aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@linux-foundation.org>2009-04-03 12:53:22 -0400
committerLinus Torvalds <torvalds@linux-foundation.org>2009-04-03 12:53:22 -0400
commit9b59f0316bc556a1b63518f0b1224cf9be48467b (patch)
treed6ffccbe5d9ce5f55c1b2efaf02220c701d4420a
parentac7c1a776dfe1a9c83ea7885f858f5f1a144d8af (diff)
parent0d8fe329a80714e0f729ae48cba8d64cbe5701cb (diff)
Merge branch 'for-linus' of git://git.open-osd.org/linux-open-osd
* 'for-linus' of git://git.open-osd.org/linux-open-osd: fs: Add exofs to Kernel build exofs: Documentation exofs: export_operations exofs: super_operations and file_system_type exofs: dir_inode and directory operations exofs: address_space_operations exofs: symlink_inode and fast_symlink_inode operations exofs: file and file_inode operations exofs: Kbuild, Headers and osd utils
Diffstat
-rw-r--r--Documentation/filesystems/exofs.txt176
-rw-r--r--fs/Kconfig2
-rw-r--r--fs/Makefile1
-rw-r--r--fs/exofs/BUGS3
-rw-r--r--fs/exofs/Kbuild16
-rw-r--r--fs/exofs/Kconfig13
-rw-r--r--fs/exofs/common.h184
-rw-r--r--fs/exofs/dir.c672
-rw-r--r--fs/exofs/exofs.h180
-rw-r--r--fs/exofs/file.c87
-rw-r--r--fs/exofs/inode.c1303
-rw-r--r--fs/exofs/namei.c342
-rw-r--r--fs/exofs/osd.c153
-rw-r--r--fs/exofs/super.c584
-rw-r--r--fs/exofs/symlink.c57
15 files changed, 3773 insertions, 0 deletions
diff --git a/Documentation/filesystems/exofs.txt b/Documentation/filesystems/exofs.txt
new file mode 100644
index 000000000000..0ced74c2f73c
--- /dev/null
+++ b/Documentation/filesystems/exofs.txt
@@ -0,0 +1,176 @@
1===============================================================================
2WHAT IS EXOFS?
3===============================================================================
4
5exofs is a file system that uses an OSD and exports the API of a normal Linux
6file system. Users access exofs like any other local file system, and exofs
7will in turn issue commands to the local OSD initiator.
8
9OSD is a new T10 command set that views storage devices not as a large/flat
10array of sectors but as a container of objects, each having a length, quota,
11time attributes and more. Each object is addressed by a 64bit ID, and is
12contained in a 64bit ID partition. Each object has associated attributes
13attached to it, which are integral part of the object and provide metadata about
14the object. The standard defines some common obligatory attributes, but user
15attributes can be added as needed.
16
17===============================================================================
18ENVIRONMENT
19===============================================================================
20
21To use this file system, you need to have an object store to run it on. You
22may download a target from:
23http://open-osd.org
24
25See Documentation/scsi/osd.txt for how to setup a working osd environment.
26
27===============================================================================
28USAGE
29===============================================================================
30
311. Download and compile exofs and open-osd initiator:
32 You need an external Kernel source tree or kernel headers from your
33 distribution. (anything based on 2.6.26 or later).
34
35 a. download open-osd including exofs source using:
36 [parent-directory]$ git clone git://git.open-osd.org/open-osd.git
37
38 b. Build the library module like this:
39 [parent-directory]$ make -C KSRC=$(KER_DIR) open-osd
40
41 This will build both the open-osd initiator as well as the exofs kernel
42 module. Use whatever parameters you compiled your Kernel with and
43 $(KER_DIR) above pointing to the Kernel you compile against. See the file
44 open-osd/top-level-Makefile for an example.
45
462. Get the OSD initiator and target set up properly, and login to the target.
47 See Documentation/scsi/osd.txt for farther instructions. Also see ./do-osd
48 for example script that does all these steps.
