aboutsummaryrefslogtreecommitdiffstats
path: root/fs
diff options
context:
space:
mode:
authorDavid S. Miller <davem@davemloft.net>2008-09-08 19:59:05 -0400
committerDavid S. Miller <davem@davemloft.net>2008-09-08 19:59:05 -0400
commit17dce5dfe38ae2fb359b61e855f5d8a3a8b7892b (patch)
tree88bb1fcf84f9ebfa4299c9a8dcd9e6330b358446 /fs
parent712d6954e3998d0de2840d8130941e8042541246 (diff)
parent82a28c794f27aac17d7a3ebd7f14d731a11a5532 (diff)
Merge branch 'master' of master.kernel.org:/pub/scm/linux/kernel/git/torvalds/linux-2.6
Conflicts: net/mac80211/mlme.c
Diffstat (limited to 'fs')
-rw-r--r--fs/9p/vfs_dir.c1
-rw-r--r--fs/Kconfig21
-rw-r--r--fs/adfs/dir.c1
-rw-r--r--fs/affs/dir.c1
-rw-r--r--fs/autofs4/root.c2
-rw-r--r--fs/befs/linuxvfs.c1
-rw-r--r--fs/bio.c48
-rw-r--r--fs/buffer.c13
-rw-r--r--fs/cifs/CHANGES10
-rw-r--r--fs/cifs/README44
-rw-r--r--fs/cifs/asn1.c11
-rw-r--r--fs/cifs/cifs_spnego.c4
-rw-r--r--fs/cifs/cifs_spnego.h2
-rw-r--r--fs/cifs/cifsencrypt.c1
-rw-r--r--fs/cifs/cifsglob.h3
-rw-r--r--fs/cifs/connect.c33
-rw-r--r--fs/cifs/dns_resolve.c7
-rw-r--r--fs/cifs/file.c4
-rw-r--r--fs/cifs/inode.c3
-rw-r--r--fs/cifs/sess.c13
-rw-r--r--fs/compat.c8
-rw-r--r--fs/configfs/dir.c17
-rw-r--r--fs/dcache.c2
-rw-r--r--fs/efs/namei.c3
-rw-r--r--fs/ext4/balloc.c3
-rw-r--r--fs/ext4/dir.c20
-rw-r--r--fs/ext4/ext4.h4
-rw-r--r--fs/ext4/ext4_extents.h4
-rw-r--r--fs/ext4/ext4_jbd2.h8
-rw-r--r--fs/ext4/extents.c113
-rw-r--r--fs/ext4/ialloc.c2
-rw-r--r--fs/ext4/inode.c478
-rw-r--r--fs/ext4/mballoc.c53
-rw-r--r--fs/ext4/migrate.c3
-rw-r--r--fs/ext4/resize.c3
-rw-r--r--fs/ext4/super.c1
-rw-r--r--fs/jffs2/jffs2_fs_i.h1
-rw-r--r--fs/nfs/super.c6
-rw-r--r--fs/nfsd/nfs4acl.c2
-rw-r--r--fs/nfsd/nfs4proc.c12
-rw-r--r--fs/ntfs/namei.c89
-rw-r--r--fs/ntfs/usnjrnl.h4
-rw-r--r--fs/ocfs2/cluster/netdebug.c26
-rw-r--r--fs/ocfs2/cluster/tcp.c44
-rw-r--r--fs/ocfs2/cluster/tcp_internal.h32
-rw-r--r--fs/ocfs2/dir.c11
-rw-r--r--fs/ocfs2/journal.c23
-rw-r--r--fs/ocfs2/stackglue.c7
-rw-r--r--fs/proc/array.c59
-rw-r--r--fs/proc/generic.c1
-rw-r--r--fs/proc/proc_misc.c7
-rw-r--r--fs/readdir.c8
-rw-r--r--fs/seq_file.c11
-rw-r--r--fs/xfs/linux-2.6/xfs_file.c1
-rw-r--r--fs/xfs/linux-2.6/xfs_iops.c2
-rw-r--r--fs/xfs/xfs_dmapi.h1
56 files changed, 746 insertions, 546 deletions
diff --git a/fs/9p/vfs_dir.c b/fs/9p/vfs_dir.c
index 88e3787c6ea9..e298fe194093 100644
--- a/fs/9p/vfs_dir.c
+++ b/fs/9p/vfs_dir.c
@@ -119,6 +119,7 @@ int v9fs_dir_release(struct inode *inode, struct file *filp)
119 119
120const struct file_operations v9fs_dir_operations = { 120const struct file_operations v9fs_dir_operations = {
121 .read = generic_read_dir, 121 .read = generic_read_dir,
122 .llseek = generic_file_llseek,
122 .readdir = v9fs_dir_readdir, 123 .readdir = v9fs_dir_readdir,
123 .open = v9fs_file_open, 124 .open = v9fs_file_open,
124 .release = v9fs_dir_release, 125 .release = v9fs_dir_release,
diff --git a/fs/Kconfig b/fs/Kconfig
index d3873583360b..abccb5dab9a8 100644
--- a/fs/Kconfig
+++ b/fs/Kconfig
@@ -1930,6 +1930,16 @@ config CIFS_WEAK_PW_HASH
1930 1930
1931 If unsure, say N. 1931 If unsure, say N.
1932 1932
1933config CIFS_UPCALL
1934 bool "Kerberos/SPNEGO advanced session setup"
1935 depends on CIFS && KEYS
1936 help
1937 Enables an upcall mechanism for CIFS which accesses
1938 userspace helper utilities to provide SPNEGO packaged (RFC 4178)
1939 Kerberos tickets which are needed to mount to certain secure servers
1940 (for which more secure Kerberos authentication is required). If
1941 unsure, say N.
1942
1933config CIFS_XATTR 1943config CIFS_XATTR
1934 bool "CIFS extended attributes" 1944 bool "CIFS extended attributes"
1935 depends on CIFS 1945 depends on CIFS
@@ -1982,17 +1992,6 @@ config CIFS_EXPERIMENTAL
1982 (which is disabled by default). See the file fs/cifs/README 1992 (which is disabled by default). See the file fs/cifs/README
1983 for more details. If unsure, say N. 1993 for more details. If unsure, say N.
1984 1994
1985config CIFS_UPCALL
1986 bool "Kerberos/SPNEGO advanced session setup (EXPERIMENTAL)"
1987 depends on CIFS_EXPERIMENTAL
1988 depends on KEYS
1989 help
1990 Enables an upcall mechanism for CIFS which accesses
1991 userspace helper utilities to provide SPNEGO packaged (RFC 4178)
1992 Kerberos tickets which are needed to mount to certain secure servers
1993 (for which more secure Kerberos authentication is required). If
1994 unsure, say N.
1995
1996config CIFS_DFS_UPCALL 1995config CIFS_DFS_UPCALL
1997 bool "DFS feature support (EXPERIMENTAL)" 1996 bool "DFS feature support (EXPERIMENTAL)"
1998 depends on CIFS_EXPERIMENTAL 1997 depends on CIFS_EXPERIMENTAL
diff --git a/fs/adfs/dir.c b/fs/adfs/dir.c
index fc1a8dc64d78..85a30e929800 100644
--- a/fs/adfs/dir.c
+++ b/fs/adfs/dir.c
@@ -197,6 +197,7 @@ out:
197 197
198const struct file_operations adfs_dir_operations = { 198const struct file_operations adfs_dir_operations = {
199 .read = generic_read_dir, 199 .read = generic_read_dir,
200 .llseek = generic_file_llseek,
200 .readdir = adfs_readdir, 201 .readdir = adfs_readdir,
201 .fsync = file_fsync, 202 .fsync = file_fsync,
202}; 203};
diff --git a/fs/affs/dir.c b/fs/affs/dir.c
index 6e3f282424b0..7b36904dbeac 100644
--- a/fs/affs/dir.c
+++ b/fs/affs/dir.c
@@ -19,6 +19,7 @@ static int affs_readdir(struct file *, void *, filldir_t);
19 19
20const struct file_operations affs_dir_operations = { 20const struct file_operations affs_dir_operations = {
21 .read = generic_read_dir, 21 .read = generic_read_dir,
22 .llseek = generic_file_llseek,
22 .readdir = affs_readdir, 23 .readdir = affs_readdir,
23 .fsync = file_fsync, 24 .fsync = file_fsync,
24}; 25};
diff --git a/fs/autofs4/root.c b/fs/autofs4/root.c
index bcfb2dc0a61b..2a41c2a7fc52 100644
--- a/fs/autofs4/root.c
+++ b/fs/autofs4/root.c
@@ -36,6 +36,7 @@ const struct file_operations autofs4_root_operations = {
36 .release = dcache_dir_close, 36 .release = dcache_dir_close,
37 .read = generic_read_dir, 37 .read = generic_read_dir,
38 .readdir = dcache_readdir, 38 .readdir = dcache_readdir,
39 .llseek = dcache_dir_lseek,
39 .ioctl = autofs4_root_ioctl, 40 .ioctl = autofs4_root_ioctl,
40}; 41};
41 42
@@ -44,6 +45,7 @@ const struct file_operations autofs4_dir_operations = {
44 .release = dcache_dir_close, 45 .release = dcache_dir_close,
45 .read = generic_read_dir, 46 .read = generic_read_dir,
46 .readdir = dcache_readdir, 47 .readdir = dcache_readdir,
48 .llseek = dcache_dir_lseek,
47}; 49};
48 50
49const struct inode_operations autofs4_indirect_root_inode_operations = { 51const struct inode_operations autofs4_indirect_root_inode_operations = {
diff --git a/fs/befs/linuxvfs.c b/fs/befs/linuxvfs.c
index 02c6e62b72f8..740f53672a8a 100644
--- a/fs/befs/linuxvfs.c
+++ b/fs/befs/linuxvfs.c
@@ -66,6 +66,7 @@ static struct kmem_cache *befs_inode_cachep;
66static const struct file_operations befs_dir_operations = { 66static const struct file_operations befs_dir_operations = {
67 .read = generic_read_dir, 67 .read = generic_read_dir,
68 .readdir = befs_readdir, 68 .readdir = befs_readdir,
69 .llseek = generic_file_llseek,
69}; 70};
70 71
71static const struct inode_operations befs_dir_inode_operations = { 72static const struct inode_operations befs_dir_inode_operations = {
diff --git a/fs/bio.c b/fs/bio.c
index 8000e2fa16cb..3cba7ae34d75 100644
--- a/fs/bio.c
+++ b/fs/bio.c
@@ -469,20 +469,21 @@ static void bio_free_map_data(struct bio_map_data *bmd)
469 kfree(bmd); 469 kfree(bmd);
470} 470}
471 471
472static struct bio_map_data *bio_alloc_map_data(int nr_segs, int iov_count) 472static struct bio_map_data *bio_alloc_map_data(int nr_segs, int iov_count,
473 gfp_t gfp_mask)
473{ 474{
474 struct bio_map_data *bmd = kmalloc(sizeof(*bmd), GFP_KERNEL); 475 struct bio_map_data *bmd = kmalloc(sizeof(*bmd), gfp_mask);
475 476
476 if (!bmd) 477 if (!bmd)
477 return NULL; 478 return NULL;
478 479
479 bmd->iovecs = kmalloc(sizeof(struct bio_vec) * nr_segs, GFP_KERNEL); 480 bmd->iovecs = kmalloc(sizeof(struct bio_vec) * nr_segs, gfp_mask);
480 if (!bmd->iovecs) { 481 if (!bmd->iovecs) {
481 kfree(bmd); 482 kfree(bmd);
482 return NULL; 483 return NULL;
483 } 484 }
484 485
485 bmd->sgvecs = kmalloc(sizeof(struct sg_iovec) * iov_count, GFP_KERNEL); 486 bmd->sgvecs = kmalloc(sizeof(struct sg_iovec) * iov_count, gfp_mask);
486 if (bmd->sgvecs) 487 if (bmd->sgvecs)
487 return bmd; 488 return bmd;
488 489
@@ -491,8 +492,8 @@ static struct bio_map_data *bio_alloc_map_data(int nr_segs, int iov_count)
491 return NULL; 492 return NULL;
492} 493}
493 494
494static int __bio_copy_iov(struct bio *bio, struct sg_iovec *iov, int iov_count, 495static int __bio_copy_iov(struct bio *bio, struct bio_vec *iovecs,
495 int uncopy) 496 struct sg_iovec *iov, int iov_count, int uncopy)
496{ 497{
497 int ret = 0, i; 498 int ret = 0, i;
498 struct bio_vec *bvec; 499 struct bio_vec *bvec;
@@ -502,7 +503,7 @@ static int __bio_copy_iov(struct bio *bio, struct sg_iovec *iov, int iov_count,
502 503
503 __bio_for_each_segment(bvec, bio, i, 0) { 504 __bio_for_each_segment(bvec, bio, i, 0) {
504 char *bv_addr = page_address(bvec->bv_page); 505 char *bv_addr = page_address(bvec->bv_page);
505 unsigned int bv_len = bvec->bv_len; 506 unsigned int bv_len = iovecs[i].bv_len;
506 507
507 while (bv_len && iov_idx < iov_count) { 508 while (bv_len && iov_idx < iov_count) {
508 unsigned int bytes; 509 unsigned int bytes;
@@ -554,7 +555,7 @@ int bio_uncopy_user(struct bio *bio)
554 struct bio_map_data *bmd = bio->bi_private; 555 struct bio_map_data *bmd = bio->bi_private;
555 int ret; 556 int ret;
556 557
557 ret = __bio_copy_iov(bio, bmd->sgvecs, bmd->nr_sgvecs, 1); 558 ret = __bio_copy_iov(bio, bmd->iovecs, bmd->sgvecs, bmd->nr_sgvecs, 1);
558 559
559 bio_free_map_data(bmd); 560 bio_free_map_data(bmd);
560 bio_put(bio); 561 bio_put(bio);
@@ -596,7 +597,7 @@ struct bio *bio_copy_user_iov(struct request_queue *q, struct sg_iovec *iov,
596 len += iov[i].iov_len; 597 len += iov[i].iov_len;
597 } 598 }
598 599
599 bmd = bio_alloc_map_data(nr_pages, iov_count); 600 bmd = bio_alloc_map_data(nr_pages, iov_count, GFP_KERNEL);
600 if (!bmd) 601 if (!bmd)
601 return ERR_PTR(-ENOMEM); 602 return ERR_PTR(-ENOMEM);
602 603
@@ -633,7 +634,7 @@ struct bio *bio_copy_user_iov(struct request_queue *q, struct sg_iovec *iov,
633 * success 634 * success
634 */ 635 */
635 if (!write_to_vm) { 636 if (!write_to_vm) {
636 ret = __bio_copy_iov(bio, iov, iov_count, 0); 637 ret = __bio_copy_iov(bio, bio->bi_io_vec, iov, iov_count, 0);
637 if (ret) 638 if (ret)
638 goto cleanup; 639 goto cleanup;
639 } 640 }
@@ -942,19 +943,22 @@ static void bio_copy_kern_endio(struct bio *bio, int err)
942{ 943{
943 struct bio_vec *bvec; 944 struct bio_vec *bvec;
944 const int read = bio_data_dir(bio) == READ; 945 const int read = bio_data_dir(bio) == READ;
945 char *p = bio->bi_private; 946 struct bio_map_data *bmd = bio->bi_private;
946 int i; 947 int i;
948 char *p = bmd->sgvecs[0].iov_base;
947 949
948 __bio_for_each_segment(bvec, bio, i, 0) { 950 __bio_for_each_segment(bvec, bio, i, 0) {
949 char *addr = page_address(bvec->bv_page); 951 char *addr = page_address(bvec->bv_page);
952 int len = bmd->iovecs[i].bv_len;
950 953
951 if (read && !err) 954 if (read && !err)
952 memcpy(p, addr, bvec->bv_len); 955 memcpy(p, addr, len);
953 956
954 __free_page(bvec->bv_page); 957 __free_page(bvec->bv_page);
955 p += bvec->bv_len; 958 p += len;
956 } 959 }
957 960
961 bio_free_map_data(bmd);
958 bio_put(bio); 962 bio_put(bio);
959} 963}
960 964
@@ -978,11 +982,21 @@ struct bio *bio_copy_kern(struct request_queue *q, void *data, unsigned int len,
978 const int nr_pages = end - start; 982 const int nr_pages = end - start;
979 struct bio *bio; 983 struct bio *bio;
980 struct bio_vec *bvec; 984 struct bio_vec *bvec;
985 struct bio_map_data *bmd;
981 int i, ret; 986 int i, ret;
987 struct sg_iovec iov;
988
989 iov.iov_base = data;
990 iov.iov_len = len;
991
992 bmd = bio_alloc_map_data(nr_pages, 1, gfp_mask);
993 if (!bmd)
994 return ERR_PTR(-ENOMEM);
982 995
996 ret = -ENOMEM;
983 bio = bio_alloc(gfp_mask, nr_pages); 997 bio = bio_alloc(gfp_mask, nr_pages);
984 if (!bio) 998 if (!bio)
985 return ERR_PTR(-ENOMEM); 999 goto out_bmd;
986 1000
987 while (len) { 1001 while (len) {
988 struct page *page; 1002 struct page *page;
@@ -1016,14 +1030,18 @@ struct bio *bio_copy_kern(struct request_queue *q, void *data, unsigned int len,
1016 } 1030 }
1017 } 1031 }
1018 1032
1019 bio->bi_private = data; 1033 bio->bi_private = bmd;
1020 bio->bi_end_io = bio_copy_kern_endio; 1034 bio->bi_end_io = bio_copy_kern_endio;
1035
1036 bio_set_map_data(bmd, bio, &iov, 1);
1021 return bio; 1037 return bio;
1022cleanup: 1038cleanup:
1023 bio_for_each_segment(bvec, bio, i) 1039 bio_for_each_segment(bvec, bio, i)
1024 __free_page(bvec->bv_page); 1040 __free_page(bvec->bv_page);
1025 1041
1026 bio_put(bio); 1042 bio_put(bio);
1043out_bmd:
1044 bio_free_map_data(bmd);
1027 1045
1028 return ERR_PTR(ret); 1046 return ERR_PTR(ret);
1029} 1047}
diff --git a/fs/buffer.c b/fs/buffer.c
index 38653e36e225..ac78d4c19b3b 100644
--- a/fs/buffer.c
+++ b/fs/buffer.c
@@ -2926,14 +2926,17 @@ int submit_bh(int rw, struct buffer_head * bh)
2926 BUG_ON(!buffer_mapped(bh)); 2926 BUG_ON(!buffer_mapped(bh));
2927 BUG_ON(!bh->b_end_io); 2927 BUG_ON(!bh->b_end_io);
2928 2928
2929 if (buffer_ordered(bh) && (rw == WRITE)) 2929 /*
2930 rw = WRITE_BARRIER; 2930 * Mask in barrier bit for a write (could be either a WRITE or a
2931 * WRITE_SYNC
2932 */
2933 if (buffer_ordered(bh) && (rw & WRITE))
2934 rw |= WRITE_BARRIER;
2931 2935
2932 /* 2936 /*
2933 * Only clear out a write error when rewriting, should this 2937 * Only clear out a write error when rewriting
2934 * include WRITE_SYNC as well?
