -ENOMEM; 94 return -ENOMEM;
@@ -138,7 +142,7 @@ static void nfs_msync_inode(struct inode *inode)
138/* 142/*
139 * Basic procedure for returning a delegation to the server 143 * Basic procedure for returning a delegation to the server
140 */ 144 */
141int nfs_inode_return_delegation(struct inode *inode) 145int __nfs_inode_return_delegation(struct inode *inode)
142{ 146{
143 struct nfs4_client *clp = NFS_SERVER(inode)->nfs4_state; 147 struct nfs4_client *clp = NFS_SERVER(inode)->nfs4_state;
144 struct nfs_inode *nfsi = NFS_I(inode); 148 struct nfs_inode *nfsi = NFS_I(inode);
diff --git a/fs/nfs/delegation.h b/fs/nfs/delegation.h
index 3f6c45a29d..8017846b56 100644
--- a/fs/nfs/delegation.h
+++ b/fs/nfs/delegation.h
@@ -25,7 +25,7 @@ struct nfs_delegation {
25 25
26int nfs_inode_set_delegation(struct inode *inode, struct rpc_cred *cred, struct nfs_openres *res); 26int nfs_inode_set_delegation(struct inode *inode, struct rpc_cred *cred, struct nfs_openres *res);
27void nfs_inode_reclaim_delegation(struct inode *inode, struct rpc_cred *cred, struct nfs_openres *res); 27void nfs_inode_reclaim_delegation(struct inode *inode, struct rpc_cred *cred, struct nfs_openres *res);
28int nfs_inode_return_delegation(struct inode *inode); 28int __nfs_inode_return_delegation(struct inode *inode);
29int nfs_async_inode_return_delegation(struct inode *inode, const nfs4_stateid *stateid); 29int nfs_async_inode_return_delegation(struct inode *inode, const nfs4_stateid *stateid);
30 30
31struct inode *nfs_delegation_find_inode(struct nfs4_client *clp, const struct nfs_fh *fhandle); 31struct inode *nfs_delegation_find_inode(struct nfs4_client *clp, const struct nfs_fh *fhandle);
@@ -47,11 +47,25 @@ static inline int nfs_have_delegation(struct inode *inode, int flags)
47 return 1; 47 return 1;
48 return 0; 48 return 0;
49} 49}
50
51static inline int nfs_inode_return_delegation(struct inode *inode)
52{
53 int err = 0;
54
55 if (NFS_I(inode)->delegation != NULL)
56 err = __nfs_inode_return_delegation(inode);
57 return err;
58}
50#else 59#else
51static inline int nfs_have_delegation(struct inode *inode, int flags) 60static inline int nfs_have_delegation(struct inode *inode, int flags)
52{ 61{
53 return 0; 62 return 0;
54} 63}
64
65static inline int nfs_inode_return_delegation(struct inode *inode)
66{
67 return 0;
68}
55#endif 69#endif
56 70
57#endif 71#endif
diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c
index c70eabd6d1..eb50c19fc2 100644
--- a/fs/nfs/dir.c
+++ b/fs/nfs/dir.c
@@ -801,6 +801,7 @@ static int nfs_dentry_delete(struct dentry *dentry)
801 */ 801 */
802static void nfs_dentry_iput(struct dentry *dentry, struct inode *inode) 802static void nfs_dentry_iput(struct dentry *dentry, struct inode *inode)
803{ 803{
804 nfs_inode_return_delegation(inode);
804 if (dentry->d_flags & DCACHE_NFSFS_RENAMED) { 805 if (dentry->d_flags & DCACHE_NFSFS_RENAMED) {
805 lock_kernel(); 806 lock_kernel();
806 inode->i_nlink--; 807 inode->i_nlink--;
@@ -914,7 +915,6 @@ static int is_atomic_open(struct inode *dir, struct nameidata *nd)
914static struct dentry *nfs_atomic_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd) 915static struct dentry *nfs_atomic_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd)
915{ 916{
916 struct dentry *res = NULL; 917 struct dentry *res = NULL;
917 struct inode *inode = NULL;
918 int error; 918 int error;
919 919
920 /* Check that we are indeed trying to open this file */ 920 /* Check that we are indeed trying to open this file */
@@ -928,8 +928,10 @@ static struct dentry *nfs_atomic_lookup(struct inode *dir, struct dentry *dentry
928 dentry->d_op = NFS_PROTO(dir)->dentry_ops; 928 dentry->d_op = NFS_PROTO(dir)->dentry_ops;
929 929
930 /* Let vfs_create() deal with O_EXCL */ 930 /* Let vfs_create() deal with O_EXCL */
931 if (nd->intent.open.flags & O_EXCL) 931 if (nd->intent.open.flags & O_EXCL) {
932 goto no_entry; 932 d_add(dentry, NULL);
933 goto out;
934 }
933 935
934 /* Open the file on the server */ 936 /* Open the file on the server */
935 lock_kernel(); 937 lock_kernel();
@@ -943,32 +945,30 @@ static struct dentry *nfs_atomic_lookup(struct inode *dir, struct dentry *dentry
943 945
944 if (nd->intent.open.flags & O_CREAT) { 946 if (nd->intent.open.flags & O_CREAT) {
945 nfs_begin_data_update(dir); 947 nfs_begin_data_update(dir);
946 inode = nfs4_atomic_open(dir, dentry, nd); 948 res = nfs4_atomic_open(dir, dentry, nd);
947 nfs_end_data_update(dir); 949 nfs_end_data_update(dir);
948 } else 950 } else
949 inode = nfs4_atomic_open(dir, dentry, nd); 951 res = nfs4_atomic_open(dir, dentry, nd);
950 unlock_kernel(); 952 unlock_kernel();
951 if (IS_ERR(inode)) { 953 if (IS_ERR(res)) {
952 error = PTR_ERR(inode); 954 error = PTR_ERR(res);
953 switch (error) { 955 switch (error) {
954 /* Make a negative dentry */ 956 /* Make a negative dentry */
955 case -ENOENT: 957 case -ENOENT:
956 inode = NULL; 958 res = NULL;
957 break; 959 goto out;
958 /* This turned out not to be a regular file */ 960 /* This turned out not to be a regular file */
961 case -EISDIR:
962 case -ENOTDIR:
963 goto no_open;
959 case -ELOOP: 964 case -ELOOP:
960 if (!(nd->intent.open.flags & O_NOFOLLOW)) 965 if (!(nd->intent.open.flags & O_NOFOLLOW))
961 goto no_open; 966 goto no_open;
962 /* case -EISDIR: */
963 /* case -EINVAL: */ 967 /* case -EINVAL: */
964 default: 968 default:
965 res = ERR_PTR(error);
966 goto out; 969 goto out;
967 } 970 }
968 } 971 } else if (res != NULL)
969no_entry:
970 res = d_add_unique(dentry, inode);
971 if (res != NULL)
972 dentry = res; 972 dentry = res;
973 nfs_renew_times(dentry); 973 nfs_renew_times(dentry);
974 nfs_set_verifier(dentry, nfs_save_change_attribute(dir)); 974 nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
@@ -1012,7 +1012,7 @@ static int nfs_open_revalidate(struct dentry *dentry, struct nameidata *nd)
1012 */ 1012 */
1013 lock_kernel(); 1013 lock_kernel();
1014 verifier = nfs_save_change_attribute(dir); 1014 verifier = nfs_save_change_attribute(dir);
1015 ret = nfs4_open_revalidate(dir, dentry, openflags); 1015 ret = nfs4_open_revalidate(dir, dentry, openflags, nd);
1016 if (!ret) 1016 if (!ret)
1017 nfs_set_verifier(dentry, verifier); 1017 nfs_set_verifier(dentry, verifier);
1018 unlock_kernel(); 1018 unlock_kernel();
@@ -1135,7 +1135,7 @@ static int nfs_create(struct inode *dir, struct dentry *dentry, int mode,
1135 1135
1136 lock_kernel(); 1136 lock_kernel();
1137 nfs_begin_data_update(dir); 1137 nfs_begin_data_update(dir);
1138 error = NFS_PROTO(dir)->create(dir, dentry, &attr, open_flags); 1138 error = NFS_PROTO(dir)->create(dir, dentry, &attr, open_flags, nd);
1139 nfs_end_data_update(dir); 1139 nfs_end_data_update(dir);
1140 if (error != 0) 1140 if (error != 0)
1141 goto out_err; 1141 goto out_err;
@@ -1330,6 +1330,7 @@ static int nfs_safe_remove(struct dentry *dentry)
1330 1330
1331 nfs_begin_data_update(dir); 1331 nfs_begin_data_update(dir);
1332 if (inode != NULL) { 1332 if (inode != NULL) {
1333 nfs_inode_return_delegation(inode);
1333 nfs_begin_data_update(inode); 1334 nfs_begin_data_update(inode);
1334 error = NFS_PROTO(dir)->remove(dir, &dentry->d_name); 1335 error = NFS_PROTO(dir)->remove(dir, &dentry->d_name);
1335 /* The VFS may want to delete this inode */ 1336 /* The VFS may want to delete this inode */
@@ -1510,9 +1511,11 @@ static int nfs_rename(struct inode *old_dir, struct dentry *old_dentry,
1510 */ 1511 */
1511 if (!new_inode) 1512 if (!new_inode)
1512 goto go_ahead; 1513 goto go_ahead;
1513 if (S_ISDIR(new_inode->i_mode)) 1514 if (S_ISDIR(new_inode->i_mode)) {
1514 goto out; 1515 error = -EISDIR;
1515 else if (atomic_read(&new_dentry->d_count) > 2) { 1516 if (!S_ISDIR(old_inode->i_mode))
1517 goto out;
1518 } else if (atomic_read(&new_dentry->d_count) > 2) {
1516 int err; 1519 int err;
1517 /* copy the target dentry's name */ 1520 /* copy the target dentry's name */
1518 dentry = d_alloc(new_dentry->d_parent, 1521 dentry = d_alloc(new_dentry->d_parent,
@@ -1548,6 +1551,7 @@ go_ahead:
1548 nfs_wb_all(old_inode); 1551 nfs_wb_all(old_inode);
1549 shrink_dcache_parent(old_dentry); 1552 shrink_dcache_parent(old_dentry);
1550 } 1553 }
1554 nfs_inode_return_delegation(old_inode);
1551 1555
1552 if (new_inode) 1556 if (new_inode)
1553 d_delete(new_dentry); 1557 d_delete(new_dentry);
diff --git a/fs/nfs/file.c b/fs/nfs/file.c
index f6b9eda925..572d859348 100644
--- a/fs/nfs/file.c
+++ b/fs/nfs/file.c
@@ -137,7 +137,8 @@ static int nfs_revalidate_file(struct inode *inode, struct file *filp)
137 struct nfs_inode *nfsi = NFS_I(inode); 137 struct nfs_inode *nfsi = NFS_I(inode);
138 int retval = 0; 138 int retval = 0;
139 139
140 if ((nfsi->cache_validity & NFS_INO_REVAL_PAGECACHE) || nfs_attribute_timeout(inode)) 140 if ((nfsi->cache_validity & (NFS_INO_REVAL_PAGECACHE|NFS_INO_INVALID_ATTR))
141 || nfs_attribute_timeout(inode))
141 retval = __nfs_revalidate_inode(NFS_SERVER(inode), inode); 142 retval = __nfs_revalidate_inode(NFS_SERVER(inode), inode);
142 nfs_revalidate_mapping(inode, filp->f_mapping); 143 nfs_revalidate_mapping(inode, filp->f_mapping);
143 return 0; 144 return 0;
@@ -375,22 +376,31 @@ out_swapfile:
375 376
376static int do_getlk(struct file *filp, int cmd, struct file_lock *fl) 377static int do_getlk(struct file *filp, int cmd, struct file_lock *fl)
377{ 378{
379 struct file_lock *cfl;
378 struct inode *inode = filp->f_mapping->host; 380 struct inode *inode = filp->f_mapping->host;
379 int status = 0; 381 int status = 0;
380 382
381 lock_kernel(); 383 lock_kernel();
382 /* Use local locking if mounted with "-onolock" */ 384 /* Try local locking first */
383 if (!(NFS_SERVER(inode)->flags & NFS_MOUNT_NONLM)) 385 cfl = posix_test_lock(filp, fl);
384 status = NFS_PROTO(inode)->lock(filp, cmd, fl); 386 if (cfl != NULL) {
385 else { 387 locks_copy_lock(fl, cfl);
386 struct file_lock *cfl = posix_test_lock(filp, fl); 388 goto out;
387
388 fl->fl_type = F_UNLCK;
389 if (cfl != NULL)
390 memcpy(fl, cfl, sizeof(*fl));
391 } 389 }
390
391 if (nfs_have_delegation(inode, FMODE_READ))
392 goto out_noconflict;
393
394 if (NFS_SERVER(inode)->flags & NFS_MOUNT_NONLM)
395 goto out_noconflict;
396
397 status = NFS_PROTO(inode)->lock(filp, cmd, fl);
398out:
392 unlock_kernel(); 399 unlock_kernel();
393 return status; 400 return status;
401out_noconflict:
402 fl->fl_type = F_UNLCK;
403 goto out;
394} 404}
395 405
396static int do_vfs_lock(struct file *file, struct file_lock *fl) 406static int do_vfs_lock(struct file *file, struct file_lock *fl)
diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c
index 358d8ef8c0..65d5ab45dd 100644
--- a/fs/nfs/inode.c
+++ b/fs/nfs/inode.c
@@ -853,6 +853,11 @@ nfs_setattr(struct dentry *dentry, struct iattr *attr)
853 filemap_fdatawait(inode->i_mapping); 853 filemap_fdatawait(inode->i_mapping);
854 nfs_wb_all(inode); 854 nfs_wb_all(inode);
855 } 855 }
856 /*
857 * Return any delegations if we're going to change ACLs
858 */
859 if ((attr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0)
860 nfs_inode_return_delegation(inode);
856 error = NFS_PROTO(inode)->setattr(dentry, &fattr, attr); 861 error = NFS_PROTO(inode)->setattr(dentry, &fattr, attr);
857 if (error == 0) 862 if (error == 0)
858 nfs_refresh_inode(inode, &fattr); 863 nfs_refresh_inode(inode, &fattr);
@@ -909,12 +914,10 @@ static int nfs_wait_on_inode(struct inode *inode)
909 sigset_t oldmask; 914 sigset_t oldmask;
910 int error; 915 int error;
911 916
912 atomic_inc(&inode->i_count);
913 rpc_clnt_sigmask(clnt, &oldmask); 917 rpc_clnt_sigmask(clnt, &oldmask);
914 error = wait_on_bit_lock(&nfsi->flags, NFS_INO_REVALIDATING, 918 error = wait_on_bit_lock(&nfsi->flags, NFS_INO_REVALIDATING,