49
503. Insmod the exofs.ko module:
51 [exofs]$ insmod exofs.ko
52
534. Make sure the directory where you want to mount exists. If not, create it.
54 (For example, mkdir /mnt/exofs)
55
565. At first run you will need to invoke the mkfs.exofs application
57
58 As an example, this will create the file system on:
59 /dev/osd0 partition ID 65536
60
61 mkfs.exofs --pid=65536 --format /dev/osd0
62
63 The --format is optional if not specified no OSD_FORMAT will be
64 preformed and a clean file system will be created in the specified pid,
65 in the available space of the target. (Use --format=size_in_meg to limit
66 the total LUN space available)
67
68 If pid already exist it will be deleted and a new one will be created in it's
69 place. Be careful.
70
71 An exofs lives inside a single OSD partition. You can create multiple exofs
72 filesystems on the same device using multiple pids.
73
74 (run mkfs.exofs without any parameters for usage help message)
75
766. Mount the file system.
77
78 For example, to mount /dev/osd0, partition ID 0x10000 on /mnt/exofs:
79
80 mount -t exofs -o pid=65536 /dev/osd0 /mnt/exofs/
81
827. For reference (See do-exofs example script):
83 do-exofs start - an example of how to perform the above steps.
84 do-exofs stop - an example of how to unmount the file system.
85 do-exofs format - an example of how to format and mkfs a new exofs.
86
878. Extra compilation flags (uncomment in fs/exofs/Kbuild):
88 CONFIG_EXOFS_DEBUG - for debug messages and extra checks.
89
90===============================================================================
91exofs mount options
92===============================================================================
93Similar to any mount command:
94 mount -t exofs -o exofs_options /dev/osdX mount_exofs_directory
95
96Where:
97 -t exofs: specifies the exofs file system
98
99 /dev/osdX: X is a decimal number. /dev/osdX was created after a successful
100 login into an OSD target.
101
102 mount_exofs_directory: The directory to mount the file system on
103
104 exofs specific options: Options are separated by commas (,)
105 pid=<integer> - The partition number to mount/create as
106 container of the filesystem.
107 This option is mandatory
108 to=<integer> - Timeout in ticks for a single command
109 default is (60 * HZ) [for debugging only]
110
111===============================================================================
112DESIGN
113===============================================================================
114
115* The file system control block (AKA on-disk superblock) resides in an object
116 with a special ID (defined in common.h).
117 Information included in the file system control block is used to fill the
118 in-memory superblock structure at mount time. This object is created before
119 the file system is used by mkexofs.c It contains information such as:
120 - The file system's magic number
121 - The next inode number to be allocated
122
123* Each file resides in its own object and contains the data (and it will be
124 possible to extend the file over multiple objects, though this has not been
125 implemented yet).
126
127* A directory is treated as a file, and essentially contains a list of <file
128 name, inode #> pairs for files that are found in that directory. The object
129 IDs correspond to the files' inode numbers and will be allocated according to
130 a bitmap (stored in a separate object). Now they are allocated using a
131 counter.
132
133* Each file's control block (AKA on-disk inode) is stored in its object's
134 attributes. This applies to both regular files and other types (directories,
135 device files, symlinks, etc.).
136
137* Credentials are generated per object (inode and superblock) when they is
138 created in memory (read off disk or created). The credential works for all
139 operations and is used as long as the object remains in memory.
140
141* Async OSD operations are used whenever possible, but the target may execute
142 them out of order. The operations that concern us are create, delete,
143 readpage, writepage, update_inode, and truncate. The following pairs of
144 operations should execute in the order written, and we need to prevent them
145 from executing in reverse order:
146 - The following are handled with the OBJ_CREATED and OBJ_2BCREATED
147 flags. OBJ_CREATED is set when we know the object exists on the OSD -
148 in create's callback function, and when we successfully do a read_inode.
149 OBJ_2BCREATED is set in the beginning of the create function, so we
150 know that we should wait.