2935 */ 2938 */
2936 if (test_set_buffer_req(bh) && (rw == WRITE || rw == WRITE_BARRIER)) 2939 if (test_set_buffer_req(bh) && (rw & WRITE))
2937 clear_buffer_write_io_error(bh); 2940 clear_buffer_write_io_error(bh);
2938 2941
2939 /* 2942 /*
diff --git a/fs/cifs/CHANGES b/fs/cifs/CHANGES
index f5d0083e09fa..06e521a945c3 100644
--- a/fs/cifs/CHANGES
+++ b/fs/cifs/CHANGES
@@ -4,7 +4,15 @@ Fix premature write failure on congested networks (we would give up
4on EAGAIN from the socket too quickly on large writes). 4on EAGAIN from the socket too quickly on large writes).
5Cifs_mkdir and cifs_create now respect the setgid bit on parent dir. 5Cifs_mkdir and cifs_create now respect the setgid bit on parent dir.
6Fix endian problems in acl (mode from/to cifs acl) on bigendian 6Fix endian problems in acl (mode from/to cifs acl) on bigendian
7architectures. 7architectures. Fix problems with preserving timestamps on copying open
8files (e.g. "cp -a") to Windows servers. For mkdir and create honor setgid bit
9on parent directory when server supports Unix Extensions but not POSIX
10create. Update cifs.upcall version to handle new Kerberos sec flags
11(this requires update of cifs.upcall program from Samba). Fix memory leak
12on dns_upcall (resolving DFS referralls). Fix plain text password
13authentication (requires setting SecurityFlags to 0x30030 to enable
14lanman and plain text though). Fix writes to be at correct offset when
15file is open with O_APPEND and file is on a directio (forcediretio) mount.
8 16
9Version 1.53 17Version 1.53
10------------ 18------------
diff --git a/fs/cifs/README b/fs/cifs/README
index 2bd6fe556f88..bd2343d4c6a6 100644
--- a/fs/cifs/README
+++ b/fs/cifs/README
@@ -542,10 +542,20 @@ SecurityFlags Flags which control security negotiation and
542 hashing mechanisms (as "must use") on the other hand 542 hashing mechanisms (as "must use") on the other hand
543 does not make much sense. Default flags are 543 does not make much sense. Default flags are
544 0x07007 544 0x07007
545 (NTLM, NTLMv2 and packet signing allowed). Maximum 545 (NTLM, NTLMv2 and packet signing allowed). The maximum
546 allowable flags if you want to allow mounts to servers 546 allowable flags if you want to allow mounts to servers
547 using weaker password hashes is 0x37037 (lanman, 547 using weaker password hashes is 0x37037 (lanman,
548 plaintext, ntlm, ntlmv2, signing allowed): 548 plaintext, ntlm, ntlmv2, signing allowed). Some
549 SecurityFlags require the corresponding menuconfig
550 options to be enabled (lanman and plaintext require
551 CONFIG_CIFS_WEAK_PW_HASH for example). Enabling
552 plaintext authentication currently requires also
553 enabling lanman authentication in the security flags
554 because the cifs module only supports sending
555 laintext passwords using the older lanman dialect
556 form of the session setup SMB. (e.g. for authentication
557 using plain text passwords, set the SecurityFlags
558 to 0x30030):
549 559
550 may use packet signing 0x00001 560 may use packet signing 0x00001
551 must use packet signing 0x01001 561 must use packet signing 0x01001
@@ -642,8 +652,30 @@ The statistics for the number of total SMBs and oplock breaks are different in
642that they represent all for that share, not just those for which the server 652that they represent all for that share, not just those for which the server
643returned success. 653returned success.
644 654
645Also note that "cat /proc/fs/cifs/DebugData" will display information about 655Also note that "cat /proc/fs/cifs/DebugData" will display information about
646the active sessions and the shares that are mounted. 656the active sessions and the shares that are mounted.
647Enabling Kerberos (extended security) works when CONFIG_CIFS_EXPERIMENTAL is 657
648on but requires a user space helper (from the Samba project). NTLM and NTLMv2 and 658Enabling Kerberos (extended security) works but requires version 1.2 or later
649LANMAN support do not require this helper. 659of the helper program cifs.upcall to be present and to be configured in the
660/etc/request-key.conf file. The cifs.upcall helper program is from the Samba
661project(http://www.samba.org). NTLM and NTLMv2 and LANMAN support do not
662require this helper. Note that NTLMv2 security (which does not require the
663cifs.upcall helper program), instead of using Kerberos, is sufficient for
664some use cases.
665
666Enabling DFS support (used to access shares transparently in an MS-DFS
667global name space) requires that CONFIG_CIFS_EXPERIMENTAL be enabled. In
668addition, DFS support for target shares which are specified as UNC
669names which begin with host names (rather than IP addresses) requires
670a user space helper (such as cifs.upcall) to be present in order to
671translate host names to ip address, and the user space helper must also
672be configured in the file /etc/request-key.conf
673
674To use cifs Kerberos and DFS support, the Linux keyutils package should be
675installed and something like the following lines should be added to the
676/etc/request-key.conf file:
677
678create cifs.spnego * * /usr/local/sbin/cifs.upcall %k
679create dns_resolver * * /usr/local/sbin/cifs.upcall %k
680
681
diff --git a/fs/cifs/asn1.c b/fs/cifs/asn1.c
index 5fabd2caf93c..1b09f1670061 100644
--- a/fs/cifs/asn1.c
+++ b/fs/cifs/asn1.c
@@ -476,6 +476,7 @@ decode_negTokenInit(unsigned char *security_blob, int length,
476 unsigned int cls, con, tag, oidlen, rc; 476 unsigned int cls, con, tag, oidlen, rc;
477 bool use_ntlmssp = false; 477 bool use_ntlmssp = false;
478 bool use_kerberos = false; 478 bool use_kerberos = false;
479 bool use_mskerberos = false;
479 480
480 *secType = NTLM; /* BB eventually make Kerberos or NLTMSSP the default*/ 481 *secType = NTLM; /* BB eventually make Kerberos or NLTMSSP the default*/
481 482
@@ -574,10 +575,12 @@ decode_negTokenInit(unsigned char *security_blob, int length,
574 *(oid + 1), *(oid + 2), *(oid + 3))); 575 *(oid + 1), *(oid + 2), *(oid + 3)));
575 576
576 if (compare_oid(oid, oidlen, MSKRB5_OID, 577 if (compare_oid(oid, oidlen, MSKRB5_OID,
577 MSKRB5_OID_LEN)) 578 MSKRB5_OID_LEN) &&
578 use_kerberos = true; 579 !use_kerberos)
580 use_mskerberos = true;
579 else if (compare_oid(oid, oidlen, KRB5_OID, 581 else if (compare_oid(oid, oidlen, KRB5_OID,
580 KRB5_OID_LEN)) 582 KRB5_OID_LEN) &&
583 !use_mskerberos)
581 use_kerberos = true; 584 use_kerberos = true;
582 else if (compare_oid(oid, oidlen, NTLMSSP_OID, 585 else if (compare_oid(oid, oidlen, NTLMSSP_OID,
583 NTLMSSP_OID_LEN)) 586 NTLMSSP_OID_LEN))
@@ -630,6 +633,8 @@ decode_negTokenInit(unsigned char *security_blob, int length,
630 633
631 if (use_kerberos) 634 if (use_kerberos)
632 *secType = Kerberos; 635 *secType = Kerberos;
636 else if (use_mskerberos)
637 *secType = MSKerberos;
633 else if (use_ntlmssp) 638 else if (use_ntlmssp)
634 *secType = NTLMSSP; 639 *secType = NTLMSSP;
635 640
diff --git a/fs/cifs/cifs_spnego.c b/fs/cifs/cifs_spnego.c
index 2434ab0e8791..117ef4bba68e 100644
--- a/fs/cifs/cifs_spnego.c
+++ b/fs/cifs/cifs_spnego.c
@@ -114,9 +114,11 @@ cifs_get_spnego_key(struct cifsSesInfo *sesInfo)
114 114
115 dp = description + strlen(description); 115 dp = description + strlen(description);
116 116
117 /* for now, only sec=krb5 is valid */ 117 /* for now, only sec=krb5 and sec=mskrb5 are valid */
118 if (server->secType == Kerberos) 118 if (server->secType == Kerberos)
119 sprintf(dp, ";sec=krb5"); 119 sprintf(dp, ";sec=krb5");
120 else if (server->secType == MSKerberos)
121 sprintf(dp, ";sec=mskrb5");
120 else 122 else
121 goto out; 123 goto out;
122 124
diff --git a/fs/cifs/cifs_spnego.h b/fs/cifs/cifs_spnego.h
index 05a34b17a1ab..e4041ec4d712 100644
--- a/fs/cifs/cifs_spnego.h
+++ b/fs/cifs/cifs_spnego.h
@@ -23,7 +23,7 @@
23#ifndef _CIFS_SPNEGO_H 23#ifndef _CIFS_SPNEGO_H
24#define _CIFS_SPNEGO_H 24#define _CIFS_SPNEGO_H
25 25
26#define CIFS_SPNEGO_UPCALL_VERSION 1 26#define CIFS_SPNEGO_UPCALL_VERSION 2
27 27
28/* 28/*
29 * The version field should always be set to CIFS_SPNEGO_UPCALL_VERSION. 29 * The version field should always be set to CIFS_SPNEGO_UPCALL_VERSION.
diff --git a/fs/cifs/cifsencrypt.c b/fs/cifs/cifsencrypt.c
index 83fd40dc1ef0..bd5f13d38450 100644
--- a/fs/cifs/cifsencrypt.c
+++ b/fs/cifs/cifsencrypt.c
@@ -294,6 +294,7 @@ void calc_lanman_hash(struct cifsSesInfo *ses, char *lnm_session_key)
294 294
295 if ((ses->server->secMode & SECMODE_PW_ENCRYPT) == 0) 295 if ((ses->server->secMode & SECMODE_PW_ENCRYPT) == 0)
296 if (extended_security & CIFSSEC_MAY_PLNTXT) { 296 if (extended_security & CIFSSEC_MAY_PLNTXT) {
297 memset(lnm_session_key, 0, CIFS_SESS_KEY_SIZE);
297 memcpy(lnm_session_key, password_with_pad, 298 memcpy(lnm_session_key, password_with_pad,
298 CIFS_ENCPWD_SIZE); 299 CIFS_ENCPWD_SIZE);
299 return; 300 return;
diff --git a/fs/cifs/cifsglob.h b/fs/cifs/cifsglob.h
index 7e1cf262effe..8dfd6f24d488 100644
--- a/fs/cifs/cifsglob.h
+++ b/fs/cifs/cifsglob.h
@@ -80,7 +80,8 @@ enum securityEnum {
80 NTLMv2, /* Legacy NTLM auth with NTLMv2 hash */ 80 NTLMv2, /* Legacy NTLM auth with NTLMv2 hash */
81 RawNTLMSSP, /* NTLMSSP without SPNEGO */ 81 RawNTLMSSP, /* NTLMSSP without SPNEGO */
82 NTLMSSP, /* NTLMSSP via SPNEGO */ 82 NTLMSSP, /* NTLMSSP via SPNEGO */
83 Kerberos /* Kerberos via SPNEGO */ 83 Kerberos, /* Kerberos via SPNEGO */
84 MSKerberos, /* MS Kerberos via SPNEGO */
84}; 85};
85 86
86enum protocolEnum { 87enum protocolEnum {
diff --git a/fs/cifs/connect.c b/fs/cifs/connect.c
index 0711db65afe8..4c13bcdb92a5 100644
--- a/fs/cifs/connect.c
+++ b/fs/cifs/connect.c
@@ -3598,19 +3598,21 @@ int cifs_setup_session(unsigned int xid, struct cifsSesInfo *pSesInfo,
3598 char ntlm_session_key[CIFS_SESS_KEY_SIZE]; 3598 char ntlm_session_key[CIFS_SESS_KEY_SIZE];
3599 bool ntlmv2_flag = false; 3599 bool ntlmv2_flag = false;
3600 int first_time = 0; 3600 int first_time = 0;
3601 struct TCP_Server_Info *server = pSesInfo->server;
3601 3602
3602 /* what if server changes its buffer size after dropping the session? */ 3603 /* what if server changes its buffer size after dropping the session? */
3603 if (pSesInfo->server->maxBuf == 0) /* no need to send on reconnect */ { 3604 if (server->maxBuf == 0) /* no need to send on reconnect */ {
3604 rc = CIFSSMBNegotiate(xid, pSesInfo); 3605 rc = CIFSSMBNegotiate(xid, pSesInfo);
3605 if (rc == -EAGAIN) /* retry only once on 1st time connection */ { 3606 if (rc == -EAGAIN) {
3607 /* retry only once on 1st time connection */
3606 rc = CIFSSMBNegotiate(xid, pSesInfo); 3608 rc = CIFSSMBNegotiate(xid, pSesInfo);
3607 if (rc == -EAGAIN) 3609 if (rc == -EAGAIN)
3608 rc = -EHOSTDOWN; 3610 rc = -EHOSTDOWN;
3609 } 3611 }
3610 if (rc == 0) { 3612 if (rc == 0) {
3611 spin_lock(&GlobalMid_Lock); 3613 spin_lock(&GlobalMid_Lock);
3612 if (pSesInfo->server->tcpStatus != CifsExiting) 3614 if (server->tcpStatus != CifsExiting)
3613 pSesInfo->server->tcpStatus = CifsGood; 3615 server->tcpStatus = CifsGood;
3614 else 3616 else
3615 rc = -EHOSTDOWN; 3617 rc = -EHOSTDOWN;
3616 spin_unlock(&GlobalMid_Lock); 3618 spin_unlock(&GlobalMid_Lock);
@@ -3623,23 +3625,22 @@ int cifs_setup_session(unsigned int xid, struct cifsSesInfo *pSesInfo,
3623 goto ss_err_exit; 3625 goto ss_err_exit;
3624 3626
3625 pSesInfo->flags = 0; 3627 pSesInfo->flags = 0;
3626 pSesInfo->capabilities = pSesInfo->server->capabilities; 3628 pSesInfo->capabilities = server->capabilities;
3627 if (linuxExtEnabled == 0) 3629 if (linuxExtEnabled == 0)
3628 pSesInfo->capabilities &= (~CAP_UNIX); 3630 pSesInfo->capabilities &= (~CAP_UNIX);
3629 /* pSesInfo->sequence_number = 0;*/ 3631 /* pSesInfo->sequence_number = 0;*/
3630 cFYI(1, ("Security Mode: 0x%x Capabilities: 0x%x TimeAdjust: %d", 3632 cFYI(1, ("Security Mode: 0x%x Capabilities: 0x%x TimeAdjust: %d",
3631 pSesInfo->server->secMode, 3633 server->secMode, server->capabilities, server->timeAdj));
3632 pSesInfo->server->capabilities, 3634
3633 pSesInfo->server->timeAdj));
3634 if (experimEnabled < 2) 3635 if (experimEnabled < 2)
3635 rc = CIFS_SessSetup(xid, pSesInfo, first_time, nls_info); 3636 rc = CIFS_SessSetup(xid, pSesInfo, first_time, nls_info);
3636 else if (extended_security 3637 else if (extended_security
3637 && (pSesInfo->capabilities & CAP_EXTENDED_SECURITY) 3638 && (pSesInfo->capabilities & CAP_EXTENDED_SECURITY)
3638 && (pSesInfo->server->secType == NTLMSSP)) { 3639 && (server->secType == NTLMSSP)) {
3639 rc = -EOPNOTSUPP; 3640 rc = -EOPNOTSUPP;
3640 } else if (extended_security 3641 } else if (extended_security
3641 && (pSesInfo->capabilities & CAP_EXTENDED_SECURITY) 3642 && (pSesInfo->capabilities & CAP_EXTENDED_SECURITY)
3642 && (pSesInfo->server->secType == RawNTLMSSP)) { 3643 && (server->secType == RawNTLMSSP)) {
3643 cFYI(1, ("NTLMSSP sesssetup")); 3644 cFYI(1, ("NTLMSSP sesssetup"));
3644 rc = CIFSNTLMSSPNegotiateSessSetup(xid, pSesInfo, &ntlmv2_flag, 3645 rc = CIFSNTLMSSPNegotiateSessSetup(xid, pSesInfo, &ntlmv2_flag,
3645 nls_info); 3646 nls_info);
@@ -3668,12 +3669,12 @@ int cifs_setup_session(unsigned int xid, struct cifsSesInfo *pSesInfo,
3668 3669
3669 } else { 3670 } else {
3670 SMBNTencrypt(pSesInfo->password, 3671 SMBNTencrypt(pSesInfo->password,
3671 pSesInfo->server->cryptKey, 3672 server->cryptKey,
3672 ntlm_session_key); 3673 ntlm_session_key);
3673 3674
3674 if (first_time) 3675 if (first_time)
3675 cifs_calculate_mac_key( 3676 cifs_calculate_mac_key(
3676 &pSesInfo->server->mac_signing_key, 3677 &server->mac_signing_key,
3677 ntlm_session_key, 3678 ntlm_session_key,
3678 pSesInfo->password); 3679 pSesInfo->password);
3679 } 3680 }
@@ -3686,13 +3687,13 @@ int cifs_setup_session(unsigned int xid, struct cifsSesInfo *pSesInfo,
3686 nls_info); 3687 nls_info);
3687 } 3688 }
3688 } else { /* old style NTLM 0.12 session setup */ 3689 } else { /* old style NTLM 0.12 session setup */
3689 SMBNTencrypt(pSesInfo->password, pSesInfo->server->cryptKey, 3690 SMBNTencrypt(pSesInfo->password, server->cryptKey,
3690 ntlm_session_key); 3691 ntlm_session_key);
3691 3692
3692 if (first_time) 3693 if (first_time)
3693 cifs_calculate_mac_key( 3694 cifs_calculate_mac_key(&server->mac_signing_key,
3694 &pSesInfo->server->mac_signing_key, 3695 ntlm_session_key,
3695 ntlm_session_key, pSesInfo->password); 3696 pSesInfo->password);
3696 3697
3697 rc = CIFSSessSetup(xid, pSesInfo, ntlm_session_key, nls_info); 3698 rc = CIFSSessSetup(xid, pSesInfo, ntlm_session_key, nls_info);
3698 } 3699 }
diff --git a/fs/cifs/dns_resolve.c b/fs/cifs/dns_resolve.c
index f730ef35499e..a2e0673e1b08 100644
--- a/fs/cifs/dns_resolve.c
+++ b/fs/cifs/dns_resolve.c
@@ -47,11 +47,18 @@ static int dns_resolver_instantiate(struct key *key, const void *data,
47 return rc; 47 return rc;
48} 48}
49 49
50static void
51dns_resolver_destroy(struct key *key)
52{
53 kfree(key->payload.data);
54}
55
50struct key_type key_type_dns_resolver = { 56struct key_type key_type_dns_resolver = {
51 .name = "dns_resolver", 57 .name = "dns_resolver",
52 .def_datalen = sizeof(struct in_addr), 58 .def_datalen = sizeof(struct in_addr),
53 .describe = user_describe, 59 .describe = user_describe,
54 .instantiate = dns_resolver_instantiate, 60 .instantiate = dns_resolver_instantiate,
61 .destroy = dns_resolver_destroy,