915 nfs_wait_schedule, TASK_INTERRUPTIBLE); 919 nfs_wait_schedule, TASK_INTERRUPTIBLE);
916 rpc_clnt_sigunmask(clnt, &oldmask); 920 rpc_clnt_sigunmask(clnt, &oldmask);
917 iput(inode);
918 921
919 return error; 922 return error;
920} 923}
@@ -1258,10 +1261,6 @@ int nfs_refresh_inode(struct inode *inode, struct nfs_fattr *fattr)
1258 loff_t cur_size, new_isize; 1261 loff_t cur_size, new_isize;
1259 int data_unstable; 1262 int data_unstable;
1260 1263
1261 /* Do we hold a delegation? */
1262 if (nfs_have_delegation(inode, FMODE_READ))
1263 return 0;
1264
1265 spin_lock(&inode->i_lock); 1264 spin_lock(&inode->i_lock);
1266 1265
1267 /* Are we in the process of updating data on the server? */ 1266 /* Are we in the process of updating data on the server? */
@@ -1382,7 +1381,8 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr, unsign
1382 nfsi->read_cache_jiffies = fattr->timestamp; 1381 nfsi->read_cache_jiffies = fattr->timestamp;
1383 1382
1384 /* Are we racing with known updates of the metadata on the server? */ 1383 /* Are we racing with known updates of the metadata on the server? */
1385 data_unstable = ! nfs_verify_change_attribute(inode, verifier); 1384 data_unstable = ! (nfs_verify_change_attribute(inode, verifier) ||
1385 (nfsi->cache_validity & NFS_INO_REVAL_PAGECACHE));
1386 1386
1387 /* Check if our cached file size is stale */ 1387 /* Check if our cached file size is stale */
1388 new_isize = nfs_size_to_loff_t(fattr->size); 1388 new_isize = nfs_size_to_loff_t(fattr->size);
@@ -1676,8 +1676,7 @@ static void nfs4_clear_inode(struct inode *inode)
1676 struct nfs_inode *nfsi = NFS_I(inode); 1676 struct nfs_inode *nfsi = NFS_I(inode);
1677 1677
1678 /* If we are holding a delegation, return it! */ 1678 /* If we are holding a delegation, return it! */
1679 if (nfsi->delegation != NULL) 1679 nfs_inode_return_delegation(inode);
1680 nfs_inode_return_delegation(inode);
1681 /* First call standard NFS clear_inode() code */ 1680 /* First call standard NFS clear_inode() code */
1682 nfs_clear_inode(inode); 1681 nfs_clear_inode(inode);
1683 /* Now clear out any remaining state */ 1682 /* Now clear out any remaining state */
diff --git a/fs/nfs/nfs3proc.c b/fs/nfs/nfs3proc.c
index edc9551404..df80477c5a 100644
--- a/fs/nfs/nfs3proc.c
+++ b/fs/nfs/nfs3proc.c
@@ -299,7 +299,7 @@ static int nfs3_proc_commit(struct nfs_write_data *cdata)
299 */ 299 */
300static int 300static int
301nfs3_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr, 301nfs3_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
302 int flags) 302 int flags, struct nameidata *nd)
303{ 303{
304 struct nfs_fh fhandle; 304 struct nfs_fh fhandle;
305 struct nfs_fattr fattr; 305 struct nfs_fattr fattr;
diff --git a/fs/nfs/nfs4_fs.h b/fs/nfs/nfs4_fs.h
index ec1a22d7b8..8a37881990 100644
--- a/fs/nfs/nfs4_fs.h
+++ b/fs/nfs/nfs4_fs.h
@@ -93,25 +93,50 @@ struct nfs4_client {
93}; 93};
94 94
95/* 95/*
96 * struct rpc_sequence ensures that RPC calls are sent in the exact
97 * order that they appear on the list.
98 */
99struct rpc_sequence {
100 struct rpc_wait_queue wait; /* RPC call delay queue */
101 spinlock_t lock; /* Protects the list */
102 struct list_head list; /* Defines sequence of RPC calls */
103};
104
105#define NFS_SEQID_CONFIRMED 1
106struct nfs_seqid_counter {
107 struct rpc_sequence *sequence;
108 int flags;
109 u32 counter;
110};
111
112struct nfs_seqid {
113 struct list_head list;
114 struct nfs_seqid_counter *sequence;
115 struct rpc_task *task;
116};
117
118static inline void nfs_confirm_seqid(struct nfs_seqid_counter *seqid, int status)
119{
120 if (seqid_mutating_err(-status))
121 seqid->flags |= NFS_SEQID_CONFIRMED;
122}
123
124/*
96 * NFS4 state_owners and lock_owners are simply labels for ordered 125 * NFS4 state_owners and lock_owners are simply labels for ordered
97 * sequences of RPC calls. Their sole purpose is to provide once-only 126 * sequences of RPC calls. Their sole purpose is to provide once-only
98 * semantics by allowing the server to identify replayed requests. 127 * semantics by allowing the server to identify replayed requests.
99 *
100 * The ->so_sema is held during all state_owner seqid-mutating operations:
101 * OPEN, OPEN_DOWNGRADE, and CLOSE. Its purpose is to properly serialize
102 * so_seqid.
103 */ 128 */
104struct nfs4_state_owner { 129struct nfs4_state_owner {
105 struct list_head so_list; /* per-clientid list of state_owners */ 130 struct list_head so_list; /* per-clientid list of state_owners */
106 struct nfs4_client *so_client; 131 struct nfs4_client *so_client;
107 u32 so_id; /* 32-bit identifier, unique */ 132 u32 so_id; /* 32-bit identifier, unique */
108 struct semaphore so_sema;
109 u32 so_seqid; /* protected by so_sema */
110 atomic_t so_count; 133 atomic_t so_count;
111 134
112 struct rpc_cred *so_cred; /* Associated cred */ 135 struct rpc_cred *so_cred; /* Associated cred */
113 struct list_head so_states; 136 struct list_head so_states;
114 struct list_head so_delegations; 137 struct list_head so_delegations;
138 struct nfs_seqid_counter so_seqid;
139 struct rpc_sequence so_sequence;
115}; 140};
116 141
117/* 142/*
@@ -132,7 +157,7 @@ struct nfs4_lock_state {
132 fl_owner_t ls_owner; /* POSIX lock owner */ 157 fl_owner_t ls_owner; /* POSIX lock owner */
133#define NFS_LOCK_INITIALIZED 1 158#define NFS_LOCK_INITIALIZED 1
134 int ls_flags; 159 int ls_flags;
135 u32 ls_seqid; 160 struct nfs_seqid_counter ls_seqid;
136 u32 ls_id; 161 u32 ls_id;
137 nfs4_stateid ls_stateid; 162 nfs4_stateid ls_stateid;
138 atomic_t ls_count; 163 atomic_t ls_count;
@@ -153,7 +178,6 @@ struct nfs4_state {
153 struct inode *inode; /* Pointer to the inode */ 178 struct inode *inode; /* Pointer to the inode */
154 179
155 unsigned long flags; /* Do we hold any locks? */ 180 unsigned long flags; /* Do we hold any locks? */
156 struct semaphore lock_sema; /* Serializes file locking operations */
157 spinlock_t state_lock; /* Protects the lock_states list */ 181 spinlock_t state_lock; /* Protects the lock_states list */
158 182
159 nfs4_stateid stateid; 183 nfs4_stateid stateid;
@@ -191,8 +215,8 @@ extern int nfs4_proc_setclientid_confirm(struct nfs4_client *);
191extern int nfs4_proc_async_renew(struct nfs4_client *); 215extern int nfs4_proc_async_renew(struct nfs4_client *);
192extern int nfs4_proc_renew(struct nfs4_client *); 216extern int nfs4_proc_renew(struct nfs4_client *);
193extern int nfs4_do_close(struct inode *inode, struct nfs4_state *state, mode_t mode); 217extern int nfs4_do_close(struct inode *inode, struct nfs4_state *state, mode_t mode);
194extern struct inode *nfs4_atomic_open(struct inode *, struct dentry *, struct nameidata *); 218extern struct dentry *nfs4_atomic_open(struct inode *, struct dentry *, struct nameidata *);
195extern int nfs4_open_revalidate(struct inode *, struct dentry *, int); 219extern int nfs4_open_revalidate(struct inode *, struct dentry *, int, struct nameidata *);
196 220
197extern struct nfs4_state_recovery_ops nfs4_reboot_recovery_ops; 221extern struct nfs4_state_recovery_ops nfs4_reboot_recovery_ops;
198extern struct nfs4_state_recovery_ops nfs4_network_partition_recovery_ops; 222extern struct nfs4_state_recovery_ops nfs4_network_partition_recovery_ops;
@@ -224,12 +248,17 @@ extern struct nfs4_state * nfs4_get_open_state(struct inode *, struct nfs4_state
224extern void nfs4_put_open_state(struct nfs4_state *); 248extern void nfs4_put_open_state(struct nfs4_state *);
225extern void nfs4_close_state(struct nfs4_state *, mode_t); 249extern void nfs4_close_state(struct nfs4_state *, mode_t);
226extern struct nfs4_state *nfs4_find_state(struct inode *, struct rpc_cred *, mode_t mode); 250extern struct nfs4_state *nfs4_find_state(struct inode *, struct rpc_cred *, mode_t mode);
227extern void nfs4_increment_seqid(int status, struct nfs4_state_owner *sp);
228extern void nfs4_schedule_state_recovery(struct nfs4_client *); 251extern void nfs4_schedule_state_recovery(struct nfs4_client *);
252extern void nfs4_put_lock_state(struct nfs4_lock_state *lsp);
229extern int nfs4_set_lock_state(struct nfs4_state *state, struct file_lock *fl); 253extern int nfs4_set_lock_state(struct nfs4_state *state, struct file_lock *fl);
230extern void nfs4_increment_lock_seqid(int status, struct nfs4_lock_state *ls);
231extern void nfs4_copy_stateid(nfs4_stateid *, struct nfs4_state *, fl_owner_t); 254extern void nfs4_copy_stateid(nfs4_stateid *, struct nfs4_state *, fl_owner_t);
232 255
256extern struct nfs_seqid *nfs_alloc_seqid(struct nfs_seqid_counter *counter);
257extern int nfs_wait_on_sequence(struct nfs_seqid *seqid, struct rpc_task *task);
258extern void nfs_increment_open_seqid(int status, struct nfs_seqid *seqid);
259extern void nfs_increment_lock_seqid(int status, struct nfs_seqid *seqid);
260extern void nfs_free_seqid(struct nfs_seqid *seqid);
261
233extern const nfs4_stateid zero_stateid; 262extern const nfs4_stateid zero_stateid;
234 263
235/* nfs4xdr.c */ 264/* nfs4xdr.c */
diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c
index 9701ca8c94..ff378126cc 100644
--- a/fs/nfs/nfs4proc.c
+++ b/fs/nfs/nfs4proc.c
@@ -47,6 +47,7 @@
47#include <linux/nfs_page.h> 47#include <linux/nfs_page.h>
48#include <linux/smp_lock.h> 48#include <linux/smp_lock.h>
49#include <linux/namei.h> 49#include <linux/namei.h>
50#include <linux/mount.h>
50 51
51#include "nfs4_fs.h" 52#include "nfs4_fs.h"
52#include "delegation.h" 53#include "delegation.h"
@@ -56,10 +57,11 @@
56#define NFS4_POLL_RETRY_MIN (1*HZ) 57#define NFS4_POLL_RETRY_MIN (1*HZ)
57#define NFS4_POLL_RETRY_MAX (15*HZ) 58#define NFS4_POLL_RETRY_MAX (15*HZ)
58 59
60static int _nfs4_proc_open_confirm(struct rpc_clnt *clnt, const struct nfs_fh *fh, struct nfs4_state_owner *sp, nfs4_stateid *stateid, struct nfs_seqid *seqid);
59static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *); 61static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
60static int nfs4_async_handle_error(struct rpc_task *, struct nfs_server *); 62static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *);
61static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry); 63static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry);
62static int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception); 64static int nfs4_handle_exception(const struct nfs_server *server, int errorcode, struct nfs4_exception *exception);
63extern u32 *nfs4_decode_dirent(u32 *p, struct nfs_entry *entry, int plus); 65extern u32 *nfs4_decode_dirent(u32 *p, struct nfs_entry *entry, int plus);
64extern struct rpc_procinfo nfs4_procedures[]; 66extern struct rpc_procinfo nfs4_procedures[];
65 67
@@ -189,6 +191,21 @@ static void update_changeattr(struct inode *inode, struct nfs4_change_info *cinf
189 nfsi->change_attr = cinfo->after; 191 nfsi->change_attr = cinfo->after;
190} 192}
191 193
194/* Helper for asynchronous RPC calls */
195static int nfs4_call_async(struct rpc_clnt *clnt, rpc_action tk_begin,
196 rpc_action tk_exit, void *calldata)
197{
198 struct rpc_task *task;
199
200 if (!(task = rpc_new_task(clnt, tk_exit, RPC_TASK_ASYNC)))
201 return -ENOMEM;
202
203 task->tk_calldata = calldata;
204 task->tk_action = tk_begin;
205 rpc_execute(task);
206 return 0;
207}
208
192static void update_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, int open_flags) 209static void update_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, int open_flags)
193{ 210{
194 struct inode *inode = state->inode; 211 struct inode *inode = state->inode;
@@ -209,7 +226,6 @@ static void update_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid,
209/* 226/*
210 * OPEN_RECLAIM: 227 * OPEN_RECLAIM:
211 * reclaim state on the server after a reboot. 228 * reclaim state on the server after a reboot.