151 - create/delete: delete should wait until the object is created
152 on the OSD.
153 - create/readpage: readpage should be able to return a page
154 full of zeroes in this case. If there was a write already
155 en-route (i.e. create, writepage, readpage) then the page
156 would be locked, and so it would really be the same as
157 create/writepage.
158 - create/writepage: if writepage is called for a sync write, it
159 should wait until the object is created on the OSD.
160 Otherwise, it should just return.
161 - create/truncate: truncate should wait until the object is
162 created on the OSD.
163 - create/update_inode: update_inode should wait until the
164 object is created on the OSD.
165 - Handled by VFS locks:
166 - readpage/delete: shouldn't happen because of page lock.
167 - writepage/delete: shouldn't happen because of page lock.
168 - readpage/writepage: shouldn't happen because of page lock.
169
170===============================================================================
171LICENSE/COPYRIGHT
172===============================================================================
173The exofs file system is based on ext2 v0.5b (distributed with the Linux kernel
174version 2.6.10). All files include the original copyrights, and the license
175is GPL version 2 (only version 2, as is true for the Linux kernel). The
176Linux kernel can be downloaded from www.kernel.org.
diff --git a/fs/Kconfig b/fs/Kconfig
index cef8b18ceaa3..ae3b34a2ea69 100644
--- a/fs/Kconfig
+++ b/fs/Kconfig
@@ -169,6 +169,8 @@ source "fs/romfs/Kconfig"
169source "fs/sysv/Kconfig" 169source "fs/sysv/Kconfig"
170source "fs/ufs/Kconfig" 170source "fs/ufs/Kconfig"
171 171
172source "fs/exofs/Kconfig"
173
172endif # MISC_FILESYSTEMS 174endif # MISC_FILESYSTEMS
173 175
174menuconfig NETWORK_FILESYSTEMS 176menuconfig NETWORK_FILESYSTEMS
diff --git a/fs/Makefile b/fs/Makefile
index b5cd8e18dd9f..15f73014a208 100644
--- a/fs/Makefile
+++ b/fs/Makefile
@@ -120,3 +120,4 @@ obj-$(CONFIG_DEBUG_FS) += debugfs/
120obj-$(CONFIG_OCFS2_FS) += ocfs2/ 120obj-$(CONFIG_OCFS2_FS) += ocfs2/
121obj-$(CONFIG_BTRFS_FS) += btrfs/ 121obj-$(CONFIG_BTRFS_FS) += btrfs/
122obj-$(CONFIG_GFS2_FS) += gfs2/ 122obj-$(CONFIG_GFS2_FS) += gfs2/
123obj-$(CONFIG_EXOFS_FS) += exofs/
diff --git a/fs/exofs/BUGS b/fs/exofs/BUGS
new file mode 100644
index 000000000000..1b2d4c63a579
--- /dev/null
+++ b/fs/exofs/BUGS
@@ -0,0 +1,3 @@
1- Out-of-space may cause a severe problem if the object (and directory entry)
2 were written, but the inode attributes failed. Then if the filesystem was
3 unmounted and mounted the kernel can get into an endless loop doing a readdir.
diff --git a/fs/exofs/Kbuild b/fs/exofs/Kbuild
new file mode 100644
index 000000000000..cc2d22db119c
--- /dev/null
+++ b/fs/exofs/Kbuild
@@ -0,0 +1,16 @@
1#
2# Kbuild for the EXOFS module
3#
4# Copyright (C) 2008 Panasas Inc. All rights reserved.