55 .match = user_match, 62 .match = user_match,
56}; 63};
57 64
diff --git a/fs/cifs/file.c b/fs/cifs/file.c
index ff14d14903a0..cbefe1f1f9fe 100644
--- a/fs/cifs/file.c
+++ b/fs/cifs/file.c
@@ -833,6 +833,10 @@ ssize_t cifs_user_write(struct file *file, const char __user *write_data,
833 return -EBADF; 833 return -EBADF;
834 open_file = (struct cifsFileInfo *) file->private_data; 834 open_file = (struct cifsFileInfo *) file->private_data;
835 835
836 rc = generic_write_checks(file, poffset, &write_size, 0);
837 if (rc)
838 return rc;
839
836 xid = GetXid(); 840 xid = GetXid();
837 841
838 if (*poffset > file->f_path.dentry->d_inode->i_size) 842 if (*poffset > file->f_path.dentry->d_inode->i_size)
diff --git a/fs/cifs/inode.c b/fs/cifs/inode.c
index 848286861c31..9c548f110102 100644
--- a/fs/cifs/inode.c
+++ b/fs/cifs/inode.c
@@ -546,7 +546,8 @@ int cifs_get_inode_info(struct inode **pinode,
546 if ((inode->i_mode & S_IWUGO) == 0 && 546 if ((inode->i_mode & S_IWUGO) == 0 &&
547 (attr & ATTR_READONLY) == 0) 547 (attr & ATTR_READONLY) == 0)
548 inode->i_mode |= (S_IWUGO & default_mode); 548 inode->i_mode |= (S_IWUGO & default_mode);
549 inode->i_mode &= ~S_IFMT; 549
550 inode->i_mode &= ~S_IFMT;
550 } 551 }
551 /* clear write bits if ATTR_READONLY is set */ 552 /* clear write bits if ATTR_READONLY is set */
552 if (attr & ATTR_READONLY) 553 if (attr & ATTR_READONLY)
diff --git a/fs/cifs/sess.c b/fs/cifs/sess.c
index ed150efbe27c..252fdc0567f1 100644
--- a/fs/cifs/sess.c
+++ b/fs/cifs/sess.c
@@ -409,6 +409,8 @@ CIFS_SessSetup(unsigned int xid, struct cifsSesInfo *ses, int first_time,
409#ifdef CONFIG_CIFS_WEAK_PW_HASH 409#ifdef CONFIG_CIFS_WEAK_PW_HASH
410 char lnm_session_key[CIFS_SESS_KEY_SIZE]; 410 char lnm_session_key[CIFS_SESS_KEY_SIZE];
411 411
412 pSMB->req.hdr.Flags2 &= ~SMBFLG2_UNICODE;
413
412 /* no capabilities flags in old lanman negotiation */ 414 /* no capabilities flags in old lanman negotiation */
413 415
414 pSMB->old_req.PasswordLength = cpu_to_le16(CIFS_SESS_KEY_SIZE); 416 pSMB->old_req.PasswordLength = cpu_to_le16(CIFS_SESS_KEY_SIZE);
@@ -505,7 +507,7 @@ CIFS_SessSetup(unsigned int xid, struct cifsSesInfo *ses, int first_time,
505 unicode_ssetup_strings(&bcc_ptr, ses, nls_cp); 507 unicode_ssetup_strings(&bcc_ptr, ses, nls_cp);
506 } else 508 } else
507 ascii_ssetup_strings(&bcc_ptr, ses, nls_cp); 509 ascii_ssetup_strings(&bcc_ptr, ses, nls_cp);
508 } else if (type == Kerberos) { 510 } else if (type == Kerberos || type == MSKerberos) {
509#ifdef CONFIG_CIFS_UPCALL 511#ifdef CONFIG_CIFS_UPCALL
510 struct cifs_spnego_msg *msg; 512 struct cifs_spnego_msg *msg;
511 spnego_key = cifs_get_spnego_key(ses); 513 spnego_key = cifs_get_spnego_key(ses);
@@ -516,6 +518,15 @@ CIFS_SessSetup(unsigned int xid, struct cifsSesInfo *ses, int first_time,
516 } 518 }
517 519
518 msg = spnego_key->payload.data; 520 msg = spnego_key->payload.data;
521 /* check version field to make sure that cifs.upcall is
522 sending us a response in an expected form */
523 if (msg->version != CIFS_SPNEGO_UPCALL_VERSION) {
524 cERROR(1, ("incorrect version of cifs.upcall (expected"
525 " %d but got %d)",
526 CIFS_SPNEGO_UPCALL_VERSION, msg->version));
527 rc = -EKEYREJECTED;
528 goto ssetup_exit;
529 }
519 /* bail out if key is too long */ 530 /* bail out if key is too long */
520 if (msg->sesskey_len > 531 if (msg->sesskey_len >
521 sizeof(ses->server->mac_signing_key.data.krb5)) { 532 sizeof(ses->server->mac_signing_key.data.krb5)) {
diff --git a/fs/compat.c b/fs/compat.c
index c9d1472e65c5..075d0509970d 100644
--- a/fs/compat.c
+++ b/fs/compat.c
@@ -792,8 +792,10 @@ static int compat_fillonedir(void *__buf, const char *name, int namlen,
792 if (buf->result) 792 if (buf->result)
793 return -EINVAL; 793 return -EINVAL;
794 d_ino = ino; 794 d_ino = ino;
795 if (sizeof(d_ino) < sizeof(ino) && d_ino != ino) 795 if (sizeof(d_ino) < sizeof(ino) && d_ino != ino) {
796 buf->result = -EOVERFLOW;
796 return -EOVERFLOW; 797 return -EOVERFLOW;
798 }
797 buf->result++; 799 buf->result++;
798 dirent = buf->dirent; 800 dirent = buf->dirent;
799 if (!access_ok(VERIFY_WRITE, dirent, 801 if (!access_ok(VERIFY_WRITE, dirent,
@@ -862,8 +864,10 @@ static int compat_filldir(void *__buf, const char *name, int namlen,
862 if (reclen > buf->count) 864 if (reclen > buf->count)
863 return -EINVAL; 865 return -EINVAL;
864 d_ino = ino; 866 d_ino = ino;
865 if (sizeof(d_ino) < sizeof(ino) && d_ino != ino) 867 if (sizeof(d_ino) < sizeof(ino) && d_ino != ino) {
868 buf->error = -EOVERFLOW;
866 return -EOVERFLOW; 869 return -EOVERFLOW;
870 }
867 dirent = buf->previous; 871 dirent = buf->previous;
868 if (dirent) { 872 if (dirent) {
869 if (__put_user(offset, &dirent->d_off)) 873 if (__put_user(offset, &dirent->d_off))
diff --git a/fs/configfs/dir.c b/fs/configfs/dir.c
index 7a8db78a91d2..8e93341f3e82 100644
--- a/fs/configfs/dir.c
+++ b/fs/configfs/dir.c
@@ -1311,16 +1311,18 @@ static int configfs_rmdir(struct inode *dir, struct dentry *dentry)
1311 * Ensure that no racing symlink() will make detach_prep() fail while 1311 * Ensure that no racing symlink() will make detach_prep() fail while
1312 * the new link is temporarily attached 1312 * the new link is temporarily attached
1313 */ 1313 */
1314 mutex_lock(&configfs_symlink_mutex);
1315 spin_lock(&configfs_dirent_lock);
1316 do { 1314 do {
1317 struct mutex *wait_mutex; 1315 struct mutex *wait_mutex;
1318 1316
1317 mutex_lock(&configfs_symlink_mutex);
1318 spin_lock(&configfs_dirent_lock);
1319 ret = configfs_detach_prep(dentry, &wait_mutex); 1319 ret = configfs_detach_prep(dentry, &wait_mutex);
1320 if (ret) { 1320 if (ret)
1321 configfs_detach_rollback(dentry); 1321 configfs_detach_rollback(dentry);
1322 spin_unlock(&configfs_dirent_lock); 1322 spin_unlock(&configfs_dirent_lock);
1323 mutex_unlock(&configfs_symlink_mutex); 1323 mutex_unlock(&configfs_symlink_mutex);
1324
1325 if (ret) {
1324 if (ret != -EAGAIN) { 1326 if (ret != -EAGAIN) {
1325 config_item_put(parent_item); 1327 config_item_put(parent_item);
1326 return ret; 1328 return ret;
@@ -1329,13 +1331,8 @@ static int configfs_rmdir(struct inode *dir, struct dentry *dentry)
1329 /* Wait until the racing operation terminates */ 1331 /* Wait until the racing operation terminates */
1330 mutex_lock(wait_mutex); 1332 mutex_lock(wait_mutex);
1331 mutex_unlock(wait_mutex); 1333 mutex_unlock(wait_mutex);
1332
1333 mutex_lock(&configfs_symlink_mutex);
1334 spin_lock(&configfs_dirent_lock);
1335 } 1334 }
1336 } while (ret == -EAGAIN); 1335 } while (ret == -EAGAIN);
1337 spin_unlock(&configfs_dirent_lock);
1338 mutex_unlock(&configfs_symlink_mutex);
1339 1336
1340 /* Get a working ref for the duration of this function */ 1337 /* Get a working ref for the duration of this function */
1341 item = configfs_get_config_item(dentry); 1338 item = configfs_get_config_item(dentry);
diff --git a/fs/dcache.c b/fs/dcache.c
index 101663d15e9f..80e93956aced 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -1236,7 +1236,7 @@ struct dentry *d_splice_alias(struct inode *inode, struct dentry *dentry)
1236 * If no entry exists with the exact case name, allocate new dentry with 1236 * If no entry exists with the exact case name, allocate new dentry with
1237 * the exact case, and return the spliced entry. 1237 * the exact case, and return the spliced entry.
1238 */ 1238 */
1239struct dentry *d_add_ci(struct inode *inode, struct dentry *dentry, 1239struct dentry *d_add_ci(struct dentry *dentry, struct inode *inode,
1240 struct qstr *name) 1240 struct qstr *name)
1241{ 1241{
1242 int error; 1242 int error;
diff --git a/fs/efs/namei.c b/fs/efs/namei.c
index 3a404e7fad53..291abb11e20e 100644
--- a/fs/efs/namei.c
+++ b/fs/efs/namei.c
@@ -74,8 +74,7 @@ struct dentry *efs_lookup(struct inode *dir, struct dentry *dentry, struct namei
74 } 74 }
75 unlock_kernel(); 75 unlock_kernel();
76 76
77 d_add(dentry, inode); 77 return d_splice_alias(inode, dentry);
78 return NULL;
79} 78}
80 79
81static struct inode *efs_nfs_get_inode(struct super_block *sb, u64 ino, 80static struct inode *efs_nfs_get_inode(struct super_block *sb, u64 ino,
diff --git a/fs/ext4/balloc.c b/fs/ext4/balloc.c
index 1ae5004e93fc..e9fa960ba6da 100644
--- a/fs/ext4/balloc.c
+++ b/fs/ext4/balloc.c
@@ -1626,6 +1626,9 @@ ext4_fsblk_t ext4_has_free_blocks(struct ext4_sb_info *sbi,
1626 free_blocks = 1626 free_blocks =
1627 percpu_counter_sum_and_set(&sbi->s_freeblocks_counter); 1627 percpu_counter_sum_and_set(&sbi->s_freeblocks_counter);
1628#endif 1628#endif
1629 if (free_blocks <= root_blocks)
1630 /* we don't have free space */
1631 return 0;
1629 if (free_blocks - root_blocks < nblocks) 1632 if (free_blocks - root_blocks < nblocks)
1630 return free_blocks - root_blocks; 1633 return free_blocks - root_blocks;
1631 return nblocks; 1634 return nblocks;
diff --git a/fs/ext4/dir.c b/fs/ext4/dir.c
index d3d23d73c08b..ec8e33b45219 100644
--- a/fs/ext4/dir.c
+++ b/fs/ext4/dir.c
@@ -411,7 +411,7 @@ static int call_filldir(struct file * filp, void * dirent,
411 get_dtype(sb, fname->file_type)); 411 get_dtype(sb, fname->file_type));
412 if (error) { 412 if (error) {
413 filp->f_pos = curr_pos; 413 filp->f_pos = curr_pos;
414 info->extra_fname = fname->next; 414 info->extra_fname = fname;
415 return error; 415 return error;
416 } 416 }
417 fname = fname->next; 417 fname = fname->next;
@@ -450,11 +450,21 @@ static int ext4_dx_readdir(struct file * filp,
450 * If there are any leftover names on the hash collision 450 * If there are any leftover names on the hash collision
451 * chain, return them first. 451 * chain, return them first.
452 */ 452 */
453 if (info->extra_fname && 453 if (info->extra_fname) {
454 call_filldir(filp, dirent, filldir, info->extra_fname)) 454 if (call_filldir(filp, dirent, filldir, info->extra_fname))
455 goto finished; 455 goto finished;
456 456
457 if (!info->curr_node) 457 info->extra_fname = NULL;
458 info->curr_node = rb_next(info->curr_node);
459 if (!info->curr_node) {
460 if (info->next_hash == ~0) {
461 filp->f_pos = EXT4_HTREE_EOF;
462 goto finished;
463 }
464 info->curr_hash = info->next_hash;
465 info->curr_minor_hash = 0;
466 }
467 } else if (!info->curr_node)
458 info->curr_node = rb_first(&info->root); 468 info->curr_node = rb_first(&info->root);
459 469
460 while (1) { 470 while (1) {
diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h
index 6c7924d9e358..295003241d3d 100644
--- a/fs/ext4/ext4.h
+++ b/fs/ext4/ext4.h
@@ -1072,6 +1072,8 @@ extern void ext4_set_inode_flags(struct inode *);
1072extern void ext4_get_inode_flags(struct ext4_inode_info *); 1072extern void ext4_get_inode_flags(struct ext4_inode_info *);
1073extern void ext4_set_aops(struct inode *inode); 1073extern void ext4_set_aops(struct inode *inode);
1074extern int ext4_writepage_trans_blocks(struct inode *); 1074extern int ext4_writepage_trans_blocks(struct inode *);
1075extern int ext4_meta_trans_blocks(struct inode *, int nrblocks, int idxblocks);
1076extern int ext4_chunk_trans_blocks(struct inode *, int nrblocks);
1075extern int ext4_block_truncate_page(handle_t *handle, 1077extern int ext4_block_truncate_page(handle_t *handle,
1076 struct address_space *mapping, loff_t from); 1078 struct address_space *mapping, loff_t from);
1077extern int ext4_page_mkwrite(struct vm_area_struct *vma, struct page *page); 1079extern int ext4_page_mkwrite(struct vm_area_struct *vma, struct page *page);
@@ -1227,6 +1229,8 @@ extern const struct inode_operations ext4_fast_symlink_inode_operations;
1227/* extents.c */ 1229/* extents.c */
1228extern int ext4_ext_tree_init(handle_t *handle, struct inode *); 1230extern int ext4_ext_tree_init(handle_t *handle, struct inode *);
1229extern int ext4_ext_writepage_trans_blocks(struct inode *, int); 1231extern int ext4_ext_writepage_trans_blocks(struct inode *, int);
1232extern int ext4_ext_index_trans_blocks(struct inode *inode, int nrblocks,
1233 int chunk);
1230extern int ext4_ext_get_blocks(handle_t *handle, struct inode *inode, 1234extern int ext4_ext_get_blocks(handle_t *handle, struct inode *inode,
1231 ext4_lblk_t iblock, 1235 ext4_lblk_t iblock,
1232 unsigned long max_blocks, struct buffer_head *bh_result, 1236 unsigned long max_blocks, struct buffer_head *bh_result,
diff --git a/fs/ext4/ext4_extents.h b/fs/ext4/ext4_extents.h
index 6c166c0a54b7..d33dc56d6986 100644
--- a/fs/ext4/ext4_extents.h
+++ b/fs/ext4/ext4_extents.h
@@ -216,7 +216,9 @@ extern int ext4_ext_calc_metadata_amount(struct inode *inode, int blocks);
216extern ext4_fsblk_t idx_pblock(struct ext4_extent_idx *); 216extern ext4_fsblk_t idx_pblock(struct ext4_extent_idx *);
217extern void ext4_ext_store_pblock(struct ext4_extent *, ext4_fsblk_t); 217extern void ext4_ext_store_pblock(struct ext4_extent *, ext4_fsblk_t);
218extern int ext4_extent_tree_init(handle_t *, struct inode *); 218extern int ext4_extent_tree_init(handle_t *, struct inode *);
219extern int ext4_ext_calc_credits_for_insert(struct inode *, struct ext4_ext_path *); 219extern int ext4_ext_calc_credits_for_single_extent(struct inode *inode,
220 int num,
221 struct ext4_ext_path *path);
220extern int ext4_ext_try_to_merge(struct inode *inode, 222extern int ext4_ext_try_to_merge(struct inode *inode,
221 struct ext4_ext_path *path, 223 struct ext4_ext_path *path,
222 struct ext4_extent *); 224 struct ext4_extent *);
diff --git a/fs/ext4/ext4_jbd2.h b/fs/ext4/ext4_jbd2.h
index eb8bc3afe6e9..b455c685a98b 100644
--- a/fs/ext4/ext4_jbd2.h
+++ b/fs/ext4/ext4_jbd2.h
@@ -51,6 +51,14 @@
51 EXT4_XATTR_TRANS_BLOCKS - 2 + \ 51 EXT4_XATTR_TRANS_BLOCKS - 2 + \
52 2*EXT4_QUOTA_TRANS_BLOCKS(sb)) 52 2*EXT4_QUOTA_TRANS_BLOCKS(sb))
53 53
54/*
55 * Define the number of metadata blocks we need to account to modify data.
56 *
57 * This include super block, inode block, quota blocks and xattr blocks
58 */
59#define EXT4_META_TRANS_BLOCKS(sb) (EXT4_XATTR_TRANS_BLOCKS + \
60 2*EXT4_QUOTA_TRANS_BLOCKS(sb))
61
54/* Delete operations potentially hit one directory's namespace plus an 62/* Delete operations potentially hit one directory's namespace plus an
55 * entire inode, plus arbitrary amounts of bitmap/indirection data. Be 63 * entire inode, plus arbitrary amounts of bitmap/indirection data. Be
56 * generous. We can grow the delete transaction later if necessary. */ 64 * generous. We can grow the delete transaction later if necessary. */
diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c
index 612c3d2c3824..b24d3c53f20c 100644
--- a/fs/ext4/extents.c
+++ b/fs/ext4/extents.c
@@ -1747,54 +1747,61 @@ static int ext4_ext_rm_idx(handle_t *handle, struct inode *inode,
1747} 1747}
1748 1748
1749/* 1749/*
1750 * ext4_ext_calc_credits_for_insert: 1750 * ext4_ext_calc_credits_for_single_extent:
1751 * This routine returns max. credits that the extent tree can consume. 1751 * This routine returns max. credits that needed to insert an extent
1752 * It should be OK for low-performance paths like ->writepage() 1752 * to the extent tree.
1753 * To allow many writing processes to fit into a single transaction, 1753 * When pass the actual path, the caller should calculate credits
1754 * the caller should calculate credits under i_data_sem and 1754 * under i_data_sem.
1755 * pass the actual path.