212 * Assumes caller is holding the sp->so_sem
213 */ 229 */
214static int _nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state) 230static int _nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
215{ 231{
@@ -218,7 +234,6 @@ static int _nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *st
218 struct nfs_delegation *delegation = NFS_I(inode)->delegation; 234 struct nfs_delegation *delegation = NFS_I(inode)->delegation;
219 struct nfs_openargs o_arg = { 235 struct nfs_openargs o_arg = {
220 .fh = NFS_FH(inode), 236 .fh = NFS_FH(inode),
221 .seqid = sp->so_seqid,
222 .id = sp->so_id, 237 .id = sp->so_id,
223 .open_flags = state->state, 238 .open_flags = state->state,
224 .clientid = server->nfs4_state->cl_clientid, 239 .clientid = server->nfs4_state->cl_clientid,
@@ -245,8 +260,13 @@ static int _nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *st
245 } 260 }
246 o_arg.u.delegation_type = delegation->type; 261 o_arg.u.delegation_type = delegation->type;
247 } 262 }
263 o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid);
264 if (o_arg.seqid == NULL)
265 return -ENOMEM;
248 status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR); 266 status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR);
249 nfs4_increment_seqid(status, sp); 267 /* Confirm the sequence as being established */
268 nfs_confirm_seqid(&sp->so_seqid, status);
269 nfs_increment_open_seqid(status, o_arg.seqid);
250 if (status == 0) { 270 if (status == 0) {
251 memcpy(&state->stateid, &o_res.stateid, sizeof(state->stateid)); 271 memcpy(&state->stateid, &o_res.stateid, sizeof(state->stateid));
252 if (o_res.delegation_type != 0) { 272 if (o_res.delegation_type != 0) {
@@ -256,6 +276,7 @@ static int _nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *st
256 nfs_async_inode_return_delegation(inode, &o_res.stateid); 276 nfs_async_inode_return_delegation(inode, &o_res.stateid);
257 } 277 }
258 } 278 }
279 nfs_free_seqid(o_arg.seqid);
259 clear_bit(NFS_DELEGATED_STATE, &state->flags); 280 clear_bit(NFS_DELEGATED_STATE, &state->flags);
260 /* Ensure we update the inode attributes */ 281 /* Ensure we update the inode attributes */
261 NFS_CACHEINV(inode); 282 NFS_CACHEINV(inode);
@@ -302,23 +323,35 @@ static int _nfs4_open_delegation_recall(struct dentry *dentry, struct nfs4_state
302 }; 323 };
303 int status = 0; 324 int status = 0;
304 325
305 down(&sp->so_sema);
306 if (!test_bit(NFS_DELEGATED_STATE, &state->flags)) 326 if (!test_bit(NFS_DELEGATED_STATE, &state->flags))
307 goto out; 327 goto out;
308 if (state->state == 0) 328 if (state->state == 0)
309 goto out; 329 goto out;
310 arg.seqid = sp->so_seqid; 330 arg.seqid = nfs_alloc_seqid(&sp->so_seqid);
331 status = -ENOMEM;
332 if (arg.seqid == NULL)
333 goto out;
311 arg.open_flags = state->state; 334 arg.open_flags = state->state;
312 memcpy(arg.u.delegation.data, state->stateid.data, sizeof(arg.u.delegation.data)); 335 memcpy(arg.u.delegation.data, state->stateid.data, sizeof(arg.u.delegation.data));
313 status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR); 336 status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR);
314 nfs4_increment_seqid(status, sp); 337 nfs_increment_open_seqid(status, arg.seqid);
338 if (status != 0)
339 goto out_free;
340 if(res.rflags & NFS4_OPEN_RESULT_CONFIRM) {
341 status = _nfs4_proc_open_confirm(server->client, NFS_FH(inode),
342 sp, &res.stateid, arg.seqid);
343 if (status != 0)
344 goto out_free;
345 }
346 nfs_confirm_seqid(&sp->so_seqid, 0);
315 if (status >= 0) { 347 if (status >= 0) {
316 memcpy(state->stateid.data, res.stateid.data, 348 memcpy(state->stateid.data, res.stateid.data,
317 sizeof(state->stateid.data)); 349 sizeof(state->stateid.data));
318 clear_bit(NFS_DELEGATED_STATE, &state->flags); 350 clear_bit(NFS_DELEGATED_STATE, &state->flags);
319 } 351 }
352out_free:
353 nfs_free_seqid(arg.seqid);
320out: 354out:
321 up(&sp->so_sema);
322 dput(parent); 355 dput(parent);
323 return status; 356 return status;
324} 357}
@@ -345,11 +378,11 @@ int nfs4_open_delegation_recall(struct dentry *dentry, struct nfs4_state *state)
345 return err; 378 return err;
346} 379}
347 380
348static inline int _nfs4_proc_open_confirm(struct rpc_clnt *clnt, const struct nfs_fh *fh, struct nfs4_state_owner *sp, nfs4_stateid *stateid) 381static int _nfs4_proc_open_confirm(struct rpc_clnt *clnt, const struct nfs_fh *fh, struct nfs4_state_owner *sp, nfs4_stateid *stateid, struct nfs_seqid *seqid)
349{ 382{
350 struct nfs_open_confirmargs arg = { 383 struct nfs_open_confirmargs arg = {
351 .fh = fh, 384 .fh = fh,
352 .seqid = sp->so_seqid, 385 .seqid = seqid,
353 .stateid = *stateid, 386 .stateid = *stateid,
354 }; 387 };
355 struct nfs_open_confirmres res; 388 struct nfs_open_confirmres res;
@@ -362,7 +395,9 @@ static inline int _nfs4_proc_open_confirm(struct rpc_clnt *clnt, const struct nf
362 int status; 395 int status;
363 396
364 status = rpc_call_sync(clnt, &msg, RPC_TASK_NOINTR); 397 status = rpc_call_sync(clnt, &msg, RPC_TASK_NOINTR);
365 nfs4_increment_seqid(status, sp); 398 /* Confirm the sequence as being established */
399 nfs_confirm_seqid(&sp->so_seqid, status);
400 nfs_increment_open_seqid(status, seqid);
366 if (status >= 0) 401 if (status >= 0)
367 memcpy(stateid, &res.stateid, sizeof(*stateid)); 402 memcpy(stateid, &res.stateid, sizeof(*stateid));
368 return status; 403 return status;
@@ -380,21 +415,37 @@ static int _nfs4_proc_open(struct inode *dir, struct nfs4_state_owner *sp, stru
380 int status; 415 int status;
381 416
382 /* Update sequence id. The caller must serialize! */ 417 /* Update sequence id. The caller must serialize! */
383 o_arg->seqid = sp->so_seqid;
384 o_arg->id = sp->so_id; 418 o_arg->id = sp->so_id;
385 o_arg->clientid = sp->so_client->cl_clientid; 419 o_arg->clientid = sp->so_client->cl_clientid;
386 420
387 status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR); 421 status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR);
388 nfs4_increment_seqid(status, sp); 422 if (status == 0) {
423 /* OPEN on anything except a regular file is disallowed in NFSv4 */
424 switch (o_res->f_attr->mode & S_IFMT) {
425 case S_IFREG:
426 break;
427 case S_IFLNK:
428 status = -ELOOP;
429 break;
430 case S_IFDIR:
431 status = -EISDIR;
432 break;
433 default:
434 status = -ENOTDIR;
435 }
436 }
437
438 nfs_increment_open_seqid(status, o_arg->seqid);
389 if (status != 0) 439 if (status != 0)
390 goto out; 440 goto out;
391 update_changeattr(dir, &o_res->cinfo); 441 update_changeattr(dir, &o_res->cinfo);
392 if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) { 442 if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
393 status = _nfs4_proc_open_confirm(server->client, &o_res->fh, 443 status = _nfs4_proc_open_confirm(server->client, &o_res->fh,
394 sp, &o_res->stateid); 444 sp, &o_res->stateid, o_arg->seqid);
395 if (status != 0) 445 if (status != 0)
396 goto out; 446 goto out;
397 } 447 }
448 nfs_confirm_seqid(&sp->so_seqid, 0);
398 if (!(o_res->f_attr->valid & NFS_ATTR_FATTR)) 449 if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
399 status = server->rpc_ops->getattr(server, &o_res->fh, o_res->f_attr); 450 status = server->rpc_ops->getattr(server, &o_res->fh, o_res->f_attr);
400out: 451out:
@@ -465,6 +516,10 @@ static int _nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *st
465 set_bit(NFS_DELEGATED_STATE, &state->flags); 516 set_bit(NFS_DELEGATED_STATE, &state->flags);
466 goto out; 517 goto out;
467 } 518 }
519 o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid);
520 status = -ENOMEM;
521 if (o_arg.seqid == NULL)
522 goto out;
468 status = _nfs4_proc_open(dir, sp, &o_arg, &o_res); 523 status = _nfs4_proc_open(dir, sp, &o_arg, &o_res);
469 if (status != 0) 524 if (status != 0)
470 goto out_nodeleg; 525 goto out_nodeleg;
@@ -490,6 +545,7 @@ static int _nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *st
490 nfs_inode_reclaim_delegation(inode, sp->so_cred, &o_res); 545 nfs_inode_reclaim_delegation(inode, sp->so_cred, &o_res);
491 } 546 }
492out_nodeleg: 547out_nodeleg:
548 nfs_free_seqid(o_arg.seqid);
493 clear_bit(NFS_DELEGATED_STATE, &state->flags); 549 clear_bit(NFS_DELEGATED_STATE, &state->flags);
494out: 550out:
495 dput(parent); 551 dput(parent);
@@ -564,7 +620,6 @@ static int _nfs4_open_delegated(struct inode *inode, int flags, struct rpc_cred
564 dprintk("%s: nfs4_get_state_owner failed!\n", __FUNCTION__); 620 dprintk("%s: nfs4_get_state_owner failed!\n", __FUNCTION__);
565 goto out_err; 621 goto out_err;
566 } 622 }
567 down(&sp->so_sema);
568 state = nfs4_get_open_state(inode, sp); 623 state = nfs4_get_open_state(inode, sp);
569 if (state == NULL) 624 if (state == NULL)
570 goto out_err; 625 goto out_err;
@@ -589,7 +644,6 @@ static int _nfs4_open_delegated(struct inode *inode, int flags, struct rpc_cred
589 set_bit(NFS_DELEGATED_STATE, &state->flags); 644 set_bit(NFS_DELEGATED_STATE, &state->flags);
590 update_open_stateid(state, &delegation->stateid, open_flags); 645 update_open_stateid(state, &delegation->stateid, open_flags);
591out_ok: 646out_ok:
592 up(&sp->so_sema);
593 nfs4_put_state_owner(sp); 647 nfs4_put_state_owner(sp);
594 up_read(&nfsi->rwsem); 648 up_read(&nfsi->rwsem);
595 up_read(&clp->cl_sem); 649 up_read(&clp->cl_sem);
@@ -600,11 +654,12 @@ out_err:
600 if (sp != NULL) { 654 if (sp != NULL) {
601 if (state != NULL) 655 if (state != NULL)
602 nfs4_put_open_state(state); 656 nfs4_put_open_state(state);
603 up(&sp->so_sema);
604 nfs4_put_state_owner(sp); 657 nfs4_put_state_owner(sp);
605 } 658 }
606 up_read(&nfsi->rwsem); 659 up_read(&nfsi->rwsem);
607 up_read(&clp->cl_sem); 660 up_read(&clp->cl_sem);
661 if (err != -EACCES)
662 nfs_inode_return_delegation(inode);
608 return err; 663 return err;
609} 664}
610 665
@@ -665,8 +720,10 @@ static int _nfs4_do_open(struct inode *dir, struct dentry *dentry, int flags, st
665 } else 720 } else
666 o_arg.u.attrs = sattr; 721 o_arg.u.attrs = sattr;
667 /* Serialization for the sequence id */ 722 /* Serialization for the sequence id */
668 down(&sp->so_sema);
669 723
724 o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid);
725 if (o_arg.seqid == NULL)
726 return -ENOMEM;
670 status = _nfs4_proc_open(dir, sp, &o_arg, &o_res); 727 status = _nfs4_proc_open(dir, sp, &o_arg, &o_res);
671 if (status != 0) 728 if (status != 0)
672 goto out_err; 729 goto out_err;
@@ -681,7 +738,7 @@ static int _nfs4_do_open(struct inode *dir, struct dentry *dentry, int flags, st
681 update_open_stateid(state, &o_res.stateid, flags); 738 update_open_stateid(state, &o_res.stateid, flags);
682 if (o_res.delegation_type != 0) 739 if (o_res.delegation_type != 0)
683 nfs_inode_set_delegation(inode, cred, &o_res); 740 nfs_inode_set_delegation(inode, cred, &o_res);
684 up(&sp->so_sema); 741 nfs_free_seqid(o_arg.seqid);
685 nfs4_put_state_owner(sp); 742 nfs4_put_state_owner(sp);
686 up_read(&clp->cl_sem); 743 up_read(&clp->cl_sem);
687 *res = state; 744 *res = state;
@@ -690,7 +747,7 @@ out_err:
690 if (sp != NULL) { 747 if (sp != NULL) {
691 if (state != NULL) 748 if (state != NULL)
692 nfs4_put_open_state(state); 749 nfs4_put_open_state(state);
693 up(&sp->so_sema); 750 nfs_free_seqid(o_arg.seqid);
694 nfs4_put_state_owner(sp); 751 nfs4_put_state_owner(sp);
695 } 752 }
696 /* Note: clp->cl_sem must be released before nfs4_put_open_state()! */ 753 /* Note: clp->cl_sem must be released before nfs4_put_open_state()! */
@@ -718,7 +775,7 @@ static struct nfs4_state *nfs4_do_open(struct inode *dir, struct dentry *dentry,
718 * It is actually a sign of a bug on the client or on the server. 775 * It is actually a sign of a bug on the client or on the server.
719 * 776 *
720 * If we receive a BAD_SEQID error in the particular case of 777 * If we receive a BAD_SEQID error in the particular case of
721 * doing an OPEN, we assume that nfs4_increment_seqid() will 778 * doing an OPEN, we assume that nfs_increment_open_seqid() will
722 * have unhashed the old state_owner for us, and that we can 779 * have unhashed the old state_owner for us, and that we can
723 * therefore safely retry using a new one. We should still warn 780 * therefore safely retry using a new one. We should still warn
724 * the user though... 781 * the user though...