5#
6# Authors:
7# Boaz Harrosh <bharrosh@panasas.com>
8#
9# This program is free software; you can redistribute it and/or modify
10# it under the terms of the GNU General Public License version 2
11#
12# Kbuild - Gets included from the Kernels Makefile and build system
13#
14
15exofs-y := osd.o inode.o file.o symlink.o namei.o dir.o super.o
16obj-$(CONFIG_EXOFS_FS) += exofs.o
diff --git a/fs/exofs/Kconfig b/fs/exofs/Kconfig
new file mode 100644
index 000000000000..86194b2f799d
--- /dev/null
+++ b/fs/exofs/Kconfig
@@ -0,0 +1,13 @@
1config EXOFS_FS
2 tristate "exofs: OSD based file system support"
3 depends on SCSI_OSD_ULD
4 help
5 EXOFS is a file system that uses an OSD storage device,
6 as its backing storage.
7
8# Debugging-related stuff
9config EXOFS_DEBUG
10 bool "Enable debugging"
11 depends on EXOFS_FS
12 help
13 This option enables EXOFS debug prints.
diff --git a/fs/exofs/common.h b/fs/exofs/common.h
new file mode 100644
index 000000000000..b1512c4bb8c7
--- /dev/null
+++ b/fs/exofs/common.h
@@ -0,0 +1,184 @@
1/*
2 * common.h - Common definitions for both Kernel and user-mode utilities
3 *
4 * Copyright (C) 2005, 2006
5 * Avishay Traeger (avishay@gmail.com) (avishay@il.ibm.com)
6 * Copyright (C) 2005, 2006
7 * International Business Machines
8 * Copyright (C) 2008, 2009
9 * Boaz Harrosh <bharrosh@panasas.com>
10 *
11 * Copyrights for code taken from ext2:
12 * Copyright (C) 1992, 1993, 1994, 1995
13 * Remy Card (card@masi.ibp.fr)
14 * Laboratoire MASI - Institut Blaise Pascal
15 * Universite Pierre et Marie Curie (Paris VI)
16 * from
17 * linux/fs/minix/inode.c
18 * Copyright (C) 1991, 1992 Linus Torvalds
19 *
20 * This file is part of exofs.
21 *
22 * exofs is free software; you can redistribute it and/or modify
23 * it under the terms of the GNU General Public License as published by
24 * the Free Software Foundation. Since it is based on ext2, and the only
25 * valid version of GPL for the Linux kernel is version 2, the only valid
26 * version of GPL for exofs is version 2.
27 *
28 * exofs is distributed in the hope that it will be useful,
29 * but WITHOUT ANY WARRANTY; without even the implied warranty of
30 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
31 * GNU General Public License for more details.
32 *
33 * You should have received a copy of the GNU General Public License
34 * along with exofs; if not, write to the Free Software
35 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
36 */
37
38#ifndef __EXOFS_COM_H__
39#define __EXOFS_COM_H__
40
41#include <linux/types.h>
42
43#include <scsi/osd_attributes.h>
44#include <scsi/osd_initiator.h>
45#include <scsi/osd_sec.h>
46
47/****************************************************************************
48 * Object ID related defines
49 * NOTE: inode# = object ID - EXOFS_OBJ_OFF
50 ****************************************************************************/
51#define EXOFS_MIN_PID 0x10000 /* Smallest partition ID */
52#define EXOFS_OBJ_OFF 0x10000 /* offset for objects */
53#define EXOFS_SUPER_ID 0x10000 /* object ID for on-disk superblock */
54#define EXOFS_ROOT_ID 0x10002 /* object ID for root directory */
55
56/* exofs Application specific page/attribute */
57# define EXOFS_APAGE_FS_DATA (OSD_APAGE_APP_DEFINED_FIRST + 3)
58# define EXOFS_ATTR_INODE_DATA 1
59
60/*
61 * The maximum number of files we can have is limited by the size of the
62 * inode number. This is the largest object ID that the file system supports.
63 * Object IDs 0, 1, and 2 are always in use (see above defines).
64 */
65enum {
66 EXOFS_MAX_INO_ID = (sizeof(ino_t) * 8 == 64) ? ULLONG_MAX :
67 (1ULL << (sizeof(ino_t) * 8ULL - 1ULL)),
68 EXOFS_MAX_ID = (EXOFS_MAX_INO_ID - 1 - EXOFS_OBJ_OFF),
69};
70
71/****************************************************************************
72 * Misc.