1756 */ 1755 */
1757int ext4_ext_calc_credits_for_insert(struct inode *inode, 1756int ext4_ext_calc_credits_for_single_extent(struct inode *inode, int nrblocks,
1758 struct ext4_ext_path *path) 1757 struct ext4_ext_path *path)
1759{ 1758{
1760 int depth, needed;
1761
1762 if (path) { 1759 if (path) {
1760 int depth = ext_depth(inode);
1761 int ret = 0;
1762
1763 /* probably there is space in leaf? */ 1763 /* probably there is space in leaf? */
1764 depth = ext_depth(inode);
1765 if (le16_to_cpu(path[depth].p_hdr->eh_entries) 1764 if (le16_to_cpu(path[depth].p_hdr->eh_entries)
1766 < le16_to_cpu(path[depth].p_hdr->eh_max)) 1765 < le16_to_cpu(path[depth].p_hdr->eh_max)) {
1767 return 1;
1768 }
1769 1766
1770 /* 1767 /*
1771 * given 32-bit logical block (4294967296 blocks), max. tree 1768 * There are some space in the leaf tree, no
1772 * can be 4 levels in depth -- 4 * 340^4 == 53453440000. 1769 * need to account for leaf block credit
1773 * Let's also add one more level for imbalance. 1770 *
1774 */ 1771 * bitmaps and block group descriptor blocks
1775 depth = 5; 1772 * and other metadat blocks still need to be
1776 1773 * accounted.
1777 /* allocation of new data block(s) */ 1774 */
1778 needed = 2; 1775 /* 1 bitmap, 1 block group descriptor */
1776 ret = 2 + EXT4_META_TRANS_BLOCKS(inode->i_sb);
1777 }
1778 }
1779 1779
1780 /* 1780 return ext4_chunk_trans_blocks(inode, nrblocks);
1781 * tree can be full, so it would need to grow in depth: 1781}
1782 * we need one credit to modify old root, credits for
1783 * new root will be added in split accounting
1784 */
1785 needed += 1;
1786 1782
1787 /* 1783/*
1788 * Index split can happen, we would need: 1784 * How many index/leaf blocks need to change/allocate to modify nrblocks?
1789 * allocate intermediate indexes (bitmap + group) 1785 *
1790 * + change two blocks at each level, but root (already included) 1786 * if nrblocks are fit in a single extent (chunk flag is 1), then
1791 */ 1787 * in the worse case, each tree level index/leaf need to be changed
1792 needed += (depth * 2) + (depth * 2); 1788 * if the tree split due to insert a new extent, then the old tree
1789 * index/leaf need to be updated too
1790 *
1791 * If the nrblocks are discontiguous, they could cause
1792 * the whole tree split more than once, but this is really rare.
1793 */
1794int ext4_ext_index_trans_blocks(struct inode *inode, int nrblocks, int chunk)
1795{
1796 int index;
1797 int depth = ext_depth(inode);
1793 1798
1794 /* any allocation modifies superblock */ 1799 if (chunk)
1795 needed += 1; 1800 index = depth * 2;
1801 else
1802 index = depth * 3;
1796 1803
1797 return needed; 1804 return index;
1798} 1805}
1799 1806
1800static int ext4_remove_blocks(handle_t *handle, struct inode *inode, 1807static int ext4_remove_blocks(handle_t *handle, struct inode *inode,
@@ -1921,9 +1928,7 @@ ext4_ext_rm_leaf(handle_t *handle, struct inode *inode,
1921 correct_index = 1; 1928 correct_index = 1;
1922 credits += (ext_depth(inode)) + 1; 1929 credits += (ext_depth(inode)) + 1;
1923 } 1930 }
1924#ifdef CONFIG_QUOTA
1925 credits += 2 * EXT4_QUOTA_TRANS_BLOCKS(inode->i_sb); 1931 credits += 2 * EXT4_QUOTA_TRANS_BLOCKS(inode->i_sb);
1926#endif
1927 1932
1928 err = ext4_ext_journal_restart(handle, credits); 1933 err = ext4_ext_journal_restart(handle, credits);
1929 if (err) 1934 if (err)
@@ -2805,7 +2810,7 @@ void ext4_ext_truncate(struct inode *inode)
2805 /* 2810 /*
2806 * probably first extent we're gonna free will be last in block 2811 * probably first extent we're gonna free will be last in block
2807 */ 2812 */
2808 err = ext4_writepage_trans_blocks(inode) + 3; 2813 err = ext4_writepage_trans_blocks(inode);
2809 handle = ext4_journal_start(inode, err); 2814 handle = ext4_journal_start(inode, err);
2810 if (IS_ERR(handle)) 2815 if (IS_ERR(handle))
2811 return; 2816 return;
@@ -2819,7 +2824,7 @@ void ext4_ext_truncate(struct inode *inode)
2819 down_write(&EXT4_I(inode)->i_data_sem); 2824 down_write(&EXT4_I(inode)->i_data_sem);
2820 ext4_ext_invalidate_cache(inode); 2825 ext4_ext_invalidate_cache(inode);
2821 2826
2822 ext4_mb_discard_inode_preallocations(inode); 2827 ext4_discard_reservation(inode);
2823 2828
2824 /* 2829 /*
2825 * TODO: optimization is possible here. 2830 * TODO: optimization is possible here.
@@ -2858,27 +2863,6 @@ out_stop:
2858 ext4_journal_stop(handle); 2863 ext4_journal_stop(handle);
2859} 2864}
2860 2865
2861/*
2862 * ext4_ext_writepage_trans_blocks:
2863 * calculate max number of blocks we could modify
2864 * in order to allocate new block for an inode
2865 */
2866int ext4_ext_writepage_trans_blocks(struct inode *inode, int num)
2867{
2868 int needed;
2869
2870 needed = ext4_ext_calc_credits_for_insert(inode, NULL);
2871
2872 /* caller wants to allocate num blocks, but note it includes sb */
2873 needed = needed * num - (num - 1);
2874
2875#ifdef CONFIG_QUOTA
2876 needed += 2 * EXT4_QUOTA_TRANS_BLOCKS(inode->i_sb);
2877#endif
2878
2879 return needed;
2880}
2881
2882static void ext4_falloc_update_inode(struct inode *inode, 2866static void ext4_falloc_update_inode(struct inode *inode,
2883 int mode, loff_t new_size, int update_ctime) 2867 int mode, loff_t new_size, int update_ctime)
2884{ 2868{
@@ -2939,10 +2923,9 @@ long ext4_fallocate(struct inode *inode, int mode, loff_t offset, loff_t len)
2939 max_blocks = (EXT4_BLOCK_ALIGN(len + offset, blkbits) >> blkbits) 2923 max_blocks = (EXT4_BLOCK_ALIGN(len + offset, blkbits) >> blkbits)
2940 - block; 2924 - block;
2941 /* 2925 /*
2942 * credits to insert 1 extent into extent tree + buffers to be able to 2926 * credits to insert 1 extent into extent tree
2943 * modify 1 super block, 1 block bitmap and 1 group descriptor.
2944 */ 2927 */
2945 credits = EXT4_DATA_TRANS_BLOCKS(inode->i_sb) + 3; 2928 credits = ext4_chunk_trans_blocks(inode, max_blocks);
2946 mutex_lock(&inode->i_mutex); 2929 mutex_lock(&inode->i_mutex);
2947retry: 2930retry:
2948 while (ret >= 0 && ret < max_blocks) { 2931 while (ret >= 0 && ret < max_blocks) {
diff --git a/fs/ext4/ialloc.c b/fs/ext4/ialloc.c
index 655e760212b8..f344834bbf58 100644
--- a/fs/ext4/ialloc.c
+++ b/fs/ext4/ialloc.c
@@ -351,7 +351,7 @@ find_close_to_parent:
351 goto found_flexbg; 351 goto found_flexbg;
352 } 352 }
353 353
354 if (best_flex < 0 || 354 if (flex_group[best_flex].free_inodes == 0 ||
355 (flex_group[i].free_blocks > 355 (flex_group[i].free_blocks >
356 flex_group[best_flex].free_blocks && 356 flex_group[best_flex].free_blocks &&
357 flex_group[i].free_inodes)) 357 flex_group[i].free_inodes))
diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
index 59fbbe899acc..7e91913e325b 100644
--- a/fs/ext4/inode.c
+++ b/fs/ext4/inode.c
@@ -41,6 +41,8 @@
41#include "acl.h" 41#include "acl.h"
42#include "ext4_extents.h" 42#include "ext4_extents.h"
43 43
44#define MPAGE_DA_EXTENT_TAIL 0x01
45
44static inline int ext4_begin_ordered_truncate(struct inode *inode, 46static inline int ext4_begin_ordered_truncate(struct inode *inode,
45 loff_t new_size) 47 loff_t new_size)
46{ 48{
@@ -1005,6 +1007,9 @@ static int ext4_indirect_calc_metadata_amount(struct inode *inode, int blocks)
1005 */ 1007 */
1006static int ext4_calc_metadata_amount(struct inode *inode, int blocks) 1008static int ext4_calc_metadata_amount(struct inode *inode, int blocks)
1007{ 1009{
1010 if (!blocks)
1011 return 0;
1012
1008 if (EXT4_I(inode)->i_flags & EXT4_EXTENTS_FL) 1013 if (EXT4_I(inode)->i_flags & EXT4_EXTENTS_FL)
1009 return ext4_ext_calc_metadata_amount(inode, blocks); 1014 return ext4_ext_calc_metadata_amount(inode, blocks);
1010 1015
@@ -1041,18 +1046,6 @@ static void ext4_da_update_reserve_space(struct inode *inode, int used)
1041 spin_unlock(&EXT4_I(inode)->i_block_reservation_lock); 1046 spin_unlock(&EXT4_I(inode)->i_block_reservation_lock);
1042} 1047}
1043 1048
1044/* Maximum number of blocks we map for direct IO at once. */
1045#define DIO_MAX_BLOCKS 4096
1046/*
1047 * Number of credits we need for writing DIO_MAX_BLOCKS:
1048 * We need sb + group descriptor + bitmap + inode -> 4
1049 * For B blocks with A block pointers per block we need:
1050 * 1 (triple ind.) + (B/A/A + 2) (doubly ind.) + (B/A + 2) (indirect).
1051 * If we plug in 4096 for B and 256 for A (for 1KB block size), we get 25.
1052 */
1053#define DIO_CREDITS 25
1054
1055
1056/* 1049/*
1057 * The ext4_get_blocks_wrap() function try to look up the requested blocks, 1050 * The ext4_get_blocks_wrap() function try to look up the requested blocks,
1058 * and returns if the blocks are already mapped. 1051 * and returns if the blocks are already mapped.
@@ -1164,19 +1157,23 @@ int ext4_get_blocks_wrap(handle_t *handle, struct inode *inode, sector_t block,
1164 return retval; 1157 return retval;
1165} 1158}
1166 1159
1160/* Maximum number of blocks we map for direct IO at once. */
1161#define DIO_MAX_BLOCKS 4096
1162
1167static int ext4_get_block(struct inode *inode, sector_t iblock, 1163static int ext4_get_block(struct inode *inode, sector_t iblock,
1168 struct buffer_head *bh_result, int create) 1164 struct buffer_head *bh_result, int create)
1169{ 1165{
1170 handle_t *handle = ext4_journal_current_handle(); 1166 handle_t *handle = ext4_journal_current_handle();
1171 int ret = 0, started = 0; 1167 int ret = 0, started = 0;
1172 unsigned max_blocks = bh_result->b_size >> inode->i_blkbits; 1168 unsigned max_blocks = bh_result->b_size >> inode->i_blkbits;
1169 int dio_credits;
1173 1170
1174 if (create && !handle) { 1171 if (create && !handle) {
1175 /* Direct IO write... */ 1172 /* Direct IO write... */
1176 if (max_blocks > DIO_MAX_BLOCKS) 1173 if (max_blocks > DIO_MAX_BLOCKS)
1177 max_blocks = DIO_MAX_BLOCKS; 1174 max_blocks = DIO_MAX_BLOCKS;
1178 handle = ext4_journal_start(inode, DIO_CREDITS + 1175 dio_credits = ext4_chunk_trans_blocks(inode, max_blocks);
1179 2 * EXT4_QUOTA_TRANS_BLOCKS(inode->i_sb)); 1176 handle = ext4_journal_start(inode, dio_credits);
1180 if (IS_ERR(handle)) { 1177 if (IS_ERR(handle)) {
1181 ret = PTR_ERR(handle); 1178 ret = PTR_ERR(handle);
1182 goto out; 1179 goto out;
@@ -1559,7 +1556,25 @@ static void ext4_da_release_space(struct inode *inode, int to_free)
1559 struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb); 1556 struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
1560 int total, mdb, mdb_free, release; 1557 int total, mdb, mdb_free, release;
1561 1558
1559 if (!to_free)
1560 return; /* Nothing to release, exit */
1561
1562 spin_lock(&EXT4_I(inode)->i_block_reservation_lock); 1562 spin_lock(&EXT4_I(inode)->i_block_reservation_lock);
1563
1564 if (!EXT4_I(inode)->i_reserved_data_blocks) {
1565 /*
1566 * if there is no reserved blocks, but we try to free some
1567 * then the counter is messed up somewhere.
1568 * but since this function is called from invalidate
1569 * page, it's harmless to return without any action
1570 */
1571 printk(KERN_INFO "ext4 delalloc try to release %d reserved "
1572 "blocks for inode %lu, but there is no reserved "
1573 "data blocks\n", to_free, inode->i_ino);
1574 spin_unlock(&EXT4_I(inode)->i_block_reservation_lock);
1575 return;
1576 }
1577
1563 /* recalculate the number of metablocks still need to be reserved */ 1578 /* recalculate the number of metablocks still need to be reserved */
1564 total = EXT4_I(inode)->i_reserved_data_blocks - to_free; 1579 total = EXT4_I(inode)->i_reserved_data_blocks - to_free;
1565 mdb = ext4_calc_metadata_amount(inode, total); 1580 mdb = ext4_calc_metadata_amount(inode, total);
@@ -1613,11 +1628,13 @@ struct mpage_da_data {
1613 unsigned long first_page, next_page; /* extent of pages */ 1628 unsigned long first_page, next_page; /* extent of pages */
1614 get_block_t *get_block; 1629 get_block_t *get_block;
1615 struct writeback_control *wbc; 1630 struct writeback_control *wbc;
1631 int io_done;
1632 long pages_written;
1616}; 1633};
1617 1634
1618/* 1635/*
1619 * mpage_da_submit_io - walks through extent of pages and try to write 1636 * mpage_da_submit_io - walks through extent of pages and try to write
1620 * them with __mpage_writepage() 1637 * them with writepage() call back
1621 * 1638 *
1622 * @mpd->inode: inode 1639 * @mpd->inode: inode
1623 * @mpd->first_page: first page of the extent 1640 * @mpd->first_page: first page of the extent
@@ -1632,18 +1649,11 @@ struct mpage_da_data {
1632static int mpage_da_submit_io(struct mpage_da_data *mpd) 1649static int mpage_da_submit_io(struct mpage_da_data *mpd)
1633{ 1650{
1634 struct address_space *mapping = mpd->inode->i_mapping; 1651 struct address_space *mapping = mpd->inode->i_mapping;
1635 struct mpage_data mpd_pp = {
1636 .bio = NULL,
1637 .last_block_in_bio = 0,
1638 .get_block = mpd->get_block,
1639 .use_writepage = 1,
1640 };
1641 int ret = 0, err, nr_pages, i; 1652 int ret = 0, err, nr_pages, i;
1642 unsigned long index, end; 1653 unsigned long index, end;
1643 struct pagevec pvec; 1654 struct pagevec pvec;
1644 1655
1645 BUG_ON(mpd->next_page <= mpd->first_page); 1656 BUG_ON(mpd->next_page <= mpd->first_page);
1646
1647 pagevec_init(&pvec, 0); 1657 pagevec_init(&pvec, 0);
1648 index = mpd->first_page; 1658 index = mpd->first_page;
1649 end = mpd->next_page - 1; 1659 end = mpd->next_page - 1;
@@ -1661,8 +1671,9 @@ static int mpage_da_submit_io(struct mpage_da_data *mpd)
1661 break; 1671 break;
1662 index++; 1672 index++;
1663 1673
1664 err = __mpage_writepage(page, mpd->wbc, &mpd_pp); 1674 err = mapping->a_ops->writepage(page, mpd->wbc);
1665 1675 if (!err)
1676 mpd->pages_written++;
1666 /* 1677 /*
1667 * In error case, we have to continue because 1678 * In error case, we have to continue because
1668 * remaining pages are still locked 1679 * remaining pages are still locked
@@ -1673,9 +1684,6 @@ static int mpage_da_submit_io(struct mpage_da_data *mpd)
1673 } 1684 }
1674 pagevec_release(&pvec); 1685 pagevec_release(&pvec);
1675 } 1686 }
1676 if (mpd_pp.bio)
1677 mpage_bio_submit(WRITE, mpd_pp.bio);
1678
1679 return ret; 1687 return ret;
1680} 1688}
1681 1689
@@ -1698,7 +1706,7 @@ static void mpage_put_bnr_to_bhs(struct mpage_da_data *mpd, sector_t logical,
1698 int blocks = exbh->b_size >> inode->i_blkbits; 1706 int blocks = exbh->b_size >> inode->i_blkbits;
1699 sector_t pblock = exbh->b_blocknr, cur_logical; 1707 sector_t pblock = exbh->b_blocknr, cur_logical;
1700 struct buffer_head *head, *bh; 1708 struct buffer_head *head, *bh;
1701 unsigned long index, end; 1709 pgoff_t index, end;
1702 struct pagevec pvec; 1710 struct pagevec pvec;
1703 int nr_pages, i; 1711 int nr_pages, i;
1704 1712
@@ -1741,6 +1749,13 @@ static void mpage_put_bnr_to_bhs(struct mpage_da_data *mpd, sector_t logical,
1741 if (buffer_delay(bh)) { 1749 if (buffer_delay(bh)) {
1742 bh->b_blocknr = pblock; 1750 bh->b_blocknr = pblock;
1743 clear_buffer_delay(bh); 1751 clear_buffer_delay(bh);
1752 bh->b_bdev = inode->i_sb->s_bdev;
1753 } else if (buffer_unwritten(bh)) {
1754 bh->b_blocknr = pblock;
1755 clear_buffer_unwritten(bh);
1756 set_buffer_mapped(bh);
1757 set_buffer_new(bh);
1758 bh->b_bdev = inode->i_sb->s_bdev;
1744 } else if (buffer_mapped(bh)) 1759 } else if (buffer_mapped(bh))
1745 BUG_ON(bh->b_blocknr != pblock); 1760 BUG_ON(bh->b_blocknr != pblock);
1746 1761
@@ -1776,13 +1791,11 @@ static inline void __unmap_underlying_blocks(struct inode *inode,
1776 * 1791 *
1777 * The function skips space we know is already mapped to disk blocks. 1792 * The function skips space we know is already mapped to disk blocks.