@@ -728,6 +785,16 @@ static struct nfs4_state *nfs4_do_open(struct inode *dir, struct dentry *dentry,
728 exception.retry = 1; 785 exception.retry = 1;
729 continue; 786 continue;
730 } 787 }
788 /*
789 * BAD_STATEID on OPEN means that the server cancelled our
790 * state before it received the OPEN_CONFIRM.
791 * Recover by retrying the request as per the discussion
792 * on Page 181 of RFC3530.
793 */
794 if (status == -NFS4ERR_BAD_STATEID) {
795 exception.retry = 1;
796 continue;
797 }
731 res = ERR_PTR(nfs4_handle_exception(NFS_SERVER(dir), 798 res = ERR_PTR(nfs4_handle_exception(NFS_SERVER(dir),
732 status, &exception)); 799 status, &exception));
733 } while (exception.retry); 800 } while (exception.retry);
@@ -789,17 +856,27 @@ struct nfs4_closedata {
789 struct nfs_closeres res; 856 struct nfs_closeres res;
790}; 857};
791 858
859static void nfs4_free_closedata(struct nfs4_closedata *calldata)
860{
861 struct nfs4_state *state = calldata->state;
862 struct nfs4_state_owner *sp = state->owner;
863
864 nfs4_put_open_state(calldata->state);
865 nfs_free_seqid(calldata->arg.seqid);
866 nfs4_put_state_owner(sp);
867 kfree(calldata);
868}
869
792static void nfs4_close_done(struct rpc_task *task) 870static void nfs4_close_done(struct rpc_task *task)
793{ 871{
794 struct nfs4_closedata *calldata = (struct nfs4_closedata *)task->tk_calldata; 872 struct nfs4_closedata *calldata = (struct nfs4_closedata *)task->tk_calldata;
795 struct nfs4_state *state = calldata->state; 873 struct nfs4_state *state = calldata->state;
796 struct nfs4_state_owner *sp = state->owner;
797 struct nfs_server *server = NFS_SERVER(calldata->inode); 874 struct nfs_server *server = NFS_SERVER(calldata->inode);
798 875
799 /* hmm. we are done with the inode, and in the process of freeing 876 /* hmm. we are done with the inode, and in the process of freeing
800 * the state_owner. we keep this around to process errors 877 * the state_owner. we keep this around to process errors
801 */ 878 */
802 nfs4_increment_seqid(task->tk_status, sp); 879 nfs_increment_open_seqid(task->tk_status, calldata->arg.seqid);
803 switch (task->tk_status) { 880 switch (task->tk_status) {
804 case 0: 881 case 0:
805 memcpy(&state->stateid, &calldata->res.stateid, 882 memcpy(&state->stateid, &calldata->res.stateid,
@@ -817,24 +894,46 @@ static void nfs4_close_done(struct rpc_task *task)
817 } 894 }
818 } 895 }
819 state->state = calldata->arg.open_flags; 896 state->state = calldata->arg.open_flags;
820 nfs4_put_open_state(state); 897 nfs4_free_closedata(calldata);
821 up(&sp->so_sema);
822 nfs4_put_state_owner(sp);
823 up_read(&server->nfs4_state->cl_sem);
824 kfree(calldata);
825} 898}
826 899
827static inline int nfs4_close_call(struct rpc_clnt *clnt, struct nfs4_closedata *calldata) 900static void nfs4_close_begin(struct rpc_task *task)
828{ 901{
902 struct nfs4_closedata *calldata = (struct nfs4_closedata *)task->tk_calldata;
903 struct nfs4_state *state = calldata->state;
829 struct rpc_message msg = { 904 struct rpc_message msg = {
830 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE], 905 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
831 .rpc_argp = &calldata->arg, 906 .rpc_argp = &calldata->arg,
832 .rpc_resp = &calldata->res, 907 .rpc_resp = &calldata->res,
833 .rpc_cred = calldata->state->owner->so_cred, 908 .rpc_cred = state->owner->so_cred,
834 }; 909 };
835 if (calldata->arg.open_flags != 0) 910 int mode = 0;
911 int status;
912
913 status = nfs_wait_on_sequence(calldata->arg.seqid, task);
914 if (status != 0)
915 return;
916 /* Don't reorder reads */
917 smp_rmb();
918 /* Recalculate the new open mode in case someone reopened the file
919 * while we were waiting in line to be scheduled.
920 */
921 if (state->nreaders != 0)
922 mode |= FMODE_READ;
923 if (state->nwriters != 0)
924 mode |= FMODE_WRITE;
925 if (test_bit(NFS_DELEGATED_STATE, &state->flags))
926 state->state = mode;
927 if (mode == state->state) {
928 nfs4_free_closedata(calldata);
929 task->tk_exit = NULL;
930 rpc_exit(task, 0);
931 return;
932 }
933 if (mode != 0)
836 msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE]; 934 msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
837 return rpc_call_async(clnt, &msg, 0, nfs4_close_done, calldata); 935 calldata->arg.open_flags = mode;
936 rpc_call_setup(task, &msg, 0);
838} 937}
839 938
840/* 939/*
@@ -851,39 +950,52 @@ static inline int nfs4_close_call(struct rpc_clnt *clnt, struct nfs4_closedata *
851int nfs4_do_close(struct inode *inode, struct nfs4_state *state, mode_t mode) 950int nfs4_do_close(struct inode *inode, struct nfs4_state *state, mode_t mode)
852{ 951{
853 struct nfs4_closedata *calldata; 952 struct nfs4_closedata *calldata;
854 int status; 953 int status = -ENOMEM;
855 954
856 /* Tell caller we're done */ 955 calldata = kmalloc(sizeof(*calldata), GFP_KERNEL);
857 if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
858 state->state = mode;
859 return 0;
860 }
861 calldata = (struct nfs4_closedata *)kmalloc(sizeof(*calldata), GFP_KERNEL);
862 if (calldata == NULL) 956 if (calldata == NULL)
863 return -ENOMEM; 957 goto out;
864 calldata->inode = inode; 958 calldata->inode = inode;
865 calldata->state = state; 959 calldata->state = state;
866 calldata->arg.fh = NFS_FH(inode); 960 calldata->arg.fh = NFS_FH(inode);
961 calldata->arg.stateid = &state->stateid;
867 /* Serialization for the sequence id */ 962 /* Serialization for the sequence id */
868 calldata->arg.seqid = state->owner->so_seqid; 963 calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid);
869 calldata->arg.open_flags = mode; 964 if (calldata->arg.seqid == NULL)
870 memcpy(&calldata->arg.stateid, &state->stateid, 965 goto out_free_calldata;
871 sizeof(calldata->arg.stateid)); 966
872 status = nfs4_close_call(NFS_SERVER(inode)->client, calldata); 967 status = nfs4_call_async(NFS_SERVER(inode)->client, nfs4_close_begin,
873 /* 968 nfs4_close_done, calldata);
874 * Return -EINPROGRESS on success in order to indicate to the 969 if (status == 0)
875 * caller that an asynchronous RPC call has been launched, and 970 goto out;
876 * that it will release the semaphores on completion. 971
877 */ 972 nfs_free_seqid(calldata->arg.seqid);
878 return (status == 0) ? -EINPROGRESS : status; 973out_free_calldata:
974 kfree(calldata);
975out:
976 return status;
977}
978
979static void nfs4_intent_set_file(struct nameidata *nd, struct dentry *dentry, struct nfs4_state *state)
980{
981 struct file *filp;
982
983 filp = lookup_instantiate_filp(nd, dentry, NULL);
984 if (!IS_ERR(filp)) {
985 struct nfs_open_context *ctx;
986 ctx = (struct nfs_open_context *)filp->private_data;
987 ctx->state = state;
988 } else
989 nfs4_close_state(state, nd->intent.open.flags);
879} 990}
880 991
881struct inode * 992struct dentry *
882nfs4_atomic_open(struct inode *dir, struct dentry *dentry, struct nameidata *nd) 993nfs4_atomic_open(struct inode *dir, struct dentry *dentry, struct nameidata *nd)
883{ 994{
884 struct iattr attr; 995 struct iattr attr;
885 struct rpc_cred *cred; 996 struct rpc_cred *cred;
886 struct nfs4_state *state; 997 struct nfs4_state *state;
998 struct dentry *res;
887 999
888 if (nd->flags & LOOKUP_CREATE) { 1000 if (nd->flags & LOOKUP_CREATE) {
889 attr.ia_mode = nd->intent.open.create_mode; 1001 attr.ia_mode = nd->intent.open.create_mode;
@@ -897,16 +1009,23 @@ nfs4_atomic_open(struct inode *dir, struct dentry *dentry, struct nameidata *nd)
897 1009
898 cred = rpcauth_lookupcred(NFS_SERVER(dir)->client->cl_auth, 0); 1010 cred = rpcauth_lookupcred(NFS_SERVER(dir)->client->cl_auth, 0);
899 if (IS_ERR(cred)) 1011 if (IS_ERR(cred))
900 return (struct inode *)cred; 1012 return (struct dentry *)cred;
901 state = nfs4_do_open(dir, dentry, nd->intent.open.flags, &attr, cred); 1013 state = nfs4_do_open(dir, dentry, nd->intent.open.flags, &attr, cred);
902 put_rpccred(cred); 1014 put_rpccred(cred);
903 if (IS_ERR(state)) 1015 if (IS_ERR(state)) {
904 return (struct inode *)state; 1016 if (PTR_ERR(state) == -ENOENT)
905 return state->inode; 1017 d_add(dentry, NULL);
1018 return (struct dentry *)state;
1019 }
1020 res = d_add_unique(dentry, state->inode);
1021 if (res != NULL)
1022 dentry = res;
1023 nfs4_intent_set_file(nd, dentry, state);
1024 return res;
906} 1025}
907 1026
908int 1027int
909nfs4_open_revalidate(struct inode *dir, struct dentry *dentry, int openflags) 1028nfs4_open_revalidate(struct inode *dir, struct dentry *dentry, int openflags, struct nameidata *nd)
910{ 1029{
911 struct rpc_cred *cred; 1030 struct rpc_cred *cred;
912 struct nfs4_state *state; 1031 struct nfs4_state *state;
@@ -919,18 +1038,30 @@ nfs4_open_revalidate(struct inode *dir, struct dentry *dentry, int openflags)
919 if (IS_ERR(state)) 1038 if (IS_ERR(state))
920 state = nfs4_do_open(dir, dentry, openflags, NULL, cred); 1039 state = nfs4_do_open(dir, dentry, openflags, NULL, cred);
921 put_rpccred(cred); 1040 put_rpccred(cred);
922 if (state == ERR_PTR(-ENOENT) && dentry->d_inode == 0) 1041 if (IS_ERR(state)) {
923 return 1; 1042 switch (PTR_ERR(state)) {
924 if (IS_ERR(state)) 1043 case -EPERM:
925 return 0; 1044 case -EACCES:
1045 case -EDQUOT:
1046 case -ENOSPC:
1047 case -EROFS:
1048 lookup_instantiate_filp(nd, (struct dentry *)state, NULL);
1049 return 1;
1050 case -ENOENT:
1051 if (dentry->d_inode == NULL)
1052 return 1;
1053 }
1054 goto out_drop;
1055 }
926 inode = state->inode; 1056 inode = state->inode;
1057 iput(inode);
927 if (inode == dentry->d_inode) { 1058 if (inode == dentry->d_inode) {
928 iput(inode); 1059 nfs4_intent_set_file(nd, dentry, state);
929 return 1; 1060 return 1;
930 } 1061 }
931 d_drop(dentry);
932 nfs4_close_state(state, openflags); 1062 nfs4_close_state(state, openflags);
933 iput(inode); 1063out_drop:
1064 d_drop(dentry);
934 return 0; 1065 return 0;
935} 1066}
936 1067
@@ -1431,7 +1562,7 @@ static int nfs4_proc_commit(struct nfs_write_data *cdata)
1431 1562
1432static int 1563static int
1433nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr, 1564nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
1434 int flags) 1565 int flags, struct nameidata *nd)
1435{ 1566{
1436 struct nfs4_state *state; 1567 struct nfs4_state *state;
1437 struct rpc_cred *cred; 1568 struct rpc_cred *cred;
@@ -1453,13 +1584,13 @@ nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
1453 struct nfs_fattr fattr; 1584 struct nfs_fattr fattr;
1454 status = nfs4_do_setattr(NFS_SERVER(dir), &fattr, 1585 status = nfs4_do_setattr(NFS_SERVER(dir), &fattr,
1455 NFS_FH(state->inode), sattr, state); 1586 NFS_FH(state->inode), sattr, state);
1456 if (status == 0) { 1587 if (status == 0)
1457 nfs_setattr_update_inode(state->inode, sattr); 1588 nfs_setattr_update_inode(state->inode, sattr);
1458 goto out; 1589 }
1459 } 1590 if (status == 0 && nd != NULL && (nd->flags & LOOKUP_OPEN))
1460 } else if (flags != 0) 1591 nfs4_intent_set_file(nd, dentry, state);
1461 goto out; 1592 else
1462 nfs4_close_state(state, flags); 1593 nfs4_close_state(state, flags);
1463out: 1594out:
1464 return status; 1595 return status;
1465} 1596}
@@ -2106,65 +2237,6 @@ nfs4_proc_renew(struct nfs4_client *clp)
2106 return 0; 2237 return 0;
2107} 2238}
2108 2239
2109/*
2110 * We will need to arrange for the VFS layer to provide an atomic open.
2111 * Until then, this open method is prone to inefficiency and race conditions
2112 * due to the lookup, potential create, and open VFS calls from sys_open()
2113 * placed on the wire.