73 ****************************************************************************/
74#define EXOFS_BLKSHIFT 12
75#define EXOFS_BLKSIZE (1UL << EXOFS_BLKSHIFT)
76
77/****************************************************************************
78 * superblock-related things
79 ****************************************************************************/
80#define EXOFS_SUPER_MAGIC 0x5DF5
81
82/*
83 * The file system control block - stored in an object's data (mainly, the one
84 * with ID EXOFS_SUPER_ID). This is where the in-memory superblock is stored
85 * on disk. Right now it just has a magic value, which is basically a sanity
86 * check on our ability to communicate with the object store.
87 */
88struct exofs_fscb {
89 __le64 s_nextid; /* Highest object ID used */
90 __le32 s_numfiles; /* Number of files on fs */
91 __le16 s_magic; /* Magic signature */
92 __le16 s_newfs; /* Non-zero if this is a new fs */
93};
94
95/****************************************************************************
96 * inode-related things
97 ****************************************************************************/
98#define EXOFS_IDATA 5
99
100/*
101 * The file control block - stored in an object's attributes. This is where
102 * the in-memory inode is stored on disk.
103 */
104struct exofs_fcb {
105 __le64 i_size; /* Size of the file */
106 __le16 i_mode; /* File mode */
107 __le16 i_links_count; /* Links count */
108 __le32 i_uid; /* Owner Uid */
109 __le32 i_gid; /* Group Id */
110 __le32 i_atime; /* Access time */
111 __le32 i_ctime; /* Creation time */
112 __le32 i_mtime; /* Modification time */
113 __le32 i_flags; /* File flags (unused for now)*/
114 __le32 i_generation; /* File version (for NFS) */
115 __le32 i_data[EXOFS_IDATA]; /* Short symlink names and device #s */
116};
117
118#define EXOFS_INO_ATTR_SIZE sizeof(struct exofs_fcb)
119
120/* This is the Attribute the fcb is stored in */
121static const struct __weak osd_attr g_attr_inode_data = ATTR_DEF(
122 EXOFS_APAGE_FS_DATA,
123 EXOFS_ATTR_INODE_DATA,
124 EXOFS_INO_ATTR_SIZE);
125
126/****************************************************************************
127 * dentry-related things
128 ****************************************************************************/
129#define EXOFS_NAME_LEN 255
130
131/*
132 * The on-disk directory entry
133 */
134struct exofs_dir_entry {
135 __le64 inode_no; /* inode number */
136 __le16 rec_len; /* directory entry length */
137 u8 name_len; /* name length */
138 u8 file_type; /* umm...file type */
139 char name[EXOFS_NAME_LEN]; /* file name */
140};
141
142enum {
143 EXOFS_FT_UNKNOWN,
144 EXOFS_FT_REG_FILE,
145 EXOFS_FT_DIR,
146 EXOFS_FT_CHRDEV,
147 EXOFS_FT_BLKDEV,
148 EXOFS_FT_FIFO,
149 EXOFS_FT_SOCK,
150 EXOFS_FT_SYMLINK,
151 EXOFS_FT_MAX
152};
153
154#define EXOFS_DIR_PAD 4
155#define EXOFS_DIR_ROUND (EXOFS_DIR_PAD - 1)
156#define EXOFS_DIR_REC_LEN(name_len) \
157 (((name_len) + offsetof(struct exofs_dir_entry, name) + \
158 EXOFS_DIR_ROUND) & ~EXOFS_DIR_ROUND)
159
160/*************************
161 * function declarations *
162 *************************/
163/* osd.c */
164void exofs_make_credential(u8 cred_a[OSD_CAP_LEN],
165 const struct osd_obj_id *obj);
166