1778 * 1793 *
1779 * The function ignores errors ->get_block() returns, thus real
1780 * error handling is postponed to __mpage_writepage()
1781 */ 1794 */
1782static void mpage_da_map_blocks(struct mpage_da_data *mpd) 1795static void mpage_da_map_blocks(struct mpage_da_data *mpd)
1783{ 1796{
1797 int err = 0;
1784 struct buffer_head *lbh = &mpd->lbh; 1798 struct buffer_head *lbh = &mpd->lbh;
1785 int err = 0, remain = lbh->b_size;
1786 sector_t next = lbh->b_blocknr; 1799 sector_t next = lbh->b_blocknr;
1787 struct buffer_head new; 1800 struct buffer_head new;
1788 1801
@@ -1792,38 +1805,36 @@ static void mpage_da_map_blocks(struct mpage_da_data *mpd)
1792 if (buffer_mapped(lbh) && !buffer_delay(lbh)) 1805 if (buffer_mapped(lbh) && !buffer_delay(lbh))
1793 return; 1806 return;
1794 1807
1795 while (remain) { 1808 new.b_state = lbh->b_state;
1796 new.b_state = lbh->b_state; 1809 new.b_blocknr = 0;
1797 new.b_blocknr = 0; 1810 new.b_size = lbh->b_size;
1798 new.b_size = remain;
1799 err = mpd->get_block(mpd->inode, next, &new, 1);
1800 if (err) {
1801 /*
1802 * Rather than implement own error handling
1803 * here, we just leave remaining blocks
1804 * unallocated and try again with ->writepage()
1805 */
1806 break;
1807 }
1808 BUG_ON(new.b_size == 0);
1809 1811
1810 if (buffer_new(&new)) 1812 /*
1811 __unmap_underlying_blocks(mpd->inode, &new); 1813 * If we didn't accumulate anything
1814 * to write simply return
1815 */
1816 if (!new.b_size)
1817 return;
1818 err = mpd->get_block(mpd->inode, next, &new, 1);
1819 if (err)
1820 return;
1821 BUG_ON(new.b_size == 0);
1812 1822
1813 /* 1823 if (buffer_new(&new))
1814 * If blocks are delayed marked, we need to 1824 __unmap_underlying_blocks(mpd->inode, &new);
1815 * put actual blocknr and drop delayed bit
1816 */
1817 if (buffer_delay(lbh))
1818 mpage_put_bnr_to_bhs(mpd, next, &new);
1819 1825
1820 /* go for the remaining blocks */ 1826 /*
1821 next += new.b_size >> mpd->inode->i_blkbits; 1827 * If blocks are delayed marked, we need to
1822 remain -= new.b_size; 1828 * put actual blocknr and drop delayed bit
1823 } 1829 */
1830 if (buffer_delay(lbh) || buffer_unwritten(lbh))
1831 mpage_put_bnr_to_bhs(mpd, next, &new);
1832
1833 return;
1824} 1834}
1825 1835
1826#define BH_FLAGS ((1 << BH_Uptodate) | (1 << BH_Mapped) | (1 << BH_Delay)) 1836#define BH_FLAGS ((1 << BH_Uptodate) | (1 << BH_Mapped) | \
1837 (1 << BH_Delay) | (1 << BH_Unwritten))
1827 1838
1828/* 1839/*
1829 * mpage_add_bh_to_extent - try to add one more block to extent of blocks 1840 * mpage_add_bh_to_extent - try to add one more block to extent of blocks
@@ -1837,41 +1848,61 @@ static void mpage_da_map_blocks(struct mpage_da_data *mpd)
1837static void mpage_add_bh_to_extent(struct mpage_da_data *mpd, 1848static void mpage_add_bh_to_extent(struct mpage_da_data *mpd,
1838 sector_t logical, struct buffer_head *bh) 1849 sector_t logical, struct buffer_head *bh)
1839{ 1850{
1840 struct buffer_head *lbh = &mpd->lbh;
1841 sector_t next; 1851 sector_t next;
1852 size_t b_size = bh->b_size;
1853 struct buffer_head *lbh = &mpd->lbh;
1854 int nrblocks = lbh->b_size >> mpd->inode->i_blkbits;
1842 1855
1843 next = lbh->b_blocknr + (lbh->b_size >> mpd->inode->i_blkbits); 1856 /* check if thereserved journal credits might overflow */
1844 1857 if (!(EXT4_I(mpd->inode)->i_flags & EXT4_EXTENTS_FL)) {
1858 if (nrblocks >= EXT4_MAX_TRANS_DATA) {
1859 /*
1860 * With non-extent format we are limited by the journal
1861 * credit available. Total credit needed to insert
1862 * nrblocks contiguous blocks is dependent on the
1863 * nrblocks. So limit nrblocks.
1864 */
1865 goto flush_it;
1866 } else if ((nrblocks + (b_size >> mpd->inode->i_blkbits)) >
1867 EXT4_MAX_TRANS_DATA) {
1868 /*
1869 * Adding the new buffer_head would make it cross the
1870 * allowed limit for which we have journal credit
1871 * reserved. So limit the new bh->b_size
1872 */
1873 b_size = (EXT4_MAX_TRANS_DATA - nrblocks) <<
1874 mpd->inode->i_blkbits;
1875 /* we will do mpage_da_submit_io in the next loop */
1876 }
1877 }
1845 /* 1878 /*
1846 * First block in the extent 1879 * First block in the extent
1847 */ 1880 */
1848 if (lbh->b_size == 0) { 1881 if (lbh->b_size == 0) {
1849 lbh->b_blocknr = logical; 1882 lbh->b_blocknr = logical;
1850 lbh->b_size = bh->b_size; 1883 lbh->b_size = b_size;
1851 lbh->b_state = bh->b_state & BH_FLAGS; 1884 lbh->b_state = bh->b_state & BH_FLAGS;
1852 return; 1885 return;
1853 } 1886 }
1854 1887
1888 next = lbh->b_blocknr + nrblocks;
1855 /* 1889 /*
1856 * Can we merge the block to our big extent? 1890 * Can we merge the block to our big extent?
1857 */ 1891 */
1858 if (logical == next && (bh->b_state & BH_FLAGS) == lbh->b_state) { 1892 if (logical == next && (bh->b_state & BH_FLAGS) == lbh->b_state) {
1859 lbh->b_size += bh->b_size; 1893 lbh->b_size += b_size;
1860 return; 1894 return;
1861 } 1895 }
1862 1896
1897flush_it:
1863 /* 1898 /*
1864 * We couldn't merge the block to our extent, so we 1899 * We couldn't merge the block to our extent, so we
1865 * need to flush current extent and start new one 1900 * need to flush current extent and start new one
1866 */ 1901 */
1867 mpage_da_map_blocks(mpd); 1902 mpage_da_map_blocks(mpd);
1868 1903 mpage_da_submit_io(mpd);
1869 /* 1904 mpd->io_done = 1;
1870 * Now start a new extent 1905 return;
1871 */
1872 lbh->b_size = bh->b_size;
1873 lbh->b_state = bh->b_state & BH_FLAGS;
1874 lbh->b_blocknr = logical;
1875} 1906}
1876 1907
1877/* 1908/*
@@ -1891,17 +1922,35 @@ static int __mpage_da_writepage(struct page *page,
1891 struct buffer_head *bh, *head, fake; 1922 struct buffer_head *bh, *head, fake;
1892 sector_t logical; 1923 sector_t logical;
1893 1924
1925 if (mpd->io_done) {
1926 /*
1927 * Rest of the page in the page_vec
1928 * redirty then and skip then. We will
1929 * try to to write them again after
1930 * starting a new transaction
1931 */
1932 redirty_page_for_writepage(wbc, page);
1933 unlock_page(page);
1934 return MPAGE_DA_EXTENT_TAIL;
1935 }
1894 /* 1936 /*
1895 * Can we merge this page to current extent? 1937 * Can we merge this page to current extent?
1896 */ 1938 */
1897 if (mpd->next_page != page->index) { 1939 if (mpd->next_page != page->index) {
1898 /* 1940 /*
1899 * Nope, we can't. So, we map non-allocated blocks 1941 * Nope, we can't. So, we map non-allocated blocks
1900 * and start IO on them using __mpage_writepage() 1942 * and start IO on them using writepage()
1901 */ 1943 */
1902 if (mpd->next_page != mpd->first_page) { 1944 if (mpd->next_page != mpd->first_page) {
1903 mpage_da_map_blocks(mpd); 1945 mpage_da_map_blocks(mpd);
1904 mpage_da_submit_io(mpd); 1946 mpage_da_submit_io(mpd);
1947 /*
1948 * skip rest of the page in the page_vec
1949 */
1950 mpd->io_done = 1;
1951 redirty_page_for_writepage(wbc, page);
1952 unlock_page(page);
1953 return MPAGE_DA_EXTENT_TAIL;
1905 } 1954 }
1906 1955
1907 /* 1956 /*
@@ -1932,6 +1981,8 @@ static int __mpage_da_writepage(struct page *page,
1932 set_buffer_dirty(bh); 1981 set_buffer_dirty(bh);
1933 set_buffer_uptodate(bh); 1982 set_buffer_uptodate(bh);
1934 mpage_add_bh_to_extent(mpd, logical, bh); 1983 mpage_add_bh_to_extent(mpd, logical, bh);
1984 if (mpd->io_done)
1985 return MPAGE_DA_EXTENT_TAIL;
1935 } else { 1986 } else {
1936 /* 1987 /*
1937 * Page with regular buffer heads, just add all dirty ones 1988 * Page with regular buffer heads, just add all dirty ones
@@ -1940,8 +1991,12 @@ static int __mpage_da_writepage(struct page *page,
1940 bh = head; 1991 bh = head;
1941 do { 1992 do {
1942 BUG_ON(buffer_locked(bh)); 1993 BUG_ON(buffer_locked(bh));
1943 if (buffer_dirty(bh)) 1994 if (buffer_dirty(bh) &&
1995 (!buffer_mapped(bh) || buffer_delay(bh))) {
1944 mpage_add_bh_to_extent(mpd, logical, bh); 1996 mpage_add_bh_to_extent(mpd, logical, bh);
1997 if (mpd->io_done)
1998 return MPAGE_DA_EXTENT_TAIL;
1999 }
1945 logical++; 2000 logical++;
1946 } while ((bh = bh->b_this_page) != head); 2001 } while ((bh = bh->b_this_page) != head);
1947 } 2002 }
@@ -1960,22 +2015,13 @@ static int __mpage_da_writepage(struct page *page,
1960 * 2015 *
1961 * This is a library function, which implements the writepages() 2016 * This is a library function, which implements the writepages()
1962 * address_space_operation. 2017 * address_space_operation.
1963 *
1964 * In order to avoid duplication of logic that deals with partial pages,
1965 * multiple bio per page, etc, we find non-allocated blocks, allocate
1966 * them with minimal calls to ->get_block() and re-use __mpage_writepage()
1967 *
1968 * It's important that we call __mpage_writepage() only once for each
1969 * involved page, otherwise we'd have to implement more complicated logic
1970 * to deal with pages w/o PG_lock or w/ PG_writeback and so on.
1971 *
1972 * See comments to mpage_writepages()
1973 */ 2018 */
1974static int mpage_da_writepages(struct address_space *mapping, 2019static int mpage_da_writepages(struct address_space *mapping,
1975 struct writeback_control *wbc, 2020 struct writeback_control *wbc,
1976 get_block_t get_block) 2021 get_block_t get_block)
1977{ 2022{
1978 struct mpage_da_data mpd; 2023 struct mpage_da_data mpd;
2024 long to_write;
1979 int ret; 2025 int ret;
1980 2026
1981 if (!get_block) 2027 if (!get_block)
@@ -1989,17 +2035,22 @@ static int mpage_da_writepages(struct address_space *mapping,
1989 mpd.first_page = 0; 2035 mpd.first_page = 0;
1990 mpd.next_page = 0; 2036 mpd.next_page = 0;
1991 mpd.get_block = get_block; 2037 mpd.get_block = get_block;
2038 mpd.io_done = 0;
2039 mpd.pages_written = 0;
2040
2041 to_write = wbc->nr_to_write;
1992 2042
1993 ret = write_cache_pages(mapping, wbc, __mpage_da_writepage, &mpd); 2043 ret = write_cache_pages(mapping, wbc, __mpage_da_writepage, &mpd);
1994 2044
1995 /* 2045 /*
1996 * Handle last extent of pages 2046 * Handle last extent of pages
1997 */ 2047 */
1998 if (mpd.next_page != mpd.first_page) { 2048 if (!mpd.io_done && mpd.next_page != mpd.first_page) {
1999 mpage_da_map_blocks(&mpd); 2049 mpage_da_map_blocks(&mpd);
2000 mpage_da_submit_io(&mpd); 2050 mpage_da_submit_io(&mpd);
2001 } 2051 }
2002 2052
2053 wbc->nr_to_write = to_write - mpd.pages_written;
2003 return ret; 2054 return ret;
2004} 2055}
2005 2056
@@ -2204,63 +2255,95 @@ static int ext4_da_writepage(struct page *page,
2204} 2255}
2205 2256
2206/* 2257/*
2207 * For now just follow the DIO way to estimate the max credits 2258 * This is called via ext4_da_writepages() to
2208 * needed to write out EXT4_MAX_WRITEBACK_PAGES. 2259 * calulate the total number of credits to reserve to fit
2209 * todo: need to calculate the max credits need for 2260 * a single extent allocation into a single transaction,
2210 * extent based files, currently the DIO credits is based on 2261 * ext4_da_writpeages() will loop calling this before
2211 * indirect-blocks mapping way. 2262 * the block allocation.
2212 *
2213 * Probably should have a generic way to calculate credits
2214 * for DIO, writepages, and truncate
2215 */ 2263 */
2216#define EXT4_MAX_WRITEBACK_PAGES DIO_MAX_BLOCKS 2264
2217#define EXT4_MAX_WRITEBACK_CREDITS DIO_CREDITS 2265static int ext4_da_writepages_trans_blocks(struct inode *inode)
2266{
2267 int max_blocks = EXT4_I(inode)->i_reserved_data_blocks;
2268
2269 /*
2270 * With non-extent format the journal credit needed to
2271 * insert nrblocks contiguous block is dependent on
2272 * number of contiguous block. So we will limit
2273 * number of contiguous block to a sane value
2274 */
2275 if (!(inode->i_flags & EXT4_EXTENTS_FL) &&
2276 (max_blocks > EXT4_MAX_TRANS_DATA))
2277 max_blocks = EXT4_MAX_TRANS_DATA;
2278
2279 return ext4_chunk_trans_blocks(inode, max_blocks);
2280}
2218 2281
2219static int ext4_da_writepages(struct address_space *mapping, 2282static int ext4_da_writepages(struct address_space *mapping,
2220 struct writeback_control *wbc) 2283 struct writeback_control *wbc)
2221{ 2284{
2222 struct inode *inode = mapping->host;
2223 handle_t *handle = NULL; 2285 handle_t *handle = NULL;
2224 int needed_blocks;
2225 int ret = 0;
2226 long to_write;
2227 loff_t range_start = 0; 2286 loff_t range_start = 0;
2287 struct inode *inode = mapping->host;
2288 int needed_blocks, ret = 0, nr_to_writebump = 0;
2289 long to_write, pages_skipped = 0;
2290 struct ext4_sb_info *sbi = EXT4_SB(mapping->host->i_sb);
2228 2291
2229 /* 2292 /*
2230 * No pages to write? This is mainly a kludge to avoid starting 2293 * No pages to write? This is mainly a kludge to avoid starting
2231 * a transaction for special inodes like journal inode on last iput() 2294 * a transaction for special inodes like journal inode on last iput()
2232 * because that could violate lock ordering on umount 2295 * because that could violate lock ordering on umount
2233 */ 2296 */
2234 if (!mapping->nrpages) 2297 if (!mapping->nrpages || !mapping_tagged(mapping, PAGECACHE_TAG_DIRTY))
2235 return 0; 2298 return 0;
2236
2237 /* 2299 /*
2238 * Estimate the worse case needed credits to write out 2300 * Make sure nr_to_write is >= sbi->s_mb_stream_request
2239 * EXT4_MAX_BUF_BLOCKS pages 2301 * This make sure small files blocks are allocated in
2302 * single attempt. This ensure that small files
2303 * get less fragmented.
2240 */ 2304 */
2241 needed_blocks = EXT4_MAX_WRITEBACK_CREDITS; 2305 if (wbc->nr_to_write < sbi->s_mb_stream_request) {
2306 nr_to_writebump = sbi->s_mb_stream_request - wbc->nr_to_write;
2307 wbc->nr_to_write = sbi->s_mb_stream_request;
2308 }
2242 2309
2243 to_write = wbc->nr_to_write; 2310 if (!wbc->range_cyclic)
2244 if (!wbc->range_cyclic) {
2245 /* 2311 /*
2246 * If range_cyclic is not set force range_cont 2312 * If range_cyclic is not set force range_cont
2247 * and save the old writeback_index 2313 * and save the old writeback_index
2248 */ 2314 */
2249 wbc->range_cont = 1; 2315 wbc->range_cont = 1;
2250 range_start = wbc->range_start;
2251 }
2252 2316
2253 while (!ret && to_write) { 2317 range_start = wbc->range_start;
2318 pages_skipped = wbc->pages_skipped;
2319
2320restart_loop:
2321 to_write = wbc->nr_to_write;
2322 while (!ret && to_write > 0) {
2323
2324 /*
2325 * we insert one extent at a time. So we need
2326 * credit needed for single extent allocation.
2327 * journalled mode is currently not supported
2328 * by delalloc
2329 */
2330 BUG_ON(ext4_should_journal_data(inode));
2331 needed_blocks = ext4_da_writepages_trans_blocks(inode);
2332
2254 /* start a new transaction*/ 2333 /* start a new transaction*/
2255 handle = ext4_journal_start(inode, needed_blocks); 2334 handle = ext4_journal_start(inode, needed_blocks);
2256 if (IS_ERR(handle)) { 2335 if (IS_ERR(handle)) {
2257 ret = PTR_ERR(handle); 2336 ret = PTR_ERR(handle);
2337 printk(KERN_EMERG "%s: jbd2_start: "
2338 "%ld pages, ino %lu; err %d\n", __func__,
2339 wbc->nr_to_write, inode->i_ino, ret);
2340 dump_stack();
2258 goto out_writepages; 2341 goto out_writepages;
2259 } 2342 }
2260 if (ext4_should_order_data(inode)) { 2343 if (ext4_should_order_data(inode)) {
2261 /* 2344 /*
2262 * With ordered mode we need to add 2345 * With ordered mode we need to add
2263 * the inode to the journal handle 2346 * the inode to the journal handl
2264 * when we do block allocation. 2347 * when we do block allocation.