2114 */
2115static int
2116nfs4_proc_file_open(struct inode *inode, struct file *filp)
2117{
2118 struct dentry *dentry = filp->f_dentry;
2119 struct nfs_open_context *ctx;
2120 struct nfs4_state *state = NULL;
2121 struct rpc_cred *cred;
2122 int status = -ENOMEM;
2123
2124 dprintk("nfs4_proc_file_open: starting on (%.*s/%.*s)\n",
2125 (int)dentry->d_parent->d_name.len,
2126 dentry->d_parent->d_name.name,
2127 (int)dentry->d_name.len, dentry->d_name.name);
2128
2129
2130 /* Find our open stateid */
2131 cred = rpcauth_lookupcred(NFS_SERVER(inode)->client->cl_auth, 0);
2132 if (IS_ERR(cred))
2133 return PTR_ERR(cred);
2134 ctx = alloc_nfs_open_context(dentry, cred);
2135 put_rpccred(cred);
2136 if (unlikely(ctx == NULL))
2137 return -ENOMEM;
2138 status = -EIO; /* ERACE actually */
2139 state = nfs4_find_state(inode, cred, filp->f_mode);
2140 if (unlikely(state == NULL))
2141 goto no_state;
2142 ctx->state = state;
2143 nfs4_close_state(state, filp->f_mode);
2144 ctx->mode = filp->f_mode;
2145 nfs_file_set_open_context(filp, ctx);
2146 put_nfs_open_context(ctx);
2147 if (filp->f_mode & FMODE_WRITE)
2148 nfs_begin_data_update(inode);
2149 return 0;
2150no_state:
2151 printk(KERN_WARNING "NFS: v4 raced in function %s\n", __FUNCTION__);
2152 put_nfs_open_context(ctx);
2153 return status;
2154}
2155
2156/*
2157 * Release our state
2158 */
2159static int
2160nfs4_proc_file_release(struct inode *inode, struct file *filp)
2161{
2162 if (filp->f_mode & FMODE_WRITE)
2163 nfs_end_data_update(inode);
2164 nfs_file_clear_open_context(filp);
2165 return 0;
2166}
2167
2168static inline int nfs4_server_supports_acls(struct nfs_server *server) 2240static inline int nfs4_server_supports_acls(struct nfs_server *server)
2169{ 2241{
2170 return (server->caps & NFS_CAP_ACLS) 2242 return (server->caps & NFS_CAP_ACLS)
@@ -2345,6 +2417,7 @@ static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen
2345 2417
2346 if (!nfs4_server_supports_acls(server)) 2418 if (!nfs4_server_supports_acls(server))
2347 return -EOPNOTSUPP; 2419 return -EOPNOTSUPP;
2420 nfs_inode_return_delegation(inode);
2348 buf_to_pages(buf, buflen, arg.acl_pages, &arg.acl_pgbase); 2421 buf_to_pages(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
2349 ret = rpc_call_sync(NFS_SERVER(inode)->client, &msg, 0); 2422 ret = rpc_call_sync(NFS_SERVER(inode)->client, &msg, 0);
2350 if (ret == 0) 2423 if (ret == 0)
@@ -2353,7 +2426,7 @@ static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen
2353} 2426}
2354 2427
2355static int 2428static int
2356nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server) 2429nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server)
2357{ 2430{
2358 struct nfs4_client *clp = server->nfs4_state; 2431 struct nfs4_client *clp = server->nfs4_state;
2359 2432
@@ -2431,7 +2504,7 @@ static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
2431/* This is the error handling routine for processes that are allowed 2504/* This is the error handling routine for processes that are allowed
2432 * to sleep. 2505 * to sleep.
2433 */ 2506 */
2434int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception) 2507int nfs4_handle_exception(const struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
2435{ 2508{
2436 struct nfs4_client *clp = server->nfs4_state; 2509 struct nfs4_client *clp = server->nfs4_state;
2437 int ret = errorcode; 2510 int ret = errorcode;
@@ -2632,7 +2705,6 @@ static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock
2632 2705
2633 down_read(&clp->cl_sem); 2706 down_read(&clp->cl_sem);
2634 nlo.clientid = clp->cl_clientid; 2707 nlo.clientid = clp->cl_clientid;
2635 down(&state->lock_sema);
2636 status = nfs4_set_lock_state(state, request); 2708 status = nfs4_set_lock_state(state, request);
2637 if (status != 0) 2709 if (status != 0)
2638 goto out; 2710 goto out;
@@ -2659,7 +2731,6 @@ static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock
2659 status = 0; 2731 status = 0;
2660 } 2732 }
2661out: 2733out:
2662 up(&state->lock_sema);
2663 up_read(&clp->cl_sem); 2734 up_read(&clp->cl_sem);
2664 return status; 2735 return status;
2665} 2736}
@@ -2696,79 +2767,149 @@ static int do_vfs_lock(struct file *file, struct file_lock *fl)
2696 return res; 2767 return res;
2697} 2768}
2698 2769
2699static int _nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request) 2770struct nfs4_unlockdata {
2771 struct nfs_lockargs arg;
2772 struct nfs_locku_opargs luargs;
2773 struct nfs_lockres res;
2774 struct nfs4_lock_state *lsp;
2775 struct nfs_open_context *ctx;
2776 atomic_t refcount;
2777 struct completion completion;
2778};
2779
2780static void nfs4_locku_release_calldata(struct nfs4_unlockdata *calldata)
2700{ 2781{
2701 struct inode *inode = state->inode; 2782 if (atomic_dec_and_test(&calldata->refcount)) {
2702 struct nfs_server *server = NFS_SERVER(inode); 2783 nfs_free_seqid(calldata->luargs.seqid);
2703 struct nfs4_client *clp = server->nfs4_state; 2784 nfs4_put_lock_state(calldata->lsp);
2704 struct nfs_lockargs arg = { 2785 put_nfs_open_context(calldata->ctx);
2705 .fh = NFS_FH(inode), 2786 kfree(calldata);
2706 .type = nfs4_lck_type(cmd, request), 2787 }
2707 .offset = request->fl_start, 2788}
2708 .length = nfs4_lck_length(request), 2789
2709 }; 2790static void nfs4_locku_complete(struct nfs4_unlockdata *calldata)
2710 struct nfs_lockres res = { 2791{
2711 .server = server, 2792 complete(&calldata->completion);
2712 }; 2793 nfs4_locku_release_calldata(calldata);
2794}
2795
2796static void nfs4_locku_done(struct rpc_task *task)
2797{
2798 struct nfs4_unlockdata *calldata = (struct nfs4_unlockdata *)task->tk_calldata;
2799
2800 nfs_increment_lock_seqid(task->tk_status, calldata->luargs.seqid);
2801 switch (task->tk_status) {
2802 case 0:
2803 memcpy(calldata->lsp->ls_stateid.data,
2804 calldata->res.u.stateid.data,
2805 sizeof(calldata->lsp->ls_stateid.data));
2806 break;
2807 case -NFS4ERR_STALE_STATEID:
2808 case -NFS4ERR_EXPIRED:
2809 nfs4_schedule_state_recovery(calldata->res.server->nfs4_state);
2810 break;
2811 default:
2812 if (nfs4_async_handle_error(task, calldata->res.server) == -EAGAIN) {
2813 rpc_restart_call(task);
2814 return;
2815 }
2816 }
2817 nfs4_locku_complete(calldata);
2818}
2819
2820static void nfs4_locku_begin(struct rpc_task *task)
2821{
2822 struct nfs4_unlockdata *calldata = (struct nfs4_unlockdata *)task->tk_calldata;
2713 struct rpc_message msg = { 2823 struct rpc_message msg = {
2714 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU], 2824 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
2715 .rpc_argp = &arg, 2825 .rpc_argp = &calldata->arg,
2716 .rpc_resp = &res, 2826 .rpc_resp = &calldata->res,
2717 .rpc_cred = state->owner->so_cred, 2827 .rpc_cred = calldata->lsp->ls_state->owner->so_cred,
2718 }; 2828 };
2829 int status;
2830
2831 status = nfs_wait_on_sequence(calldata->luargs.seqid, task);
2832 if (status != 0)
2833 return;
2834 if ((calldata->lsp->ls_flags & NFS_LOCK_INITIALIZED) == 0) {
2835 nfs4_locku_complete(calldata);
2836 task->tk_exit = NULL;
2837 rpc_exit(task, 0);
2838 return;
2839 }
2840 rpc_call_setup(task, &msg, 0);
2841}
2842
2843static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
2844{
2845 struct nfs4_unlockdata *calldata;
2846 struct inode *inode = state->inode;
2847 struct nfs_server *server = NFS_SERVER(inode);
2719 struct nfs4_lock_state *lsp; 2848 struct nfs4_lock_state *lsp;
2720 struct nfs_locku_opargs luargs;
2721 int status; 2849 int status;
2722 2850
2723 down_read(&clp->cl_sem);
2724 down(&state->lock_sema);
2725 status = nfs4_set_lock_state(state, request); 2851 status = nfs4_set_lock_state(state, request);
2726 if (status != 0) 2852 if (status != 0)
2727 goto out; 2853 return status;
2728 lsp = request->fl_u.nfs4_fl.owner; 2854 lsp = request->fl_u.nfs4_fl.owner;
2729 /* We might have lost the locks! */ 2855 /* We might have lost the locks! */
2730 if ((lsp->ls_flags & NFS_LOCK_INITIALIZED) == 0) 2856 if ((lsp->ls_flags & NFS_LOCK_INITIALIZED) == 0)
2731 goto out; 2857 return 0;
2732 luargs.seqid = lsp->ls_seqid; 2858 calldata = kmalloc(sizeof(*calldata), GFP_KERNEL);
2733 memcpy(&luargs.stateid, &lsp->ls_stateid, sizeof(luargs.stateid)); 2859 if (calldata == NULL)
2734 arg.u.locku = &luargs; 2860 return -ENOMEM;
2735 status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR); 2861 calldata->luargs.seqid = nfs_alloc_seqid(&lsp->ls_seqid);
2736 nfs4_increment_lock_seqid(status, lsp); 2862 if (calldata->luargs.seqid == NULL) {
2737 2863 kfree(calldata);
2738 if (status == 0) 2864 return -ENOMEM;
2739 memcpy(&lsp->ls_stateid, &res.u.stateid, 2865 }
2740 sizeof(lsp->ls_stateid)); 2866 calldata->luargs.stateid = &lsp->ls_stateid;
2741out: 2867 calldata->arg.fh = NFS_FH(inode);
2742 up(&state->lock_sema); 2868 calldata->arg.type = nfs4_lck_type(cmd, request);
2869 calldata->arg.offset = request->fl_start;
2870 calldata->arg.length = nfs4_lck_length(request);
2871 calldata->arg.u.locku = &calldata->luargs;
2872 calldata->res.server = server;
2873 calldata->lsp = lsp;
2874 atomic_inc(&lsp->ls_count);
2875
2876 /* Ensure we don't close file until we're done freeing locks! */
2877 calldata->ctx = get_nfs_open_context((struct nfs_open_context*)request->fl_file->private_data);
2878
2879 atomic_set(&calldata->refcount, 2);
2880 init_completion(&calldata->completion);
2881
2882 status = nfs4_call_async(NFS_SERVER(inode)->client, nfs4_locku_begin,
2883 nfs4_locku_done, calldata);
2743 if (status == 0) 2884 if (status == 0)
2744 do_vfs_lock(request->fl_file, request); 2885 wait_for_completion_interruptible(&calldata->completion);
2745 up_read(&clp->cl_sem); 2886 do_vfs_lock(request->fl_file, request);
2887 nfs4_locku_release_calldata(calldata);
2746 return status; 2888 return status;
2747} 2889}
2748 2890
2749static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
2750{
2751 struct nfs4_exception exception = { };
2752 int err;
2753
2754 do {
2755 err = nfs4_handle_exception(NFS_SERVER(state->inode),
2756 _nfs4_proc_unlck(state, cmd, request),
2757 &exception);
2758 } while (exception.retry);
2759 return err;
2760}
2761
2762static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *request, int reclaim) 2891static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *request, int reclaim)
2763{ 2892{
2764 struct inode *inode = state->inode; 2893 struct inode *inode = state->inode;
2765 struct nfs_server *server = NFS_SERVER(inode); 2894 struct nfs_server *server = NFS_SERVER(inode);
2766 struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner; 2895 struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
2896 struct nfs_lock_opargs largs = {
2897 .lock_stateid = &lsp->ls_stateid,
2898 .open_stateid = &state->stateid,
2899 .lock_owner = {
2900 .clientid = server->nfs4_state->cl_clientid,
2901 .id = lsp->ls_id,
2902 },
2903 .reclaim = reclaim,
2904 };