167int exofs_check_ok_resid(struct osd_request *or, u64 *in_resid, u64 *out_resid);
168static inline int exofs_check_ok(struct osd_request *or)
169{
170 return exofs_check_ok_resid(or, NULL, NULL);
171}
172int exofs_sync_op(struct osd_request *or, int timeout, u8 *cred);
173int exofs_async_op(struct osd_request *or,
174 osd_req_done_fn *async_done, void *caller_context, u8 *cred);
175
176int extract_attr_from_req(struct osd_request *or, struct osd_attr *attr);
177
178int osd_req_read_kern(struct osd_request *or,
179 const struct osd_obj_id *obj, u64 offset, void *buff, u64 len);
180
181int osd_req_write_kern(struct osd_request *or,
182 const struct osd_obj_id *obj, u64 offset, void *buff, u64 len);
183
184#endif /*ifndef __EXOFS_COM_H__*/
diff --git a/fs/exofs/dir.c b/fs/exofs/dir.c
new file mode 100644
index 000000000000..65b0c8c776a1
--- /dev/null
+++ b/fs/exofs/dir.c
@@ -0,0 +1,672 @@
1/*
2 * Copyright (C) 2005, 2006
3 * Avishay Traeger (avishay@gmail.com) (avishay@il.ibm.com)
4 * Copyright (C) 2005, 2006
5 * International Business Machines
6 * Copyright (C) 2008, 2009
7 * Boaz Harrosh <bharrosh@panasas.com>
8 *
9 * Copyrights for code taken from ext2:
10 * Copyright (C) 1992, 1993, 1994, 1995
11 * Remy Card (card@masi.ibp.fr)
12 * Laboratoire MASI - Institut Blaise Pascal
13 * Universite Pierre et Marie Curie (Paris VI)
14 * from
15 * linux/fs/minix/inode.c
16 * Copyright (C) 1991, 1992 Linus Torvalds
17 *
18 * This file is part of exofs.
19 *
20 * exofs is free software; you can redistribute it and/or modify
21 * it under the terms of the GNU General Public License as published by
22 * the Free Software Foundation. Since it is based on ext2, and the only
23 * valid version of GPL for the Linux kernel is version 2, the only valid
24 * version of GPL for exofs is version 2.
25 *
26 * exofs is distributed in the hope that it will be useful,
27 * but WITHOUT ANY WARRANTY; without even the implied warranty of
28 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
29 * GNU General Public License for more details.
30 *
31 * You should have received a copy of the GNU General Public License
32 * along with exofs; if not, write to the Free Software
33 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
34 */
35
36#include "exofs.h"
37
38static inline unsigned exofs_chunk_size(struct inode *inode)
39{
40 return inode->i_sb->s_blocksize;
41}
42
43static inline void exofs_put_page(struct page *page)
44{
45 kunmap(page);
46 page_cache_release(page);
47}
48
49/* Accesses dir's inode->i_size must be called under inode lock */
50static inline unsigned long dir_pages(struct inode *inode)
51{
52 return (inode->i_size + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
53}
54
55static unsigned exofs_last_byte(struct inode *inode, unsigned long page_nr)
56{
57 loff_t last_byte = inode->i_size;
58
59 last_byte -= page_nr << PAGE_CACHE_SHIFT;
60 if (last_byte > PAGE_CACHE_SIZE)
61 last_byte = PAGE_CACHE_SIZE;
62 return last_byte;
63}
64
65static int exofs_commit_chunk(struct page *page, loff_t pos, unsigned len)
66{
67 struct address_space *mapping = page->mapping;
68 struct inode *dir = mapping->host;
69 int err = 0;
70
71 dir->i_version++;
72
73 if (!PageUptodate(page))