2265 */ 2348 */
2266 ret = ext4_jbd2_file_inode(handle, inode); 2349 ret = ext4_jbd2_file_inode(handle, inode);
@@ -2268,20 +2351,20 @@ static int ext4_da_writepages(struct address_space *mapping,
2268 ext4_journal_stop(handle); 2351 ext4_journal_stop(handle);
2269 goto out_writepages; 2352 goto out_writepages;
2270 } 2353 }
2271
2272 } 2354 }
2273 /*
2274 * set the max dirty pages could be write at a time
2275 * to fit into the reserved transaction credits
2276 */
2277 if (wbc->nr_to_write > EXT4_MAX_WRITEBACK_PAGES)
2278 wbc->nr_to_write = EXT4_MAX_WRITEBACK_PAGES;
2279 2355
2280 to_write -= wbc->nr_to_write; 2356 to_write -= wbc->nr_to_write;
2281 ret = mpage_da_writepages(mapping, wbc, 2357 ret = mpage_da_writepages(mapping, wbc,
2282 ext4_da_get_block_write); 2358 ext4_da_get_block_write);
2283 ext4_journal_stop(handle); 2359 ext4_journal_stop(handle);
2284 if (wbc->nr_to_write) { 2360 if (ret == MPAGE_DA_EXTENT_TAIL) {
2361 /*
2362 * got one extent now try with
2363 * rest of the pages
2364 */
2365 to_write += wbc->nr_to_write;
2366 ret = 0;
2367 } else if (wbc->nr_to_write) {
2285 /* 2368 /*
2286 * There is no more writeout needed 2369 * There is no more writeout needed
2287 * or we requested for a noblocking writeout 2370 * or we requested for a noblocking writeout
@@ -2293,10 +2376,18 @@ static int ext4_da_writepages(struct address_space *mapping,
2293 wbc->nr_to_write = to_write; 2376 wbc->nr_to_write = to_write;
2294 } 2377 }
2295 2378
2296out_writepages: 2379 if (wbc->range_cont && (pages_skipped != wbc->pages_skipped)) {
2297 wbc->nr_to_write = to_write; 2380 /* We skipped pages in this loop */
2298 if (range_start)
2299 wbc->range_start = range_start; 2381 wbc->range_start = range_start;
2382 wbc->nr_to_write = to_write +
2383 wbc->pages_skipped - pages_skipped;
2384 wbc->pages_skipped = pages_skipped;
2385 goto restart_loop;
2386 }
2387
2388out_writepages:
2389 wbc->nr_to_write = to_write - nr_to_writebump;
2390 wbc->range_start = range_start;
2300 return ret; 2391 return ret;
2301} 2392}
2302 2393
@@ -3486,6 +3577,9 @@ void ext4_truncate(struct inode *inode)
3486 * modify the block allocation tree. 3577 * modify the block allocation tree.
3487 */ 3578 */
3488 down_write(&ei->i_data_sem); 3579 down_write(&ei->i_data_sem);
3580
3581 ext4_discard_reservation(inode);
3582
3489 /* 3583 /*
3490 * The orphan list entry will now protect us from any crash which 3584 * The orphan list entry will now protect us from any crash which
3491 * occurs before the truncate completes, so it is now safe to propagate 3585 * occurs before the truncate completes, so it is now safe to propagate
@@ -3555,8 +3649,6 @@ do_indirects:
3555 ; 3649 ;
3556 } 3650 }
3557 3651
3558 ext4_discard_reservation(inode);
3559
3560 up_write(&ei->i_data_sem); 3652 up_write(&ei->i_data_sem);
3561 inode->i_mtime = inode->i_ctime = ext4_current_time(inode); 3653 inode->i_mtime = inode->i_ctime = ext4_current_time(inode);
3562 ext4_mark_inode_dirty(handle, inode); 3654 ext4_mark_inode_dirty(handle, inode);
@@ -4324,57 +4416,129 @@ int ext4_getattr(struct vfsmount *mnt, struct dentry *dentry,
4324 return 0; 4416 return 0;
4325} 4417}
4326 4418
4419static int ext4_indirect_trans_blocks(struct inode *inode, int nrblocks,
4420 int chunk)
4421{
4422 int indirects;
4423
4424 /* if nrblocks are contiguous */
4425 if (chunk) {
4426 /*
4427 * With N contiguous data blocks, it need at most
4428 * N/EXT4_ADDR_PER_BLOCK(inode->i_sb) indirect blocks
4429 * 2 dindirect blocks
4430 * 1 tindirect block
4431 */
4432 indirects = nrblocks / EXT4_ADDR_PER_BLOCK(inode->i_sb);
4433 return indirects + 3;
4434 }
4435 /*
4436 * if nrblocks are not contiguous, worse case, each block touch
4437 * a indirect block, and each indirect block touch a double indirect
4438 * block, plus a triple indirect block
4439 */
4440 indirects = nrblocks * 2 + 1;
4441 return indirects;
4442}
4443
4444static int ext4_index_trans_blocks(struct inode *inode, int nrblocks, int chunk)
4445{
4446 if (!(EXT4_I(inode)->i_flags & EXT4_EXTENTS_FL))
4447 return ext4_indirect_trans_blocks(inode, nrblocks, 0);
4448 return ext4_ext_index_trans_blocks(inode, nrblocks, 0);
4449}
4327/* 4450/*
4328 * How many blocks doth make a writepage()? 4451 * Account for index blocks, block groups bitmaps and block group
4329 * 4452 * descriptor blocks if modify datablocks and index blocks
4330 * With N blocks per page, it may be: 4453 * worse case, the indexs blocks spread over different block groups
4331 * N data blocks
4332 * 2 indirect block
4333 * 2 dindirect
4334 * 1 tindirect
4335 * N+5 bitmap blocks (from the above)
4336 * N+5 group descriptor summary blocks
4337 * 1 inode block
4338 * 1 superblock.
4339 * 2 * EXT4_SINGLEDATA_TRANS_BLOCKS for the quote files
4340 * 4454 *
4341 * 3 * (N + 5) + 2 + 2 * EXT4_SINGLEDATA_TRANS_BLOCKS 4455 * If datablocks are discontiguous, they are possible to spread over
4456 * different block groups too. If they are contiugous, with flexbg,
4457 * they could still across block group boundary.
4342 * 4458 *
4343 * With ordered or writeback data it's the same, less the N data blocks. 4459 * Also account for superblock, inode, quota and xattr blocks
4460 */
4461int ext4_meta_trans_blocks(struct inode *inode, int nrblocks, int chunk)
4462{
4463 int groups, gdpblocks;
4464 int idxblocks;
4465 int ret = 0;
4466
4467 /*
4468 * How many index blocks need to touch to modify nrblocks?
4469 * The "Chunk" flag indicating whether the nrblocks is
4470 * physically contiguous on disk
4471 *
4472 * For Direct IO and fallocate, they calls get_block to allocate
4473 * one single extent at a time, so they could set the "Chunk" flag
4474 */
4475 idxblocks = ext4_index_trans_blocks(inode, nrblocks, chunk);
4476
4477 ret = idxblocks;
4478
4479 /*
4480 * Now let's see how many group bitmaps and group descriptors need
4481 * to account
4482 */
4483 groups = idxblocks;
4484 if (chunk)
4485 groups += 1;
4486 else
4487 groups += nrblocks;
4488
4489 gdpblocks = groups;
4490 if (groups > EXT4_SB(inode->i_sb)->s_groups_count)
4491 groups = EXT4_SB(inode->i_sb)->s_groups_count;
4492 if (groups > EXT4_SB(inode->i_sb)->s_gdb_count)
4493 gdpblocks = EXT4_SB(inode->i_sb)->s_gdb_count;
4494
4495 /* bitmaps and block group descriptor blocks */
4496 ret += groups + gdpblocks;
4497
4498 /* Blocks for super block, inode, quota and xattr blocks */
4499 ret += EXT4_META_TRANS_BLOCKS(inode->i_sb);
4500
4501 return ret;
4502}
4503
4504/*
4505 * Calulate the total number of credits to reserve to fit
4506 * the modification of a single pages into a single transaction,
4507 * which may include multiple chunks of block allocations.
4344 * 4508 *
4345 * If the inode's direct blocks can hold an integral number of pages then a 4509 * This could be called via ext4_write_begin()
4346 * page cannot straddle two indirect blocks, and we can only touch one indirect
4347 * and dindirect block, and the "5" above becomes "3".
4348 * 4510 *
4349 * This still overestimates under most circumstances. If we were to pass the 4511 * We need to consider the worse case, when
4350 * start and end offsets in here as well we could do block_to_path() on each 4512 * one new block per extent.
4351 * block and work out the exact number of indirects which are touched. Pah.
4352 */ 4513 */
4353
4354int ext4_writepage_trans_blocks(struct inode *inode) 4514int ext4_writepage_trans_blocks(struct inode *inode)
4355{ 4515{
4356 int bpp = ext4_journal_blocks_per_page(inode); 4516 int bpp = ext4_journal_blocks_per_page(inode);
4357 int indirects = (EXT4_NDIR_BLOCKS % bpp) ? 5 : 3;
4358 int ret; 4517 int ret;
4359 4518
4360 if (EXT4_I(inode)->i_flags & EXT4_EXTENTS_FL) 4519 ret = ext4_meta_trans_blocks(inode, bpp, 0);
4361 return ext4_ext_writepage_trans_blocks(inode, bpp);
4362 4520
4521 /* Account for data blocks for journalled mode */
4363 if (ext4_should_journal_data(inode)) 4522 if (ext4_should_journal_data(inode))
4364 ret = 3 * (bpp + indirects) + 2; 4523 ret += bpp;
4365 else
4366 ret = 2 * (bpp + indirects) + 2;
4367
4368#ifdef CONFIG_QUOTA
4369 /* We know that structure was already allocated during DQUOT_INIT so
4370 * we will be updating only the data blocks + inodes */
4371 ret += 2*EXT4_QUOTA_TRANS_BLOCKS(inode->i_sb);
4372#endif
4373
4374 return ret; 4524 return ret;
4375} 4525}
4376 4526
4377/* 4527/*
4528 * Calculate the journal credits for a chunk of data modification.
4529 *
4530 * This is called from DIO, fallocate or whoever calling
4531 * ext4_get_blocks_wrap() to map/allocate a chunk of contigous disk blocks.
4532 *
4533 * journal buffers for data blocks are not included here, as DIO
4534 * and fallocate do no need to journal data buffers.
4535 */
4536int ext4_chunk_trans_blocks(struct inode *inode, int nrblocks)
4537{
4538 return ext4_meta_trans_blocks(inode, nrblocks, 1);
4539}
4540
4541/*
4378 * The caller must have previously called ext4_reserve_inode_write(). 4542 * The caller must have previously called ext4_reserve_inode_write().
4379 * Give this, we know that the caller already has write access to iloc->bh. 4543 * Give this, we know that the caller already has write access to iloc->bh.
4380 */ 4544 */
diff --git a/fs/ext4/mballoc.c b/fs/ext4/mballoc.c
index 865e9ddb44d4..e0e3a5eb1ddb 100644
--- a/fs/ext4/mballoc.c
+++ b/fs/ext4/mballoc.c
@@ -3282,6 +3282,35 @@ static void ext4_mb_use_group_pa(struct ext4_allocation_context *ac,
3282} 3282}
3283 3283
3284/* 3284/*
3285 * Return the prealloc space that have minimal distance
3286 * from the goal block. @cpa is the prealloc
3287 * space that is having currently known minimal distance
3288 * from the goal block.
3289 */
3290static struct ext4_prealloc_space *
3291ext4_mb_check_group_pa(ext4_fsblk_t goal_block,
3292 struct ext4_prealloc_space *pa,
3293 struct ext4_prealloc_space *cpa)
3294{
3295 ext4_fsblk_t cur_distance, new_distance;
3296
3297 if (cpa == NULL) {
3298 atomic_inc(&pa->pa_count);
3299 return pa;
3300 }
3301 cur_distance = abs(goal_block - cpa->pa_pstart);
3302 new_distance = abs(goal_block - pa->pa_pstart);
3303
3304 if (cur_distance < new_distance)
3305 return cpa;
3306
3307 /* drop the previous reference */
3308 atomic_dec(&cpa->pa_count);
3309 atomic_inc(&pa->pa_count);
3310 return pa;
3311}
3312
3313/*
3285 * search goal blocks in preallocated space 3314 * search goal blocks in preallocated space
3286 */ 3315 */
3287static noinline_for_stack int 3316static noinline_for_stack int
@@ -3290,7 +3319,8 @@ ext4_mb_use_preallocated(struct ext4_allocation_context *ac)
3290 int order, i; 3319 int order, i;
3291 struct ext4_inode_info *ei = EXT4_I(ac->ac_inode); 3320 struct ext4_inode_info *ei = EXT4_I(ac->ac_inode);
3292 struct ext4_locality_group *lg; 3321 struct ext4_locality_group *lg;
3293 struct ext4_prealloc_space *pa; 3322 struct ext4_prealloc_space *pa, *cpa = NULL;
3323 ext4_fsblk_t goal_block;
3294 3324
3295 /* only data can be preallocated */ 3325 /* only data can be preallocated */
3296 if (!(ac->ac_flags & EXT4_MB_HINT_DATA)) 3326 if (!(ac->ac_flags & EXT4_MB_HINT_DATA))
@@ -3333,6 +3363,13 @@ ext4_mb_use_preallocated(struct ext4_allocation_context *ac)
3333 /* The max size of hash table is PREALLOC_TB_SIZE */ 3363 /* The max size of hash table is PREALLOC_TB_SIZE */
3334 order = PREALLOC_TB_SIZE - 1; 3364 order = PREALLOC_TB_SIZE - 1;
3335 3365
3366 goal_block = ac->ac_g_ex.fe_group * EXT4_BLOCKS_PER_GROUP(ac->ac_sb) +
3367 ac->ac_g_ex.fe_start +
3368 le32_to_cpu(EXT4_SB(ac->ac_sb)->s_es->s_first_data_block);
3369 /*
3370 * search for the prealloc space that is having
3371 * minimal distance from the goal block.
3372 */
3336 for (i = order; i < PREALLOC_TB_SIZE; i++) { 3373 for (i = order; i < PREALLOC_TB_SIZE; i++) {
3337 rcu_read_lock(); 3374 rcu_read_lock();
3338 list_for_each_entry_rcu(pa, &lg->lg_prealloc_list[i], 3375 list_for_each_entry_rcu(pa, &lg->lg_prealloc_list[i],
@@ -3340,17 +3377,19 @@ ext4_mb_use_preallocated(struct ext4_allocation_context *ac)
3340 spin_lock(&pa->pa_lock); 3377 spin_lock(&pa->pa_lock);
3341 if (pa->pa_deleted == 0 && 3378 if (pa->pa_deleted == 0 &&
3342 pa->pa_free >= ac->ac_o_ex.fe_len) { 3379 pa->pa_free >= ac->ac_o_ex.fe_len) {
3343 atomic_inc(&pa->pa_count); 3380
3344 ext4_mb_use_group_pa(ac, pa); 3381 cpa = ext4_mb_check_group_pa(goal_block,
3345 spin_unlock(&pa->pa_lock); 3382 pa, cpa);
3346 ac->ac_criteria = 20;
3347 rcu_read_unlock();
3348 return 1;
3349 } 3383 }
3350 spin_unlock(&pa->pa_lock); 3384 spin_unlock(&pa->pa_lock);
3351 } 3385 }
3352 rcu_read_unlock(); 3386 rcu_read_unlock();
3353 } 3387 }
3388 if (cpa) {
3389 ext4_mb_use_group_pa(ac, cpa);
3390 ac->ac_criteria = 20;
3391 return 1;
3392 }
3354 return 0; 3393 return 0;
3355} 3394}
3356 3395
diff --git a/fs/ext4/migrate.c b/fs/ext4/migrate.c
index b9e077ba07e9..46fc0b5b12ba 100644
--- a/fs/ext4/migrate.c
+++ b/fs/ext4/migrate.c
@@ -53,7 +53,8 @@ static int finish_range(handle_t *handle, struct inode *inode,
53 * credit. But below we try to not accumalate too much 53 * credit. But below we try to not accumalate too much
54 * of them by restarting the journal. 54 * of them by restarting the journal.