2767 struct nfs_lockargs arg = { 2905 struct nfs_lockargs arg = {
2768 .fh = NFS_FH(inode), 2906 .fh = NFS_FH(inode),
2769 .type = nfs4_lck_type(cmd, request), 2907 .type = nfs4_lck_type(cmd, request),
2770 .offset = request->fl_start, 2908 .offset = request->fl_start,
2771 .length = nfs4_lck_length(request), 2909 .length = nfs4_lck_length(request),
2910 .u = {
2911 .lock = &largs,
2912 },
2772 }; 2913 };
2773 struct nfs_lockres res = { 2914 struct nfs_lockres res = {
2774 .server = server, 2915 .server = server,
@@ -2779,53 +2920,39 @@ static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *r
2779 .rpc_resp = &res, 2920 .rpc_resp = &res,
2780 .rpc_cred = state->owner->so_cred, 2921 .rpc_cred = state->owner->so_cred,
2781 }; 2922 };
2782 struct nfs_lock_opargs largs = { 2923 int status = -ENOMEM;
2783 .reclaim = reclaim,
2784 .new_lock_owner = 0,
2785 };
2786 int status;
2787 2924
2788 if (!(lsp->ls_flags & NFS_LOCK_INITIALIZED)) { 2925 largs.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid);
2926 if (largs.lock_seqid == NULL)
2927 return -ENOMEM;
2928 if (!(lsp->ls_seqid.flags & NFS_SEQID_CONFIRMED)) {
2789 struct nfs4_state_owner *owner = state->owner; 2929 struct nfs4_state_owner *owner = state->owner;
2790 struct nfs_open_to_lock otl = { 2930
2791 .lock_owner = { 2931 largs.open_seqid = nfs_alloc_seqid(&owner->so_seqid);
2792 .clientid = server->nfs4_state->cl_clientid, 2932 if (largs.open_seqid == NULL)
2793 }, 2933 goto out;
2794 };
2795
2796 otl.lock_seqid = lsp->ls_seqid;
2797 otl.lock_owner.id = lsp->ls_id;
2798 memcpy(&otl.open_stateid, &state->stateid, sizeof(otl.open_stateid));
2799 largs.u.open_lock = &otl;
2800 largs.new_lock_owner = 1; 2934 largs.new_lock_owner = 1;
2801 arg.u.lock = &largs;
2802 down(&owner->so_sema);
2803 otl.open_seqid = owner->so_seqid;
2804 status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR); 2935 status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR);
2805 /* increment open_owner seqid on success, and 2936 /* increment open seqid on success, and seqid mutating errors */
2806 * seqid mutating errors */ 2937 if (largs.new_lock_owner != 0) {
2807 nfs4_increment_seqid(status, owner); 2938 nfs_increment_open_seqid(status, largs.open_seqid);
2808 up(&owner->so_sema); 2939 if (status == 0)
2809 if (status == 0) { 2940 nfs_confirm_seqid(&lsp->ls_seqid, 0);
2810 lsp->ls_flags |= NFS_LOCK_INITIALIZED;
2811 lsp->ls_seqid++;
2812 } 2941 }
2813 } else { 2942 nfs_free_seqid(largs.open_seqid);
2814 struct nfs_exist_lock el = { 2943 } else
2815 .seqid = lsp->ls_seqid,
2816 };
2817 memcpy(&el.stateid, &lsp->ls_stateid, sizeof(el.stateid));
2818 largs.u.exist_lock = &el;
2819 arg.u.lock = &largs;
2820 status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR); 2944 status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR);
2821 /* increment seqid on success, and * seqid mutating errors*/ 2945 /* increment lock seqid on success, and seqid mutating errors*/
2822 nfs4_increment_lock_seqid(status, lsp); 2946 nfs_increment_lock_seqid(status, largs.lock_seqid);
2823 }
2824 /* save the returned stateid. */ 2947 /* save the returned stateid. */
2825 if (status == 0) 2948 if (status == 0) {
2826 memcpy(&lsp->ls_stateid, &res.u.stateid, sizeof(nfs4_stateid)); 2949 memcpy(lsp->ls_stateid.data, res.u.stateid.data,
2827 else if (status == -NFS4ERR_DENIED) 2950 sizeof(lsp->ls_stateid.data));
2951 lsp->ls_flags |= NFS_LOCK_INITIALIZED;
2952 } else if (status == -NFS4ERR_DENIED)
2828 status = -EAGAIN; 2953 status = -EAGAIN;
2954out:
2955 nfs_free_seqid(largs.lock_seqid);
2829 return status; 2956 return status;
2830} 2957}
2831 2958
@@ -2865,11 +2992,9 @@ static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock
2865 int status; 2992 int status;
2866 2993
2867 down_read(&clp->cl_sem); 2994 down_read(&clp->cl_sem);
2868 down(&state->lock_sema);
2869 status = nfs4_set_lock_state(state, request); 2995 status = nfs4_set_lock_state(state, request);
2870 if (status == 0) 2996 if (status == 0)
2871 status = _nfs4_do_setlk(state, cmd, request, 0); 2997 status = _nfs4_do_setlk(state, cmd, request, 0);
2872 up(&state->lock_sema);
2873 if (status == 0) { 2998 if (status == 0) {
2874 /* Note: we always want to sleep here! */ 2999 /* Note: we always want to sleep here! */
2875 request->fl_flags |= FL_SLEEP; 3000 request->fl_flags |= FL_SLEEP;
@@ -3024,8 +3149,8 @@ struct nfs_rpc_ops nfs_v4_clientops = {
3024 .read_setup = nfs4_proc_read_setup, 3149 .read_setup = nfs4_proc_read_setup,
3025 .write_setup = nfs4_proc_write_setup, 3150 .write_setup = nfs4_proc_write_setup,
3026 .commit_setup = nfs4_proc_commit_setup, 3151 .commit_setup = nfs4_proc_commit_setup,
3027 .file_open = nfs4_proc_file_open, 3152 .file_open = nfs_open,
3028 .file_release = nfs4_proc_file_release, 3153 .file_release = nfs_release,
3029 .lock = nfs4_proc_lock, 3154 .lock = nfs4_proc_lock,
3030 .clear_acl_cache = nfs4_zap_acl_attr, 3155 .clear_acl_cache = nfs4_zap_acl_attr,
3031}; 3156};
diff --git a/fs/nfs/nfs4state.c b/fs/nfs/nfs4state.c
index afe587d82f..23834c8fb7 100644
--- a/fs/nfs/nfs4state.c
+++ b/fs/nfs/nfs4state.c
@@ -264,13 +264,15 @@ nfs4_alloc_state_owner(void)
264{ 264{
265 struct nfs4_state_owner *sp; 265 struct nfs4_state_owner *sp;
266 266
267 sp = kmalloc(sizeof(*sp),GFP_KERNEL); 267 sp = kzalloc(sizeof(*sp),GFP_KERNEL);
268 if (!sp) 268 if (!sp)
269 return NULL; 269 return NULL;
270 init_MUTEX(&sp->so_sema);
271 sp->so_seqid = 0; /* arbitrary */
272 INIT_LIST_HEAD(&sp->so_states); 270 INIT_LIST_HEAD(&sp->so_states);
273 INIT_LIST_HEAD(&sp->so_delegations); 271 INIT_LIST_HEAD(&sp->so_delegations);
272 rpc_init_wait_queue(&sp->so_sequence.wait, "Seqid_waitqueue");
273 sp->so_seqid.sequence = &sp->so_sequence;
274 spin_lock_init(&sp->so_sequence.lock);
275 INIT_LIST_HEAD(&sp->so_sequence.list);
274 atomic_set(&sp->so_count, 1); 276 atomic_set(&sp->so_count, 1);
275 return sp; 277 return sp;
276} 278}
@@ -359,7 +361,6 @@ nfs4_alloc_open_state(void)
359 memset(state->stateid.data, 0, sizeof(state->stateid.data)); 361 memset(state->stateid.data, 0, sizeof(state->stateid.data));
360 atomic_set(&state->count, 1); 362 atomic_set(&state->count, 1);
361 INIT_LIST_HEAD(&state->lock_states); 363 INIT_LIST_HEAD(&state->lock_states);
362 init_MUTEX(&state->lock_sema);
363 spin_lock_init(&state->state_lock); 364 spin_lock_init(&state->state_lock);
364 return state; 365 return state;
365} 366}
@@ -441,7 +442,6 @@ nfs4_get_open_state(struct inode *inode, struct nfs4_state_owner *owner)
441 state = __nfs4_find_state_byowner(inode, owner); 442 state = __nfs4_find_state_byowner(inode, owner);
442 if (state == NULL && new != NULL) { 443 if (state == NULL && new != NULL) {
443 state = new; 444 state = new;
444 /* Caller *must* be holding owner->so_sem */
445 /* Note: The reclaim code dictates that we add stateless 445 /* Note: The reclaim code dictates that we add stateless
446 * and read-only stateids to the end of the list */ 446 * and read-only stateids to the end of the list */
447 list_add_tail(&state->open_states, &owner->so_states); 447 list_add_tail(&state->open_states, &owner->so_states);
@@ -461,7 +461,7 @@ out:
461 461
462/* 462/*
463 * Beware! Caller must be holding exactly one 463 * Beware! Caller must be holding exactly one
464 * reference to clp->cl_sem and owner->so_sema! 464 * reference to clp->cl_sem!
465 */ 465 */
466void nfs4_put_open_state(struct nfs4_state *state) 466void nfs4_put_open_state(struct nfs4_state *state)
467{ 467{
@@ -481,19 +481,15 @@ void nfs4_put_open_state(struct nfs4_state *state)
481} 481}
482 482
483/* 483/*
484 * Beware! Caller must be holding no references to clp->cl_sem! 484 * Close the current file.
485 * of owner->so_sema!
486 */ 485 */
487void nfs4_close_state(struct nfs4_state *state, mode_t mode) 486void nfs4_close_state(struct nfs4_state *state, mode_t mode)
488{ 487{
489 struct inode *inode = state->inode; 488 struct inode *inode = state->inode;
490 struct nfs4_state_owner *owner = state->owner; 489 struct nfs4_state_owner *owner = state->owner;
491 struct nfs4_client *clp = owner->so_client;
492 int newstate; 490 int newstate;
493 491
494 atomic_inc(&owner->so_count); 492 atomic_inc(&owner->so_count);
495 down_read(&clp->cl_sem);
496 down(&owner->so_sema);
497 /* Protect against nfs4_find_state() */ 493 /* Protect against nfs4_find_state() */
498 spin_lock(&inode->i_lock); 494 spin_lock(&inode->i_lock);
499 if (mode & FMODE_READ) 495 if (mode & FMODE_READ)
@@ -515,14 +511,16 @@ void nfs4_close_state(struct nfs4_state *state, mode_t mode)
515 newstate |= FMODE_WRITE; 511 newstate |= FMODE_WRITE;
516 if (state->state == newstate) 512 if (state->state == newstate)
517 goto out; 513 goto out;
518 if (nfs4_do_close(inode, state, newstate) == -EINPROGRESS) 514 if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
515 state->state = newstate;
516 goto out;
517 }
518 if (nfs4_do_close(inode, state, newstate) == 0)
519 return; 519 return;
520 } 520 }
521out: 521out:
522 nfs4_put_open_state(state); 522 nfs4_put_open_state(state);
523 up(&owner->so_sema);
524 nfs4_put_state_owner(owner); 523 nfs4_put_state_owner(owner);
525 up_read(&clp->cl_sem);
526} 524}
527 525
528/* 526/*
@@ -546,19 +544,16 @@ __nfs4_find_lock_state(struct nfs4_state *state, fl_owner_t fl_owner)
546 * Return a compatible lock_state. If no initialized lock_state structure 544 * Return a compatible lock_state. If no initialized lock_state structure
547 * exists, return an uninitialized one. 545 * exists, return an uninitialized one.
548 * 546 *
549 * The caller must be holding state->lock_sema
550 */ 547 */
551static struct nfs4_lock_state *nfs4_alloc_lock_state(struct nfs4_state *state, fl_owner_t fl_owner) 548static struct nfs4_lock_state *nfs4_alloc_lock_state(struct nfs4_state *state, fl_owner_t fl_owner)
552{ 549{
553 struct nfs4_lock_state *lsp; 550 struct nfs4_lock_state *lsp;
554 struct nfs4_client *clp = state->owner->so_client; 551 struct nfs4_client *clp = state->owner->so_client;
555 552
556 lsp = kmalloc(sizeof(*lsp), GFP_KERNEL); 553 lsp = kzalloc(sizeof(*lsp), GFP_KERNEL);
557 if (lsp == NULL) 554 if (lsp == NULL)
558 return NULL; 555 return NULL;
559 lsp->ls_flags = 0; 556 lsp->ls_seqid.sequence = &state->owner->so_sequence;
560 lsp->ls_seqid = 0; /* arbitrary */
561 memset(lsp->ls_stateid.data, 0, sizeof(lsp->ls_stateid.data));
562 atomic_set(&lsp->ls_count, 1); 557 atomic_set(&lsp->ls_count, 1);
563 lsp->ls_owner = fl_owner; 558 lsp->ls_owner = fl_owner;
564 spin_lock(&clp->cl_lock); 559 spin_lock(&clp->cl_lock);
@@ -572,7 +567,7 @@ static struct nfs4_lock_state *nfs4_alloc_lock_state(struct nfs4_state *state, f
572 * Return a compatible lock_state. If no initialized lock_state structure 567 * Return a compatible lock_state. If no initialized lock_state structure
573 * exists, return an uninitialized one. 568 * exists, return an uninitialized one.