55 */ 55 */
56 needed = ext4_ext_calc_credits_for_insert(inode, path); 56 needed = ext4_ext_calc_credits_for_single_extent(inode,
57 lb->last_block - lb->first_block + 1, path);
57 58
58 /* 59 /*
59 * Make sure the credit we accumalated is not really high 60 * Make sure the credit we accumalated is not really high
diff --git a/fs/ext4/resize.c b/fs/ext4/resize.c
index 0a9265164265..b3d35604ea18 100644
--- a/fs/ext4/resize.c
+++ b/fs/ext4/resize.c
@@ -773,7 +773,8 @@ int ext4_group_add(struct super_block *sb, struct ext4_new_group_data *input)
773 773
774 if (reserved_gdb || gdb_off == 0) { 774 if (reserved_gdb || gdb_off == 0) {
775 if (!EXT4_HAS_COMPAT_FEATURE(sb, 775 if (!EXT4_HAS_COMPAT_FEATURE(sb,
776 EXT4_FEATURE_COMPAT_RESIZE_INODE)){ 776 EXT4_FEATURE_COMPAT_RESIZE_INODE)
777 || !le16_to_cpu(es->s_reserved_gdt_blocks)) {
777 ext4_warning(sb, __func__, 778 ext4_warning(sb, __func__,
778 "No reserved GDT blocks, can't resize"); 779 "No reserved GDT blocks, can't resize");
779 return -EPERM; 780 return -EPERM;
diff --git a/fs/ext4/super.c b/fs/ext4/super.c
index d5d77958b861..566344b926b7 100644
--- a/fs/ext4/super.c
+++ b/fs/ext4/super.c
@@ -568,6 +568,7 @@ static struct inode *ext4_alloc_inode(struct super_block *sb)
568#endif 568#endif
569 ei->i_block_alloc_info = NULL; 569 ei->i_block_alloc_info = NULL;
570 ei->vfs_inode.i_version = 1; 570 ei->vfs_inode.i_version = 1;
571 ei->vfs_inode.i_data.writeback_index = 0;
571 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache)); 572 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
572 INIT_LIST_HEAD(&ei->i_prealloc_list); 573 INIT_LIST_HEAD(&ei->i_prealloc_list);
573 spin_lock_init(&ei->i_prealloc_lock); 574 spin_lock_init(&ei->i_prealloc_lock);
diff --git a/fs/jffs2/jffs2_fs_i.h b/fs/jffs2/jffs2_fs_i.h
index 31559f45fdde..4c41db91eaa4 100644
--- a/fs/jffs2/jffs2_fs_i.h
+++ b/fs/jffs2/jffs2_fs_i.h
@@ -12,7 +12,6 @@
12#ifndef _JFFS2_FS_I 12#ifndef _JFFS2_FS_I
13#define _JFFS2_FS_I 13#define _JFFS2_FS_I
14 14
15#include <linux/version.h>
16#include <linux/rbtree.h> 15#include <linux/rbtree.h>
17#include <linux/posix_acl.h> 16#include <linux/posix_acl.h>
18#include <linux/mutex.h> 17#include <linux/mutex.h>
diff --git a/fs/nfs/super.c b/fs/nfs/super.c
index 9abcd2b329f7..e9b20173fef3 100644
--- a/fs/nfs/super.c
+++ b/fs/nfs/super.c
@@ -1279,6 +1279,12 @@ static int nfs_parse_mount_options(char *raw,
1279 } 1279 }
1280 } 1280 }
1281 1281
1282 if (errors > 0) {
1283 dfprintk(MOUNT, "NFS: parsing encountered %d error%s\n",
1284 errors, (errors == 1 ? "" : "s"));
1285 if (!sloppy)
1286 return 0;
1287 }
1282 return 1; 1288 return 1;
1283 1289
1284out_nomem: 1290out_nomem:
diff --git a/fs/nfsd/nfs4acl.c b/fs/nfsd/nfs4acl.c
index b6ed38380ab8..54b8b4140c8f 100644
--- a/fs/nfsd/nfs4acl.c
+++ b/fs/nfsd/nfs4acl.c
@@ -443,7 +443,7 @@ init_state(struct posix_acl_state *state, int cnt)
443 * enough space for either: 443 * enough space for either:
444 */ 444 */
445 alloc = sizeof(struct posix_ace_state_array) 445 alloc = sizeof(struct posix_ace_state_array)
446 + cnt*sizeof(struct posix_ace_state); 446 + cnt*sizeof(struct posix_user_ace_state);
447 state->users = kzalloc(alloc, GFP_KERNEL); 447 state->users = kzalloc(alloc, GFP_KERNEL);
448 if (!state->users) 448 if (!state->users)
449 return -ENOMEM; 449 return -ENOMEM;
diff --git a/fs/nfsd/nfs4proc.c b/fs/nfsd/nfs4proc.c
index 2e51adac65de..e5b51ffafc6c 100644
--- a/fs/nfsd/nfs4proc.c
+++ b/fs/nfsd/nfs4proc.c
@@ -867,11 +867,6 @@ nfsd4_proc_compound(struct svc_rqst *rqstp,
867 int slack_bytes; 867 int slack_bytes;
868 __be32 status; 868 __be32 status;
869 869
870 status = nfserr_resource;
871 cstate = cstate_alloc();
872 if (cstate == NULL)
873 goto out;
874
875 resp->xbuf = &rqstp->rq_res; 870 resp->xbuf = &rqstp->rq_res;
876 resp->p = rqstp->rq_res.head[0].iov_base + rqstp->rq_res.head[0].iov_len; 871 resp->p = rqstp->rq_res.head[0].iov_base + rqstp->rq_res.head[0].iov_len;
877 resp->tagp = resp->p; 872 resp->tagp = resp->p;
@@ -890,6 +885,11 @@ nfsd4_proc_compound(struct svc_rqst *rqstp,
890 if (args->minorversion > NFSD_SUPPORTED_MINOR_VERSION) 885 if (args->minorversion > NFSD_SUPPORTED_MINOR_VERSION)
891 goto out; 886 goto out;
892 887
888 status = nfserr_resource;
889 cstate = cstate_alloc();
890 if (cstate == NULL)
891 goto out;
892
893 status = nfs_ok; 893 status = nfs_ok;
894 while (!status && resp->opcnt < args->opcnt) { 894 while (!status && resp->opcnt < args->opcnt) {
895 op = &args->ops[resp->opcnt++]; 895 op = &args->ops[resp->opcnt++];
@@ -957,9 +957,9 @@ encode_op:
957 nfsd4_increment_op_stats(op->opnum); 957 nfsd4_increment_op_stats(op->opnum);
958 } 958 }
959 959
960 cstate_free(cstate);
960out: 961out:
961 nfsd4_release_compoundargs(args); 962 nfsd4_release_compoundargs(args);
962 cstate_free(cstate);
963 dprintk("nfsv4 compound returned %d\n", ntohl(status)); 963 dprintk("nfsv4 compound returned %d\n", ntohl(status));
964 return status; 964 return status;
965} 965}
diff --git a/fs/ntfs/namei.c b/fs/ntfs/namei.c
index e1781c8b1650..9e8a95be7a1e 100644
--- a/fs/ntfs/namei.c
+++ b/fs/ntfs/namei.c
@@ -174,7 +174,6 @@ static struct dentry *ntfs_lookup(struct inode *dir_ino, struct dentry *dent,
174 // TODO: Consider moving this lot to a separate function! (AIA) 174 // TODO: Consider moving this lot to a separate function! (AIA)
175handle_name: 175handle_name:
176 { 176 {
177 struct dentry *real_dent, *new_dent;
178 MFT_RECORD *m; 177 MFT_RECORD *m;
179 ntfs_attr_search_ctx *ctx; 178 ntfs_attr_search_ctx *ctx;
180 ntfs_inode *ni = NTFS_I(dent_inode); 179 ntfs_inode *ni = NTFS_I(dent_inode);
@@ -255,93 +254,9 @@ handle_name:
255 } 254 }
256 nls_name.hash = full_name_hash(nls_name.name, nls_name.len); 255 nls_name.hash = full_name_hash(nls_name.name, nls_name.len);
257 256
258 /* 257 dent = d_add_ci(dent, dent_inode, &nls_name);
259 * Note: No need for dent->d_lock lock as i_mutex is held on the
260 * parent inode.
261 */
262
263 /* Does a dentry matching the nls_name exist already? */
264 real_dent = d_lookup(dent->d_parent, &nls_name);
265 /* If not, create it now. */
266 if (!real_dent) {
267 real_dent = d_alloc(dent->d_parent, &nls_name);
268 kfree(nls_name.name);
269 if (!real_dent) {
270 err = -ENOMEM;
271 goto err_out;
272 }
273 new_dent = d_splice_alias(dent_inode, real_dent);
274 if (new_dent)
275 dput(real_dent);
276 else
277 new_dent = real_dent;
278 ntfs_debug("Done. (Created new dentry.)");
279 return new_dent;
280 }
281 kfree(nls_name.name); 258 kfree(nls_name.name);
282 /* Matching dentry exists, check if it is negative. */ 259 return dent;
283 if (real_dent->d_inode) {
284 if (unlikely(real_dent->d_inode != dent_inode)) {
285 /* This can happen because bad inodes are unhashed. */
286 BUG_ON(!is_bad_inode(dent_inode));
287 BUG_ON(!is_bad_inode(real_dent->d_inode));
288 }
289 /*
290 * Already have the inode and the dentry attached, decrement
291 * the reference count to balance the ntfs_iget() we did
292 * earlier on. We found the dentry using d_lookup() so it
293 * cannot be disconnected and thus we do not need to worry
294 * about any NFS/disconnectedness issues here.
295 */
296 iput(dent_inode);
297 ntfs_debug("Done. (Already had inode and dentry.)");
298 return real_dent;
299 }
300 /*
301 * Negative dentry: instantiate it unless the inode is a directory and
302 * has a 'disconnected' dentry (i.e. IS_ROOT and DCACHE_DISCONNECTED),
303 * in which case d_move() that in place of the found dentry.
304 */
305 if (!S_ISDIR(dent_inode->i_mode)) {
306 /* Not a directory; everything is easy. */
307 d_instantiate(real_dent, dent_inode);
308 ntfs_debug("Done. (Already had negative file dentry.)");
309 return real_dent;
310 }
311 spin_lock(&dcache_lock);
312 if (list_empty(&dent_inode->i_dentry)) {
313 /*
314 * Directory without a 'disconnected' dentry; we need to do
315 * d_instantiate() by hand because it takes dcache_lock which
316 * we already hold.
317 */
318 list_add(&real_dent->d_alias, &dent_inode->i_dentry);
319 real_dent->d_inode = dent_inode;
320 spin_unlock(&dcache_lock);
321 security_d_instantiate(real_dent, dent_inode);
322 ntfs_debug("Done. (Already had negative directory dentry.)");
323 return real_dent;
324 }
325 /*
326 * Directory with a 'disconnected' dentry; get a reference to the
327 * 'disconnected' dentry.
328 */
329 new_dent = list_entry(dent_inode->i_dentry.next, struct dentry,
330 d_alias);
331 dget_locked(new_dent);
332 spin_unlock(&dcache_lock);
333 /* Do security vodoo. */
334 security_d_instantiate(real_dent, dent_inode);
335 /* Move new_dent in place of real_dent. */
336 d_move(new_dent, real_dent);
337 /* Balance the ntfs_iget() we did above. */
338 iput(dent_inode);
339 /* Throw away real_dent. */
340 dput(real_dent);
341 /* Use new_dent as the actual dentry. */
342 ntfs_debug("Done. (Already had negative, disconnected directory "
343 "dentry.)");
344 return new_dent;
345 260
346eio_err_out: 261eio_err_out:
347 ntfs_error(vol->sb, "Illegal file name attribute. Run chkdsk."); 262 ntfs_error(vol->sb, "Illegal file name attribute. Run chkdsk.");
diff --git a/fs/ntfs/usnjrnl.h b/fs/ntfs/usnjrnl.h
index 3a8af75351e8..4087fbdac327 100644
--- a/fs/ntfs/usnjrnl.h
+++ b/fs/ntfs/usnjrnl.h
@@ -113,7 +113,7 @@ typedef struct {
113 * Reason flags (32-bit). Cumulative flags describing the change(s) to the 113 * Reason flags (32-bit). Cumulative flags describing the change(s) to the
114 * file since it was last opened. I think the names speak for themselves but 114 * file since it was last opened. I think the names speak for themselves but
115 * if you disagree check out the descriptions in the Linux NTFS project NTFS 115 * if you disagree check out the descriptions in the Linux NTFS project NTFS
116 * documentation: http://linux-ntfs.sourceforge.net/ntfs/files/usnjrnl.html 116 * documentation: http://www.linux-ntfs.org/
117 */ 117 */
118enum { 118enum {
119 USN_REASON_DATA_OVERWRITE = const_cpu_to_le32(0x00000001), 119 USN_REASON_DATA_OVERWRITE = const_cpu_to_le32(0x00000001),
@@ -145,7 +145,7 @@ typedef le32 USN_REASON_FLAGS;
145 * Source info flags (32-bit). Information about the source of the change(s) 145 * Source info flags (32-bit). Information about the source of the change(s)
146 * to the file. For detailed descriptions of what these mean, see the Linux 146 * to the file. For detailed descriptions of what these mean, see the Linux
147 * NTFS project NTFS documentation: 147 * NTFS project NTFS documentation:
148 * http://linux-ntfs.sourceforge.net/ntfs/files/usnjrnl.html 148 * http://www.linux-ntfs.org/
149 */ 149 */
150enum { 150enum {
151 USN_SOURCE_DATA_MANAGEMENT = const_cpu_to_le32(0x00000001), 151 USN_SOURCE_DATA_MANAGEMENT = const_cpu_to_le32(0x00000001),
diff --git a/fs/ocfs2/cluster/netdebug.c b/fs/ocfs2/cluster/netdebug.c
index d8bfa0eb41b2..52276c02f710 100644
--- a/fs/ocfs2/cluster/netdebug.c
+++ b/fs/ocfs2/cluster/netdebug.c
@@ -138,20 +138,20 @@ static int nst_seq_show(struct seq_file *seq, void *v)
138 " message id: %d\n" 138 " message id: %d\n"
139 " message type: %u\n" 139 " message type: %u\n"
140 " message key: 0x%08x\n" 140 " message key: 0x%08x\n"
141 " sock acquiry: %lu.%lu\n" 141 " sock acquiry: %lu.%ld\n"
142 " send start: %lu.%lu\n" 142 " send start: %lu.%ld\n"
143 " wait start: %lu.%lu\n", 143 " wait start: %lu.%ld\n",
144 nst, (unsigned long)nst->st_task->pid, 144 nst, (unsigned long)nst->st_task->pid,
145 (unsigned long)nst->st_task->tgid, 145 (unsigned long)nst->st_task->tgid,
146 nst->st_task->comm, nst->st_node, 146 nst->st_task->comm, nst->st_node,
147 nst->st_sc, nst->st_id, nst->st_msg_type, 147 nst->st_sc, nst->st_id, nst->st_msg_type,
148 nst->st_msg_key, 148 nst->st_msg_key,
149 nst->st_sock_time.tv_sec, 149 nst->st_sock_time.tv_sec,
150 (unsigned long)nst->st_sock_time.tv_usec, 150 (long)nst->st_sock_time.tv_usec,
151 nst->st_send_time.tv_sec, 151 nst->st_send_time.tv_sec,
152 (unsigned long)nst->st_send_time.tv_usec, 152 (long)nst->st_send_time.tv_usec,
153 nst->st_status_time.tv_sec, 153 nst->st_status_time.tv_sec,
154 nst->st_status_time.tv_usec); 154 (long)nst->st_status_time.tv_usec);
155 } 155 }
156 156
157 spin_unlock(&o2net_debug_lock); 157 spin_unlock(&o2net_debug_lock);
@@ -276,7 +276,7 @@ static void *sc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
276 return sc; /* unused, just needs to be null when done */ 276 return sc; /* unused, just needs to be null when done */
277} 277}
278 278
279#define TV_SEC_USEC(TV) TV.tv_sec, (unsigned long)TV.tv_usec 279#define TV_SEC_USEC(TV) TV.tv_sec, (long)TV.tv_usec
280 280
281static int sc_seq_show(struct seq_file *seq, void *v) 281static int sc_seq_show(struct seq_file *seq, void *v)
282{ 282{
@@ -309,12 +309,12 @@ static int sc_seq_show(struct seq_file *seq, void *v)
309 " remote node: %s\n" 309 " remote node: %s\n"
310 " page off: %zu\n" 310 " page off: %zu\n"
311 " handshake ok: %u\n" 311 " handshake ok: %u\n"
312 " timer: %lu.%lu\n" 312 " timer: %lu.%ld\n"
313 " data ready: %lu.%lu\n" 313 " data ready: %lu.%ld\n"
314 " advance start: %lu.%lu\n" 314 " advance start: %lu.%ld\n"
315 " advance stop: %lu.%lu\n" 315 " advance stop: %lu.%ld\n"
316 " func start: %lu.%lu\n" 316 " func start: %lu.%ld\n"
317 " func stop: %lu.%lu\n" 317 " func stop: %lu.%ld\n"
318 " func key: %u\n" 318 " func key: %u\n"
319 " func type: %u\n", 319 " func type: %u\n",
320 sc, 320 sc,
diff --git a/fs/ocfs2/cluster/tcp.c b/fs/ocfs2/cluster/tcp.c
index a27d61581bd6..2bcf706d9dd3 100644
--- a/fs/ocfs2/cluster/tcp.c
+++ b/fs/ocfs2/cluster/tcp.c
@@ -143,8 +143,8 @@ static void o2net_sc_postpone_idle(struct o2net_sock_container *sc);
143static void o2net_sc_reset_idle_timer(struct o2net_sock_container *sc); 143static void o2net_sc_reset_idle_timer(struct o2net_sock_container *sc);
144 144
145#ifdef CONFIG_DEBUG_FS 145#ifdef CONFIG_DEBUG_FS
146void o2net_init_nst(struct o2net_send_tracking *nst, u32 msgtype, 146static void o2net_init_nst(struct o2net_send_tracking *nst, u32 msgtype,
147 u32 msgkey, struct task_struct *task, u8 node) 147 u32 msgkey, struct task_struct *task, u8 node)
148{ 148{
149 INIT_LIST_HEAD(&nst->st_net_debug_item); 149 INIT_LIST_HEAD(&nst->st_net_debug_item);
150 nst->st_task = task; 150 nst->st_task = task;
@@ -153,31 +153,61 @@ void o2net_init_nst(struct o2net_send_tracking *nst, u32 msgtype,
153 nst->st_node = node; 153 nst->st_node = node;
154} 154}
155 155
156void o2net_set_nst_sock_time(struct o2net_send_tracking *nst) 156static void o2net_set_nst_sock_time(struct o2net_send_tracking *nst)
157{ 157{
158 do_gettimeofday(&nst->st_sock_time); 158 do_gettimeofday(&nst->st_sock_time);
159} 159}
160 160
161void o2net_set_nst_send_time(struct o2net_send_tracking *nst) 161static void o2net_set_nst_send_time(struct o2net_send_tracking *nst)
162{ 162{
163 do_gettimeofday(&nst->st_send_time); 163 do_gettimeofday(&nst->st_send_time);
164} 164}
165 165
166void o2net_set_nst_status_time(struct o2net_send_tracking *nst) 166static void o2net_set_nst_status_time(struct o2net_send_tracking *nst)
167{ 167{
168 do_gettimeofday(&nst->st_status_time); 168 do_gettimeofday(&nst->st_status_time);
169} 169}
170 170
171void o2net_set_nst_sock_container(struct o2net_send_tracking *nst, 171static void o2net_set_nst_sock_container(struct o2net_send_tracking *nst,
172 struct o2net_sock_container *sc) 172 struct o2net_sock_container *sc)
173{ 173{
174 nst->st_sc = sc; 174 nst->st_sc = sc;
175} 175}
176 176
177void o2net_set_nst_msg_id(struct o2net_send_tracking *nst, u32 msg_id) 177static void o2net_set_nst_msg_id(struct o2net_send_tracking *nst, u32 msg_id)
178{ 178{
179 nst->st_id = msg_id; 179 nst->st_id = msg_id;
180} 180}
181
182#else /* CONFIG_DEBUG_FS */
183
184static inline void o2net_init_nst(struct o2net_send_tracking *nst, u32 msgtype,
185 u32 msgkey, struct task_struct *task, u8 node)
186{
187}
188
189static inline void o2net_set_nst_sock_time(struct o2net_send_tracking *nst)
190{
191}
192
193static inline void o2net_set_nst_send_time(struct o2net_send_tracking *nst)
194{
195}
196
197static inline void o2net_set_nst_status_time(struct o2net_send_tracking *nst)
198{
199}
200
201static inline void o2net_set_nst_sock_container(struct o2net_send_tracking *nst,
202 struct o2net_sock_container *sc)
203{
204}
205
206static inline void o2net_set_nst_msg_id(struct o2net_send_tracking *nst,
207 u32 msg_id)
208{
209}
210
181#endif /* CONFIG_DEBUG_FS */ 211#endif /* CONFIG_DEBUG_FS */
182 212
183static inline int o2net_reconnect_delay(void) 213static inline int o2net_reconnect_delay(void)
diff --git a/fs/ocfs2/cluster/tcp_internal.h b/fs/ocfs2/cluster/tcp_internal.h
index 18307ff81b77..8d58cfe410b1 100644
--- a/fs/ocfs2/cluster/tcp_internal.h
+++ b/fs/ocfs2/cluster/tcp_internal.h
@@ -224,42 +224,10 @@ struct o2net_send_tracking {
224 struct timeval st_send_time; 224 struct timeval st_send_time;
225 struct timeval st_status_time; 225 struct timeval st_status_time;
226}; 226};
227
228void o2net_init_nst(struct o2net_send_tracking *nst, u32 msgtype,
229 u32 msgkey, struct task_struct *task, u8 node);
230void o2net_set_nst_sock_time(struct o2net_send_tracking *nst);
231void o2net_set_nst_send_time(struct o2net_send_tracking *nst);
232void o2net_set_nst_status_time(struct o2net_send_tracking *nst);
233void o2net_set_nst_sock_container(struct o2net_send_tracking *nst,
234 struct o2net_sock_container *sc);
235void o2net_set_nst_msg_id(struct o2net_send_tracking *nst, u32 msg_id);
236
237#else 227#else
238struct o2net_send_tracking { 228struct o2net_send_tracking {
239 u32 dummy; 229 u32 dummy;
240}; 230};
241
242static inline void o2net_init_nst(struct o2net_send_tracking *nst, u32 msgtype,
243 u32 msgkey, struct task_struct *task, u8 node)
244{
245}
246static inline void o2net_set_nst_sock_time(struct o2net_send_tracking *nst)
247{
248}
249static inline void o2net_set_nst_send_time(struct o2net_send_tracking *nst)
250{
251}
252static inline void o2net_set_nst_status_time(struct o2net_send_tracking *nst)
253{
254}
255static inline void o2net_set_nst_sock_container(struct o2net_send_tracking *nst,
256 struct o2net_sock_container *sc)
257{
258}
259static inline void o2net_set_nst_msg_id(struct o2net_send_tracking *nst,
260 u32 msg_id)
261{
262}
263#endif /* CONFIG_DEBUG_FS */ 231#endif /* CONFIG_DEBUG_FS */
264 232
265#endif /* O2CLUSTER_TCP_INTERNAL_H */ 233#endif /* O2CLUSTER_TCP_INTERNAL_H */
diff --git a/fs/ocfs2/dir.c b/fs/ocfs2/dir.c
index 8a1875848080..9cce563fd627 100644
--- a/fs/ocfs2/dir.c
+++ b/fs/ocfs2/dir.c
@@ -1300,7 +1300,6 @@ static int ocfs2_expand_inline_dir(struct inode *dir, struct buffer_head *di_bh,
1300 di->i_size = cpu_to_le64(sb->s_blocksize); 1300 di->i_size = cpu_to_le64(sb->s_blocksize);
1301 di->i_ctime = di->i_mtime = cpu_to_le64(dir->i_ctime.tv_sec); 1301 di->i_ctime = di->i_mtime = cpu_to_le64(dir->i_ctime.tv_sec);
1302 di->i_ctime_nsec = di->i_mtime_nsec = cpu_to_le32(dir->i_ctime.tv_nsec); 1302 di->i_ctime_nsec = di->i_mtime_nsec = cpu_to_le32(dir->i_ctime.tv_nsec);
1303 dir->i_blocks = ocfs2_inode_sector_count(dir);
1304 1303
1305 /* 1304 /*
1306 * This should never fail as our extent list is empty and all 1305 * This should never fail as our extent list is empty and all
@@ -1310,9 +1309,15 @@ static int ocfs2_expand_inline_dir(struct inode *dir, struct buffer_head *di_bh,
1310 NULL); 1309 NULL);
1311 if (ret) { 1310 if (ret) {
1312 mlog_errno(ret); 1311 mlog_errno(ret);
1313 goto out; 1312 goto out_commit;
1314 } 1313 }
1315 1314
1315 /*
1316 * Set i_blocks after the extent insert for the most up to
1317 * date ip_clusters value.