574 * 569 *
575 * The caller must be holding state->lock_sema and clp->cl_sem 570 * The caller must be holding clp->cl_sem
576 */ 571 */
577static struct nfs4_lock_state *nfs4_get_lock_state(struct nfs4_state *state, fl_owner_t owner) 572static struct nfs4_lock_state *nfs4_get_lock_state(struct nfs4_state *state, fl_owner_t owner)
578{ 573{
@@ -605,7 +600,7 @@ static struct nfs4_lock_state *nfs4_get_lock_state(struct nfs4_state *state, fl_
605 * Release reference to lock_state, and free it if we see that 600 * Release reference to lock_state, and free it if we see that
606 * it is no longer in use 601 * it is no longer in use
607 */ 602 */
608static void nfs4_put_lock_state(struct nfs4_lock_state *lsp) 603void nfs4_put_lock_state(struct nfs4_lock_state *lsp)
609{ 604{
610 struct nfs4_state *state; 605 struct nfs4_state *state;
611 606
@@ -673,29 +668,98 @@ void nfs4_copy_stateid(nfs4_stateid *dst, struct nfs4_state *state, fl_owner_t f
673 nfs4_put_lock_state(lsp); 668 nfs4_put_lock_state(lsp);
674} 669}
675 670
676/* 671struct nfs_seqid *nfs_alloc_seqid(struct nfs_seqid_counter *counter)
677* Called with state->lock_sema and clp->cl_sem held. 672{
678*/ 673 struct rpc_sequence *sequence = counter->sequence;
679void nfs4_increment_lock_seqid(int status, struct nfs4_lock_state *lsp) 674 struct nfs_seqid *new;
675
676 new = kmalloc(sizeof(*new), GFP_KERNEL);
677 if (new != NULL) {
678 new->sequence = counter;
679 new->task = NULL;
680 spin_lock(&sequence->lock);
681 list_add_tail(&new->list, &sequence->list);
682 spin_unlock(&sequence->lock);
683 }
684 return new;
685}
686
687void nfs_free_seqid(struct nfs_seqid *seqid)
680{ 688{
681 if (status == NFS_OK || seqid_mutating_err(-status)) 689 struct rpc_sequence *sequence = seqid->sequence->sequence;
682 lsp->ls_seqid++; 690 struct rpc_task *next = NULL;
691
692 spin_lock(&sequence->lock);
693 list_del(&seqid->list);
694 if (!list_empty(&sequence->list)) {
695 next = list_entry(sequence->list.next, struct nfs_seqid, list)->task;
696 if (next)
697 rpc_wake_up_task(next);
698 }
699 spin_unlock(&sequence->lock);
700 kfree(seqid);
683} 701}
684 702
685/* 703/*
686* Called with sp->so_sema and clp->cl_sem held. 704 * Increment the seqid if the OPEN/OPEN_DOWNGRADE/CLOSE succeeded, or
687* 705 * failed with a seqid incrementing error -
688* Increment the seqid if the OPEN/OPEN_DOWNGRADE/CLOSE succeeded, or 706 * see comments nfs_fs.h:seqid_mutating_error()
689* failed with a seqid incrementing error - 707 */
690* see comments nfs_fs.h:seqid_mutating_error() 708static inline void nfs_increment_seqid(int status, struct nfs_seqid *seqid)
691*/ 709{
692void nfs4_increment_seqid(int status, struct nfs4_state_owner *sp) 710 switch (status) {
693{ 711 case 0:
694 if (status == NFS_OK || seqid_mutating_err(-status)) 712 break;
695 sp->so_seqid++; 713 case -NFS4ERR_BAD_SEQID:
696 /* If the server returns BAD_SEQID, unhash state_owner here */ 714 case -NFS4ERR_STALE_CLIENTID:
697 if (status == -NFS4ERR_BAD_SEQID) 715 case -NFS4ERR_STALE_STATEID:
716 case -NFS4ERR_BAD_STATEID:
717 case -NFS4ERR_BADXDR:
718 case -NFS4ERR_RESOURCE:
719 case -NFS4ERR_NOFILEHANDLE:
720 /* Non-seqid mutating errors */
721 return;
722 };
723 /*
724 * Note: no locking needed as we are guaranteed to be first
725 * on the sequence list
726 */
727 seqid->sequence->counter++;
728}
729
730void nfs_increment_open_seqid(int status, struct nfs_seqid *seqid)
731{
732 if (status == -NFS4ERR_BAD_SEQID) {
733 struct nfs4_state_owner *sp = container_of(seqid->sequence,
734 struct nfs4_state_owner, so_seqid);
698 nfs4_drop_state_owner(sp); 735 nfs4_drop_state_owner(sp);
736 }
737 return nfs_increment_seqid(status, seqid);
738}
739
740/*
741 * Increment the seqid if the LOCK/LOCKU succeeded, or
742 * failed with a seqid incrementing error -
743 * see comments nfs_fs.h:seqid_mutating_error()
744 */
745void nfs_increment_lock_seqid(int status, struct nfs_seqid *seqid)
746{
747 return nfs_increment_seqid(status, seqid);
748}
749
750int nfs_wait_on_sequence(struct nfs_seqid *seqid, struct rpc_task *task)
751{
752 struct rpc_sequence *sequence = seqid->sequence->sequence;
753 int status = 0;
754
755 spin_lock(&sequence->lock);
756 if (sequence->list.next != &seqid->list) {
757 seqid->task = task;
758 rpc_sleep_on(&sequence->wait, task, NULL, NULL);
759 status = -EAGAIN;
760 }
761 spin_unlock(&sequence->lock);
762 return status;
699} 763}
700 764
701static int reclaimer(void *); 765static int reclaimer(void *);
@@ -791,8 +855,6 @@ static int nfs4_reclaim_open_state(struct nfs4_state_recovery_ops *ops, struct n
791 if (state->state == 0) 855 if (state->state == 0)
792 continue; 856 continue;
793 status = ops->recover_open(sp, state); 857 status = ops->recover_open(sp, state);
794 list_for_each_entry(lock, &state->lock_states, ls_locks)
795 lock->ls_flags &= ~NFS_LOCK_INITIALIZED;
796 if (status >= 0) { 858 if (status >= 0) {
797 status = nfs4_reclaim_locks(ops, state); 859 status = nfs4_reclaim_locks(ops, state);
798 if (status < 0) 860 if (status < 0)
@@ -831,6 +893,26 @@ out_err:
831 return status; 893 return status;
832} 894}
833 895
896static void nfs4_state_mark_reclaim(struct nfs4_client *clp)
897{
898 struct nfs4_state_owner *sp;
899 struct nfs4_state *state;
900 struct nfs4_lock_state *lock;
901
902 /* Reset all sequence ids to zero */
903 list_for_each_entry(sp, &clp->cl_state_owners, so_list) {
904 sp->so_seqid.counter = 0;
905 sp->so_seqid.flags = 0;
906 list_for_each_entry(state, &sp->so_states, open_states) {
907 list_for_each_entry(lock, &state->lock_states, ls_locks) {
908 lock->ls_seqid.counter = 0;
909 lock->ls_seqid.flags = 0;
910 lock->ls_flags &= ~NFS_LOCK_INITIALIZED;
911 }
912 }
913 }
914}
915
834static int reclaimer(void *ptr) 916static int reclaimer(void *ptr)
835{ 917{
836 struct reclaimer_args *args = (struct reclaimer_args *)ptr; 918 struct reclaimer_args *args = (struct reclaimer_args *)ptr;
@@ -864,6 +946,7 @@ restart_loop:
864 default: 946 default:
865 ops = &nfs4_network_partition_recovery_ops; 947 ops = &nfs4_network_partition_recovery_ops;
866 }; 948 };
949 nfs4_state_mark_reclaim(clp);
867 status = __nfs4_init_client(clp); 950 status = __nfs4_init_client(clp);
868 if (status) 951 if (status)
869 goto out_error; 952 goto out_error;
diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c
index 6c564ef948..c5c75235c5 100644
--- a/fs/nfs/nfs4xdr.c
+++ b/fs/nfs/nfs4xdr.c
@@ -602,10 +602,10 @@ static int encode_close(struct xdr_stream *xdr, const struct nfs_closeargs *arg)
602{ 602{
603 uint32_t *p; 603 uint32_t *p;
604 604
605 RESERVE_SPACE(8+sizeof(arg->stateid.data)); 605 RESERVE_SPACE(8+sizeof(arg->stateid->data));
606 WRITE32(OP_CLOSE); 606 WRITE32(OP_CLOSE);
607 WRITE32(arg->seqid); 607 WRITE32(arg->seqid->sequence->counter);
608 WRITEMEM(arg->stateid.data, sizeof(arg->stateid.data)); 608 WRITEMEM(arg->stateid->data, sizeof(arg->stateid->data));
609 609
610 return 0; 610 return 0;
611} 611}
@@ -729,22 +729,18 @@ static int encode_lock(struct xdr_stream *xdr, const struct nfs_lockargs *arg)
729 WRITE64(arg->length); 729 WRITE64(arg->length);
730 WRITE32(opargs->new_lock_owner); 730 WRITE32(opargs->new_lock_owner);
731 if (opargs->new_lock_owner){ 731 if (opargs->new_lock_owner){
732 struct nfs_open_to_lock *ol = opargs->u.open_lock;
733
734 RESERVE_SPACE(40); 732 RESERVE_SPACE(40);
735 WRITE32(ol->open_seqid); 733 WRITE32(opargs->open_seqid->sequence->counter);
736 WRITEMEM(&ol->open_stateid, sizeof(ol->open_stateid)); 734 WRITEMEM(opargs->open_stateid->data, sizeof(opargs->open_stateid->data));
737 WRITE32(ol->lock_seqid); 735 WRITE32(opargs->lock_seqid->sequence->counter);
738 WRITE64(ol->lock_owner.clientid); 736 WRITE64(opargs->lock_owner.clientid);
739 WRITE32(4); 737 WRITE32(4);
740 WRITE32(ol->lock_owner.id); 738 WRITE32(opargs->lock_owner.id);
741 } 739 }
742 else { 740 else {
743 struct nfs_exist_lock *el = opargs->u.exist_lock;
744
745 RESERVE_SPACE(20); 741 RESERVE_SPACE(20);
746 WRITEMEM(&el->stateid, sizeof(el->stateid)); 742 WRITEMEM(opargs->lock_stateid->data, sizeof(opargs->lock_stateid->data));
747 WRITE32(el->seqid); 743 WRITE32(opargs->lock_seqid->sequence->counter);
748 } 744 }
749 745
750 return 0; 746 return 0;
@@ -775,8 +771,8 @@ static int encode_locku(struct xdr_stream *xdr, const struct nfs_lockargs *arg)
775 RESERVE_SPACE(44); 771 RESERVE_SPACE(44);
776 WRITE32(OP_LOCKU); 772 WRITE32(OP_LOCKU);
777 WRITE32(arg->type); 773 WRITE32(arg->type);
778 WRITE32(opargs->seqid); 774 WRITE32(opargs->seqid->sequence->counter);
779 WRITEMEM(&opargs->stateid, sizeof(opargs->stateid)); 775 WRITEMEM(opargs->stateid->data, sizeof(opargs->stateid->data));
780 WRITE64(arg->offset); 776 WRITE64(arg->offset);
781 WRITE64(arg->length); 777 WRITE64(arg->length);
782 778
@@ -826,7 +822,7 @@ static inline void encode_openhdr(struct xdr_stream *xdr, const struct nfs_opena
826 */ 822 */
827 RESERVE_SPACE(8); 823 RESERVE_SPACE(8);
828 WRITE32(OP_OPEN); 824 WRITE32(OP_OPEN);
829 WRITE32(arg->seqid); 825 WRITE32(arg->seqid->sequence->counter);
830 encode_share_access(xdr, arg->open_flags); 826 encode_share_access(xdr, arg->open_flags);
831 RESERVE_SPACE(16); 827 RESERVE_SPACE(16);
832 WRITE64(arg->clientid); 828 WRITE64(arg->clientid);
@@ -941,7 +937,7 @@ static int encode_open_confirm(struct xdr_stream *xdr, const struct nfs_open_con
941 RESERVE_SPACE(8+sizeof(arg->stateid.data)); 937 RESERVE_SPACE(8+sizeof(arg->stateid.data));
942 WRITE32(OP_OPEN_CONFIRM); 938 WRITE32(OP_OPEN_CONFIRM);
943 WRITEMEM(arg->stateid.data, sizeof(arg->stateid.data)); 939 WRITEMEM(arg->stateid.data, sizeof(arg->stateid.data));
944 WRITE32(arg->seqid); 940 WRITE32(arg->seqid->sequence->counter);
945 941
946 return 0; 942 return 0;
947} 943}
@@ -950,10 +946,10 @@ static int encode_open_downgrade(struct xdr_stream *xdr, const struct nfs_closea
950{ 946{
951 uint32_t *p; 947 uint32_t *p;
952 948
953 RESERVE_SPACE(8+sizeof(arg->stateid.data)); 949 RESERVE_SPACE(8+sizeof(arg->stateid->data));
954 WRITE32(OP_OPEN_DOWNGRADE); 950 WRITE32(OP_OPEN_DOWNGRADE);
955 WRITEMEM(arg->stateid.data, sizeof(arg->stateid.data)); 951 WRITEMEM(arg->stateid->data, sizeof(arg->stateid->data));
956 WRITE32(arg->seqid); 952 WRITE32(arg->seqid->sequence->counter);
957 encode_share_access(xdr, arg->open_flags); 953 encode_share_access(xdr, arg->open_flags);
958 return 0; 954 return 0;
959} 955}
@@ -1437,6 +1433,9 @@ static int nfs4_xdr_enc_open(struct rpc_rqst *req, uint32_t *p, struct nfs_opena
1437 }; 1433 };
1438 int status; 1434 int status;
1439 1435
1436 status = nfs_wait_on_sequence(args->seqid, req->rq_task);
1437 if (status != 0)
1438 goto out;
1440 xdr_init_encode(&xdr, &req->rq_snd_buf, p); 1439 xdr_init_encode(&xdr, &req->rq_snd_buf, p);
1441 encode_compound_hdr(&xdr, &hdr); 1440 encode_compound_hdr(&xdr, &hdr);
1442 status = encode_putfh(&xdr, args->fh); 1441 status = encode_putfh(&xdr, args->fh);
@@ -1464,6 +1463,9 @@ static int nfs4_xdr_enc_open_confirm(struct rpc_rqst *req, uint32_t *p, struct n
1464 }; 1463 };
1465 int status; 1464 int status;
1466 1465
1466 status = nfs_wait_on_sequence(args->seqid, req->rq_task);
1467 if (status != 0)
1468 goto out;
1467 xdr_init_encode(&xdr, &req->rq_snd_buf, p); 1469 xdr_init_encode(&xdr, &req->rq_snd_buf, p);
1468 encode_compound_hdr(&xdr, &hdr); 1470 encode_compound_hdr(&xdr, &hdr);
1469 status = encode_putfh(&xdr, args->fh); 1471 status = encode_putfh(&xdr, args->fh);
@@ -1485,6 +1487,9 @@ static int nfs4_xdr_enc_open_noattr(struct rpc_rqst *req, uint32_t *p, struct nf
1485 }; 1487 };
1486 int status; 1488 int status;
1487 1489
1490 status = nfs_wait_on_sequence(args->seqid, req->rq_task);
1491 if (status != 0)
1492 goto out;
1488 xdr_init_encode(&xdr, &req->rq_snd_buf, p); 1493 xdr_init_encode(&xdr, &req->rq_snd_buf, p);
1489 encode_compound_hdr(&xdr, &hdr); 1494 encode_compound_hdr(&xdr, &hdr);
1490 status = encode_putfh(&xdr, args->fh); 1495 status = encode_putfh(&xdr, args->fh);
@@ -1525,8 +1530,15 @@ static int nfs4_xdr_enc_lock(struct rpc_rqst *req, uint32_t *p, struct nfs_locka
1525 struct compound_hdr hdr = { 1530 struct compound_hdr hdr = {
1526 .nops = 2, 1531 .nops = 2,
1527 }; 1532 };
1533 struct nfs_lock_opargs *opargs = args->u.lock;
1528 int status; 1534 int status;
1529 1535
1536 status = nfs_wait_on_sequence(opargs->lock_seqid, req->rq_task);
1537 if (status != 0)
1538 goto out;
1539 /* Do we need to do an open_to_lock_owner? */
1540 if (opargs->lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)
1541 opargs->new_lock_owner = 0;
1530 xdr_init_encode(&xdr, &req->rq_snd_buf, p); 1542 xdr_init_encode(&xdr, &req->rq_snd_buf, p);
1531 encode_compound_hdr(&xdr, &hdr); 1543 encode_compound_hdr(&xdr, &hdr);
1532 status = encode_putfh(&xdr, args->fh); 1544 status = encode_putfh(&xdr, args->fh);
@@ -2890,8 +2902,8 @@ static int decode_lock(struct xdr_stream *xdr, struct nfs_lockres *res)
2890 2902
2891 status = decode_op_hdr(xdr, OP_LOCK); 2903 status = decode_op_hdr(xdr, OP_LOCK);
2892 if (status == 0) { 2904 if (status == 0) {
2893 READ_BUF(sizeof(nfs4_stateid)); 2905 READ_BUF(sizeof(res->u.stateid.data));
2894 COPYMEM(&res->u.stateid, sizeof(res->u.stateid)); 2906 COPYMEM(res->u.stateid.data, sizeof(res->u.stateid.data));
2895 } else if (status == -NFS4ERR_DENIED) 2907 } else if (status == -NFS4ERR_DENIED)
2896 return decode_lock_denied(xdr, &res->u.denied); 2908 return decode_lock_denied(xdr, &res->u.denied);
2897 return status; 2909 return status;
@@ -2913,8 +2925,8 @@ static int decode_locku(struct xdr_stream *xdr, struct nfs_lockres *res)
2913 2925
2914 status = decode_op_hdr(xdr, OP_LOCKU); 2926 status = decode_op_hdr(xdr, OP_LOCKU);
2915 if (status == 0) { 2927 if (status == 0) {
2916 READ_BUF(sizeof(nfs4_stateid)); 2928 READ_BUF(sizeof(res->u.stateid.data));
2917 COPYMEM(&res->u.stateid, sizeof(res->u.stateid)); 2929 COPYMEM(res->u.stateid.data, sizeof(res->u.stateid.data));
2918 } 2930 }
2919 return status; 2931 return status;
2920} 2932}
diff --git a/fs/nfs/proc.c b/fs/nfs/proc.c
index be23c3fb92..8fef86523d 100644
--- a/fs/nfs/proc.c
+++ b/fs/nfs/proc.c
@@ -216,7 +216,7 @@ static int nfs_proc_write(struct nfs_write_data *wdata)
216 216
217static int 217static int
218nfs_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr, 218nfs_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
219 int flags) 219 int flags, struct nameidata *nd)
220{ 220{
221 struct nfs_fh fhandle; 221 struct nfs_fh fhandle;
222 struct nfs_fattr fattr; 222 struct nfs_fattr fattr;
diff --git a/fs/open.c b/fs/open.c
index f0d90cf049..8d06ec911f 100644
--- a/fs/open.c
+++ b/fs/open.c
@@ -739,7 +739,8 @@ asmlinkage long sys_fchown(unsigned int fd, uid_t user, gid_t group)
739} 739}
740 740
741static struct file *__dentry_open(struct dentry *dentry, struct vfsmount *mnt, 741static struct file *__dentry_open(struct dentry *dentry, struct vfsmount *mnt,
742 int flags, struct file *f) 742 int flags, struct file *f,
743 int (*open)(struct inode *, struct file *))
743{ 744{
744 struct inode *inode; 745 struct inode *inode;
745 int error; 746 int error;
@@ -761,11 +762,14 @@ static struct file *__dentry_open(struct dentry *dentry, struct vfsmount *mnt,
761 f->f_op = fops_get(inode->i_fop); 762 f->f_op = fops_get(inode->i_fop);
762 file_move(f, &inode->i_sb->s_files); 763 file_move(f, &inode->i_sb->s_files);
763 764
764 if (f->f_op && f->f_op->open) { 765 if (!open && f->f_op)
765 error = f->f_op->open(inode,f); 766 open = f->f_op->open;
767 if (open) {
768 error = open(inode, f);
766 if (error) 769 if (error)
767 goto cleanup_all; 770 goto cleanup_all;
768 } 771 }
772
769 f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC); 773 f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
770 774
771 file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping); 775 file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
@@ -814,28 +818,75 @@ struct file *filp_open(const char * filename, int flags, int mode)
814{ 818{
815 int namei_flags, error; 819 int namei_flags, error;
816 struct nameidata nd; 820 struct nameidata nd;
817 struct file *f;
818 821
819 namei_flags = flags; 822 namei_flags = flags;
820 if ((namei_flags+1) & O_ACCMODE) 823 if ((namei_flags+1) & O_ACCMODE)
821 namei_flags++; 824 namei_flags++;
822 if (namei_flags & O_TRUNC)
823 namei_flags |= 2;
824
825 error = -ENFILE;
826 f = get_empty_filp();
827 if (f == NULL)
828 return ERR_PTR(error);
829 825
830 error = open_namei(filename, namei_flags, mode, &nd); 826 error = open_namei(filename, namei_flags, mode, &nd);
831 if (!error) 827 if (!error)
832 return __dentry_open(nd.dentry, nd.mnt, flags, f); 828 return nameidata_to_filp(&nd, flags);
833 829
834 put_filp(f);
835 return ERR_PTR(error); 830 return ERR_PTR(error);
836} 831}
837EXPORT_SYMBOL(filp_open); 832EXPORT_SYMBOL(filp_open);
838 833
834/**
835 * lookup_instantiate_filp - instantiates the open intent filp
836 * @nd: pointer to nameidata
837 * @dentry: pointer to dentry
838 * @open: open callback
839 *
840 * Helper for filesystems that want to use lookup open intents and pass back
841 * a fully instantiated struct file to the caller.
842 * This function is meant to be called from within a filesystem's
843 * lookup method.
844 * Note that in case of error, nd->intent.open.file is destroyed, but the
845 * path information remains valid.
846 * If the open callback is set to NULL, then the standard f_op->open()
847 * filesystem callback is substituted.
848 */
849struct file *lookup_instantiate_filp(struct nameidata *nd, struct dentry *dentry,
850 int (*open)(struct inode *, struct file *))
851{
852 if (IS_ERR(nd->intent.open.file))
853 goto out;
854 if (IS_ERR(dentry))
855 goto out_err;
856 nd->intent.open.file = __dentry_open(dget(dentry), mntget(nd->mnt),
857 nd->intent.open.flags - 1,
858 nd->intent.open.file,
859 open);
860out:
861 return nd->intent.open.file;
862out_err:
863 release_open_intent(nd);
864 nd->intent.open.file = (struct file *)dentry;
865 goto out;
866}
867EXPORT_SYMBOL_GPL(lookup_instantiate_filp);
868
869/**
870 * nameidata_to_filp - convert a nameidata to an open filp.
871 * @nd: pointer to nameidata
872 * @flags: open flags
873 *
874 * Note that this function destroys the original nameidata
875 */
876struct file *nameidata_to_filp(struct nameidata *nd, int flags)
877{
878 struct file *filp;
879
880 /* Pick up the filp from the open intent */
881 filp = nd->intent.open.file;
882 /* Has the filesystem initialised the file for us? */
883 if (filp->f_dentry == NULL)
884 filp = __dentry_open(nd->dentry, nd->mnt, flags, filp, NULL);
885 else
886 path_release(nd);
887 return filp;
888}
889
839struct file *dentry_open(struct dentry *dentry, struct vfsmount *mnt, int flags) 890struct file *dentry_open(struct dentry *dentry, struct vfsmount *mnt, int flags)
840{ 891{
841 int error; 892 int error;
@@ -846,7 +897,7 @@ struct file *dentry_open(struct dentry *dentry, struct vfsmount *mnt, int flags)
846 if (f == NULL) 897 if (f == NULL)
847 return ERR_PTR(error); 898 return ERR_PTR(error);
848 899
849 return __dentry_open(dentry, mnt, flags, f); 900 return __dentry_open(dentry, mnt, flags, f, NULL);
850} 901}
851EXPORT_SYMBOL(dentry_open); 902EXPORT_SYMBOL(dentry_open);
852 903
diff --git a/fs/proc/base.c b/fs/proc/base.c
index 3b33f94020..a170450aad 100644
--- a/fs/proc/base.c
+++ b/fs/proc/base.c
@@ -103,7 +103,9 @@ enum pid_directory_inos {
103 PROC_TGID_NUMA_MAPS, 103 PROC_TGID_NUMA_MAPS,
104 PROC_TGID_MOUNTS, 104 PROC_TGID_MOUNTS,
105 PROC_TGID_WCHAN, 105 PROC_TGID_WCHAN,
106#ifdef CONFIG_MMU
106 PROC_TGID_SMAPS, 107 PROC_TGID_SMAPS,
108#endif
107#ifdef CONFIG_SCHEDSTATS 109#ifdef CONFIG_SCHEDSTATS
108 PROC_TGID_SCHEDSTAT, 110 PROC_TGID_SCHEDSTAT,
109#endif 111#endif
@@ -141,7 +143,9 @@ enum pid_directory_inos {
141 PROC_TID_NUMA_MAPS, 143 PROC_TID_NUMA_MAPS,
142 PROC_TID_MOUNTS, 144 PROC_TID_MOUNTS,
143 PROC_TID_WCHAN, 145 PROC_TID_WCHAN,
146#ifdef CONFIG_MMU
144 PROC_TID_SMAPS, 147 PROC_TID_SMAPS,
148#endif
145#ifdef CONFIG_SCHEDSTATS 149#ifdef CONFIG_SCHEDSTATS
146 PROC_TID_SCHEDSTAT, 150 PROC_TID_SCHEDSTAT,
147#endif 151#endif
@@ -195,7 +199,9 @@ static struct pid_entry tgid_base_stuff[] = {
195 E(PROC_TGID_ROOT, "root", S_IFLNK|S_IRWXUGO), 199 E(PROC_TGID_ROOT, "root", S_IFLNK|S_IRWXUGO),
196 E(PROC_TGID_EXE, "exe", S_IFLNK|S_IRWXUGO), 200 E(PROC_TGID_EXE, "exe", S_IFLNK|S_IRWXUGO),
197 E(PROC_TGID_MOUNTS, "mounts", S_IFREG|S_IRUGO), 201 E(PROC_TGID_MOUNTS, "mounts", S_IFREG|S_IRUGO),
202#ifdef CONFIG_MMU
198 E(PROC_TGID_SMAPS, "smaps", S_IFREG|S_IRUGO), 203 E(PROC_TGID_SMAPS, "smaps", S_IFREG|S_IRUGO),
204#endif
199#ifdef CONFIG_SECURITY 205#ifdef CONFIG_SECURITY
200 E(PROC_TGID_ATTR, "attr", S_IFDIR|S_IRUGO|S_IXUGO), 206 E(PROC_TGID_ATTR, "attr", S_IFDIR|S_IRUGO|S_IXUGO),
201#endif 207#endif
@@ -235,7 +241,9 @@ static struct pid_entry tid_base_stuff[] = {
235 E(PROC_TID_ROOT, "root", S_IFLNK|S_IRWXUGO), 241 E(PROC_TID_ROOT, "root", S_IFLNK|S_IRWXUGO),
236 E(PROC_TID_EXE, "exe", S_IFLNK|S_IRWXUGO), 242 E(PROC_TID_EXE, "exe", S_IFLNK|S_IRWXUGO),
237 E(PROC_TID_MOUNTS, "mounts", S_IFREG|S_IRUGO), 243 E(PROC_TID_MOUNTS, "mounts", S_IFREG|S_IRUGO),
244#ifdef CONFIG_MMU
238 E(PROC_TID_SMAPS, "smaps", S_IFREG|S_IRUGO), 245 E(PROC_TID_SMAPS, "smaps", S_IFREG|S_IRUGO),
246#endif
239#ifdef CONFIG_SECURITY 247#ifdef CONFIG_SECURITY
240 E(PROC_TID_ATTR, "attr", S_IFDIR|S_IRUGO|S_IXUGO), 248 E(PROC_TID_ATTR, "attr", S_IFDIR|S_IRUGO|S_IXUGO),
241#endif 249#endif
@@ -630,6 +638,7 @@ static struct file_operations proc_numa_maps_operations = {
630}; 638};
631#endif 639#endif
632 640
641#ifdef CONFIG_MMU
633extern struct seq_operations proc_pid_smaps_op; 642extern struct seq_operations proc_pid_smaps_op;
634static int smaps_open(struct inode *inode, struct file *file) 643static int smaps_open(struct inode *inode, struct file *file)
635{ 644{
@@ -648,6 +657,7 @@ static struct file_operations proc_smaps_operations = {
648 .llseek = seq_lseek, 657 .llseek = seq_lseek,
649 .release = seq_release, 658 .release = seq_release,
650}; 659};
660#endif
651 661
652extern struct seq_operations mounts_op; 662extern struct seq_operations mounts_op;
653static int mounts_open(struct inode *inode, struct file *file) 663static int mounts_open(struct inode *inode, struct file *file)
@@ -1681,10 +1691,12 @@ static struct dentry *proc_pident_lookup(struct inode *dir,
1681 case PROC_TGID_MOUNTS: 1691 case PROC_TGID_MOUNTS:
1682 inode->i_fop = &proc_mounts_operations; 1692 inode->i_fop = &proc_mounts_operations;
1683 break; 1693 break;
1694#ifdef CONFIG_MMU
1684 case PROC_TID_SMAPS: 1695 case PROC_TID_SMAPS:
1685 case PROC_TGID_SMAPS: 1696 case PROC_TGID_SMAPS:
1686 inode->i_fop = &proc_smaps_operations; 1697 inode->i_fop = &proc_smaps_operations;
1687 break; 1698 break;
1699#endif
1688#ifdef CONFIG_SECURITY 1700#ifdef CONFIG_SECURITY
1689 case PROC_TID_ATTR: 1701 case PROC_TID_ATTR:
1690 inode->i_nlink = 2; 1702 inode->i_nlink = 2;
diff --git a/fs/proc/nommu.c b/fs/proc/nommu.c
index f3bf016d5e..cff10ab1af 100644
--- a/fs/proc/nommu.c
+++ b/fs/proc/nommu.c
@@ -91,6 +91,7 @@ static void *nommu_vma_list_start(struct seq_file *m, loff_t *_pos)
91 next = _rb; 91 next = _rb;
92 break; 92 break;
93 } 93 }
94 pos--;
94 } 95 }
95 96
96 return next; 97 return next;
generated by cgit v1.2.2 (git 2.25.0) at 2025-03-29 07:19:21 -0400