1318 */
1319 dir->i_blocks = ocfs2_inode_sector_count(dir);
1320
1316 ret = ocfs2_journal_dirty(handle, di_bh); 1321 ret = ocfs2_journal_dirty(handle, di_bh);
1317 if (ret) { 1322 if (ret) {
1318 mlog_errno(ret); 1323 mlog_errno(ret);
@@ -1336,7 +1341,7 @@ static int ocfs2_expand_inline_dir(struct inode *dir, struct buffer_head *di_bh,
1336 len, 0, NULL); 1341 len, 0, NULL);
1337 if (ret) { 1342 if (ret) {
1338 mlog_errno(ret); 1343 mlog_errno(ret);
1339 goto out; 1344 goto out_commit;
1340 } 1345 }
1341 } 1346 }
1342 1347
diff --git a/fs/ocfs2/journal.c b/fs/ocfs2/journal.c
index 7a37240f7a31..c47bc2a809c2 100644
--- a/fs/ocfs2/journal.c
+++ b/fs/ocfs2/journal.c
@@ -1418,13 +1418,13 @@ int ocfs2_mark_dead_nodes(struct ocfs2_super *osb)
1418{ 1418{
1419 unsigned int node_num; 1419 unsigned int node_num;
1420 int status, i; 1420 int status, i;
1421 u32 gen;
1421 struct buffer_head *bh = NULL; 1422 struct buffer_head *bh = NULL;
1422 struct ocfs2_dinode *di; 1423 struct ocfs2_dinode *di;
1423 1424
1424 /* This is called with the super block cluster lock, so we 1425 /* This is called with the super block cluster lock, so we
1425 * know that the slot map can't change underneath us. */ 1426 * know that the slot map can't change underneath us. */
1426 1427
1427 spin_lock(&osb->osb_lock);
1428 for (i = 0; i < osb->max_slots; i++) { 1428 for (i = 0; i < osb->max_slots; i++) {
1429 /* Read journal inode to get the recovery generation */ 1429 /* Read journal inode to get the recovery generation */
1430 status = ocfs2_read_journal_inode(osb, i, &bh, NULL); 1430 status = ocfs2_read_journal_inode(osb, i, &bh, NULL);
@@ -1433,23 +1433,31 @@ int ocfs2_mark_dead_nodes(struct ocfs2_super *osb)
1433 goto bail; 1433 goto bail;
1434 } 1434 }
1435 di = (struct ocfs2_dinode *)bh->b_data; 1435 di = (struct ocfs2_dinode *)bh->b_data;
1436 osb->slot_recovery_generations[i] = 1436 gen = ocfs2_get_recovery_generation(di);
1437 ocfs2_get_recovery_generation(di);
1438 brelse(bh); 1437 brelse(bh);
1439 bh = NULL; 1438 bh = NULL;
1440 1439
1440 spin_lock(&osb->osb_lock);
1441 osb->slot_recovery_generations[i] = gen;
1442
1441 mlog(0, "Slot %u recovery generation is %u\n", i, 1443 mlog(0, "Slot %u recovery generation is %u\n", i,
1442 osb->slot_recovery_generations[i]); 1444 osb->slot_recovery_generations[i]);
1443 1445
1444 if (i == osb->slot_num) 1446 if (i == osb->slot_num) {
1447 spin_unlock(&osb->osb_lock);
1445 continue; 1448 continue;
1449 }
1446 1450
1447 status = ocfs2_slot_to_node_num_locked(osb, i, &node_num); 1451 status = ocfs2_slot_to_node_num_locked(osb, i, &node_num);
1448 if (status == -ENOENT) 1452 if (status == -ENOENT) {
1453 spin_unlock(&osb->osb_lock);
1449 continue; 1454 continue;
1455 }
1450 1456
1451 if (__ocfs2_recovery_map_test(osb, node_num)) 1457 if (__ocfs2_recovery_map_test(osb, node_num)) {
1458 spin_unlock(&osb->osb_lock);
1452 continue; 1459 continue;
1460 }
1453 spin_unlock(&osb->osb_lock); 1461 spin_unlock(&osb->osb_lock);
1454 1462
1455 /* Ok, we have a slot occupied by another node which 1463 /* Ok, we have a slot occupied by another node which
@@ -1465,10 +1473,7 @@ int ocfs2_mark_dead_nodes(struct ocfs2_super *osb)
1465 mlog_errno(status); 1473 mlog_errno(status);
1466 goto bail; 1474 goto bail;
1467 } 1475 }
1468
1469 spin_lock(&osb->osb_lock);
1470 } 1476 }
1471 spin_unlock(&osb->osb_lock);
1472 1477
1473 status = 0; 1478 status = 0;
1474bail: 1479bail:
diff --git a/fs/ocfs2/stackglue.c b/fs/ocfs2/stackglue.c
index 10e149ae5e3a..07f348b8d721 100644
--- a/fs/ocfs2/stackglue.c
+++ b/fs/ocfs2/stackglue.c
@@ -97,13 +97,14 @@ static int ocfs2_stack_driver_request(const char *stack_name,
97 goto out; 97 goto out;
98 } 98 }
99 99
100 /* Ok, the stack is pinned */
101 p->sp_count++;
102 active_stack = p; 100 active_stack = p;
103
104 rc = 0; 101 rc = 0;
105 102
106out: 103out:
104 /* If we found it, pin it */
105 if (!rc)
106 active_stack->sp_count++;
107
107 spin_unlock(&ocfs2_stack_lock); 108 spin_unlock(&ocfs2_stack_lock);
108 return rc; 109 return rc;
109} 110}
diff --git a/fs/proc/array.c b/fs/proc/array.c
index 0d6eb33597c6..71c9be59c9c2 100644
--- a/fs/proc/array.c
+++ b/fs/proc/array.c
@@ -337,65 +337,6 @@ int proc_pid_status(struct seq_file *m, struct pid_namespace *ns,
337 return 0; 337 return 0;
338} 338}
339 339
340/*
341 * Use precise platform statistics if available:
342 */
343#ifdef CONFIG_VIRT_CPU_ACCOUNTING
344static cputime_t task_utime(struct task_struct *p)
345{
346 return p->utime;
347}
348
349static cputime_t task_stime(struct task_struct *p)
350{
351 return p->stime;
352}
353#else
354static cputime_t task_utime(struct task_struct *p)
355{
356 clock_t utime = cputime_to_clock_t(p->utime),
357 total = utime + cputime_to_clock_t(p->stime);
358 u64 temp;
359
360 /*
361 * Use CFS's precise accounting:
362 */
363 temp = (u64)nsec_to_clock_t(p->se.sum_exec_runtime);
364
365 if (total) {
366 temp *= utime;
367 do_div(temp, total);
368 }
369 utime = (clock_t)temp;
370
371 p->prev_utime = max(p->prev_utime, clock_t_to_cputime(utime));
372 return p->prev_utime;
373}
374
375static cputime_t task_stime(struct task_struct *p)
376{
377 clock_t stime;
378
379 /*
380 * Use CFS's precise accounting. (we subtract utime from
381 * the total, to make sure the total observed by userspace
382 * grows monotonically - apps rely on that):
383 */
384 stime = nsec_to_clock_t(p->se.sum_exec_runtime) -
385 cputime_to_clock_t(task_utime(p));
386
387 if (stime >= 0)
388 p->prev_stime = max(p->prev_stime, clock_t_to_cputime(stime));
389
390 return p->prev_stime;
391}
392#endif
393
394static cputime_t task_gtime(struct task_struct *p)
395{
396 return p->gtime;
397}
398
399static int do_task_stat(struct seq_file *m, struct pid_namespace *ns, 340static int do_task_stat(struct seq_file *m, struct pid_namespace *ns,
400 struct pid *pid, struct task_struct *task, int whole) 341 struct pid *pid, struct task_struct *task, int whole)
401{ 342{
diff --git a/fs/proc/generic.c b/fs/proc/generic.c
index 4fb81e9c94e3..bca0f81eb687 100644
--- a/fs/proc/generic.c
+++ b/fs/proc/generic.c
@@ -330,6 +330,7 @@ retry:
330 spin_lock(&proc_inum_lock); 330 spin_lock(&proc_inum_lock);
331 ida_remove(&proc_inum_ida, i); 331 ida_remove(&proc_inum_ida, i);
332 spin_unlock(&proc_inum_lock); 332 spin_unlock(&proc_inum_lock);
333 return 0;
333 } 334 }
334 return PROC_DYNAMIC_FIRST + i; 335 return PROC_DYNAMIC_FIRST + i;
335} 336}
diff --git a/fs/proc/proc_misc.c b/fs/proc/proc_misc.c
index ded969862960..00f10a2dcf12 100644
--- a/fs/proc/proc_misc.c
+++ b/fs/proc/proc_misc.c
@@ -24,6 +24,7 @@
24#include <linux/tty.h> 24#include <linux/tty.h>
25#include <linux/string.h> 25#include <linux/string.h>
26#include <linux/mman.h> 26#include <linux/mman.h>
27#include <linux/quicklist.h>
27#include <linux/proc_fs.h> 28#include <linux/proc_fs.h>
28#include <linux/ioport.h> 29#include <linux/ioport.h>
29#include <linux/mm.h> 30#include <linux/mm.h>
@@ -189,7 +190,8 @@ static int meminfo_read_proc(char *page, char **start, off_t off,
189 "Committed_AS: %8lu kB\n" 190 "Committed_AS: %8lu kB\n"
190 "VmallocTotal: %8lu kB\n" 191 "VmallocTotal: %8lu kB\n"
191 "VmallocUsed: %8lu kB\n" 192 "VmallocUsed: %8lu kB\n"
192 "VmallocChunk: %8lu kB\n", 193 "VmallocChunk: %8lu kB\n"
194 "Quicklists: %8lu kB\n",
193 K(i.totalram), 195 K(i.totalram),
194 K(i.freeram), 196 K(i.freeram),
195 K(i.bufferram), 197 K(i.bufferram),
@@ -221,7 +223,8 @@ static int meminfo_read_proc(char *page, char **start, off_t off,
221 K(committed), 223 K(committed),
222 (unsigned long)VMALLOC_TOTAL >> 10, 224 (unsigned long)VMALLOC_TOTAL >> 10,
223 vmi.used >> 10, 225 vmi.used >> 10,
224 vmi.largest_chunk >> 10 226 vmi.largest_chunk >> 10,
227 K(quicklist_total_size())
225 ); 228 );
226 229
227 len += hugetlb_report_meminfo(page + len); 230 len += hugetlb_report_meminfo(page + len);
diff --git a/fs/readdir.c b/fs/readdir.c
index 4e026e5407fb..93a7559bbfd8 100644
--- a/fs/readdir.c
+++ b/fs/readdir.c
@@ -80,8 +80,10 @@ static int fillonedir(void * __buf, const char * name, int namlen, loff_t offset
80 if (buf->result) 80 if (buf->result)
81 return -EINVAL; 81 return -EINVAL;
82 d_ino = ino; 82 d_ino = ino;
83 if (sizeof(d_ino) < sizeof(ino) && d_ino != ino) 83 if (sizeof(d_ino) < sizeof(ino) && d_ino != ino) {
84 buf->result = -EOVERFLOW;
84 return -EOVERFLOW; 85 return -EOVERFLOW;
86 }
85 buf->result++; 87 buf->result++;
86 dirent = buf->dirent; 88 dirent = buf->dirent;
87 if (!access_ok(VERIFY_WRITE, dirent, 89 if (!access_ok(VERIFY_WRITE, dirent,
@@ -155,8 +157,10 @@ static int filldir(void * __buf, const char * name, int namlen, loff_t offset,
155 if (reclen > buf->count) 157 if (reclen > buf->count)
156 return -EINVAL; 158 return -EINVAL;
157 d_ino = ino; 159 d_ino = ino;
158 if (sizeof(d_ino) < sizeof(ino) && d_ino != ino) 160 if (sizeof(d_ino) < sizeof(ino) && d_ino != ino) {
161 buf->error = -EOVERFLOW;
159 return -EOVERFLOW; 162 return -EOVERFLOW;
163 }
160 dirent = buf->previous; 164 dirent = buf->previous;
161 if (dirent) { 165 if (dirent) {
162 if (__put_user(offset, &dirent->d_off)) 166 if (__put_user(offset, &dirent->d_off))
diff --git a/fs/seq_file.c b/fs/seq_file.c
index 5d54205e486b..bd20f7f5a933 100644
--- a/fs/seq_file.c
+++ b/fs/seq_file.c
@@ -108,9 +108,9 @@ ssize_t seq_read(struct file *file, char __user *buf, size_t size, loff_t *ppos)
108 goto Done; 108 goto Done;
109 } 109 }
110 /* we need at least one record in buffer */ 110 /* we need at least one record in buffer */
111 pos = m->index;
112 p = m->op->start(m, &pos);
111 while (1) { 113 while (1) {
112 pos = m->index;
113 p = m->op->start(m, &pos);
114 err = PTR_ERR(p); 114 err = PTR_ERR(p);
115 if (!p || IS_ERR(p)) 115 if (!p || IS_ERR(p))
116 break; 116 break;
@@ -119,6 +119,11 @@ ssize_t seq_read(struct file *file, char __user *buf, size_t size, loff_t *ppos)
119 break; 119 break;
120 if (unlikely(err)) 120 if (unlikely(err))
121 m->count = 0; 121 m->count = 0;
122 if (unlikely(!m->count)) {
123 p = m->op->next(m, p, &pos);
124 m->index = pos;
125 continue;
126 }
122 if (m->count < m->size) 127 if (m->count < m->size)
123 goto Fill; 128 goto Fill;
124 m->op->stop(m, p); 129 m->op->stop(m, p);
@@ -128,6 +133,8 @@ ssize_t seq_read(struct file *file, char __user *buf, size_t size, loff_t *ppos)
128 goto Enomem; 133 goto Enomem;
129 m->count = 0; 134 m->count = 0;
130 m->version = 0; 135 m->version = 0;
136 pos = m->index;
137 p = m->op->start(m, &pos);
131 } 138 }
132 m->op->stop(m, p); 139 m->op->stop(m, p);
133 m->count = 0; 140 m->count = 0;
diff --git a/fs/xfs/linux-2.6/xfs_file.c b/fs/xfs/linux-2.6/xfs_file.c
index 5f60363b9343..5311c1acdd40 100644
--- a/fs/xfs/linux-2.6/xfs_file.c
+++ b/fs/xfs/linux-2.6/xfs_file.c
@@ -475,6 +475,7 @@ const struct file_operations xfs_invis_file_operations = {
475const struct file_operations xfs_dir_file_operations = { 475const struct file_operations xfs_dir_file_operations = {
476 .read = generic_read_dir, 476 .read = generic_read_dir,
477 .readdir = xfs_file_readdir, 477 .readdir = xfs_file_readdir,
478 .llseek = generic_file_llseek,
478 .unlocked_ioctl = xfs_file_ioctl, 479 .unlocked_ioctl = xfs_file_ioctl,
479#ifdef CONFIG_COMPAT 480#ifdef CONFIG_COMPAT
480 .compat_ioctl = xfs_file_compat_ioctl, 481 .compat_ioctl = xfs_file_compat_ioctl,
diff --git a/fs/xfs/linux-2.6/xfs_iops.c b/fs/xfs/linux-2.6/xfs_iops.c
index 91bcd979242c..095d271f3434 100644
--- a/fs/xfs/linux-2.6/xfs_iops.c
+++ b/fs/xfs/linux-2.6/xfs_iops.c
@@ -355,7 +355,7 @@ xfs_vn_ci_lookup(
355 /* else case-insensitive match... */ 355 /* else case-insensitive match... */
356 dname.name = ci_name.name; 356 dname.name = ci_name.name;
357 dname.len = ci_name.len; 357 dname.len = ci_name.len;
358 dentry = d_add_ci(VFS_I(ip), dentry, &dname); 358 dentry = d_add_ci(dentry, VFS_I(ip), &dname);
359 kmem_free(ci_name.name); 359 kmem_free(ci_name.name);
360 return dentry; 360 return dentry;
361} 361}
diff --git a/fs/xfs/xfs_dmapi.h b/fs/xfs/xfs_dmapi.h
index cdc2d3464a1a..2813cdd72375 100644
--- a/fs/xfs/xfs_dmapi.h
+++ b/fs/xfs/xfs_dmapi.h
@@ -18,7 +18,6 @@
18#ifndef __XFS_DMAPI_H__ 18#ifndef __XFS_DMAPI_H__
19#define __XFS_DMAPI_H__ 19#define __XFS_DMAPI_H__
20 20
21#include <linux/version.h>
22/* Values used to define the on-disk version of dm_attrname_t. All 21/* Values used to define the on-disk version of dm_attrname_t. All
23 * on-disk attribute names start with the 8-byte string "SGI_DMI_". 22 * on-disk attribute names start with the 8-byte string "SGI_DMI_".
24 * 23 *