74 SetPageUptodate(page);
75
76 if (pos+len > dir->i_size) {
77 i_size_write(dir, pos+len);
78 mark_inode_dirty(dir);
79 }
80 set_page_dirty(page);
81
82 if (IS_DIRSYNC(dir))
83 err = write_one_page(page, 1);
84 else
85 unlock_page(page);
86
87 return err;
88}
89
90static void exofs_check_page(struct page *page)
91{
92 struct inode *dir = page->mapping->host;
93 unsigned chunk_size = exofs_chunk_size(dir);
94 char *kaddr = page_address(page);
95 unsigned offs, rec_len;
96 unsigned limit = PAGE_CACHE_SIZE;
97 struct exofs_dir_entry *p;
98 char *error;
99
100 /* if the page is the last one in the directory */
101 if ((dir->i_size >> PAGE_CACHE_SHIFT) == page->index) {
102 limit = dir->i_size & ~PAGE_CACHE_MASK;
103 if (limit & (chunk_size - 1))
104 goto Ebadsize;
105 if (!limit)
106 goto out;
107 }
108 for (offs = 0; offs <= limit - EXOFS_DIR_REC_LEN(1); offs += rec_len) {
109 p = (struct exofs_dir_entry *)(kaddr + offs);
110 rec_len = le16_to_cpu(p->rec_len);
111
112 if (rec_len < EXOFS_DIR_REC_LEN(1))
113 goto Eshort;
114 if (rec_len & 3)
115 goto Ealign;
116 if (rec_len < EXOFS_DIR_REC_LEN(p->name_len))
117 goto Enamelen;
118 if (((offs + rec_len - 1) ^ offs) & ~(chunk_size-1))
119 goto Espan;
120 }
121 if (offs != limit)
122 goto Eend;
123out:
124 SetPageChecked(page);
125 return;
126
127Ebadsize:
128 EXOFS_ERR("ERROR [exofs_check_page]: "
129 "size of directory #%lu is not a multiple of chunk size",
130 dir->i_ino
131 );
132 goto fail;
133Eshort:
134 error = "rec_len is smaller than minimal";
135 goto bad_entry;
136Ealign:
137 error = "unaligned directory entry";
138 goto bad_entry;
139Enamelen:
140 error = "rec_len is too small for name_len";
141 goto bad_entry;
142Espan:
143 error = "directory entry across blocks";
144 goto bad_entry;
145bad_entry:
146 EXOFS_ERR(
147 "ERROR [exofs_check_page]: bad entry in directory #%lu: %s - "
148 "offset=%lu, inode=%llu, rec_len=%d, name_len=%d",
149 dir->i_ino, error, (page->index<<PAGE_CACHE_SHIFT)+offs,
150 _LLU(le64_to_cpu(p->inode_no)),
151 rec_len, p->name_len);
152 goto fail;
153Eend:
154 p = (struct exofs_dir_entry *)(kaddr + offs);
155 EXOFS_ERR("ERROR [exofs_check_page]: "
156 "entry in directory #%lu spans the page boundary"
157 "offset=%lu, inode=%llu",
158 dir->i_ino, (page->index<<PAGE_CACHE_SHIFT)+offs,
159 _LLU(le64_to_cpu(p->inode_no)));
160fail:
161 SetPageChecked(page);
162 SetPageError(page);
163}
164
165static struct page *exofs_get_page(struct inode *dir, unsigned long n)
166{
167 struct address_space *mapping = dir->i_mapping;
168 struct page *page = read_mapping_page(mapping, n, NULL);
169
170 if (!IS_ERR(page)) {
171 kmap(page);
172 if (!PageChecked(page))
173 exofs_check_page(page);
174 if (PageError(page))
175 goto fail;
176 }
177 return page;
178
179fail:
180 exofs_put_page(page);
181 return ERR_PTR(-EIO);
182}
183
184static inline int exofs_match(int len, const unsigned char *name,
185 struct exofs_dir_entry *de)
186{
187 if (len != de->name_len)
188 return 0;
189 if (!de->inode_no)
190 return 0;
191 return !memcmp(name, de->name, len);
192}
193
194static inline
195struct exofs_dir_entry *exofs_next_entry(struct exofs_dir_entry *p)
196{
197 return (struct exofs_dir_entry *)((char *)p + le16_to_cpu(p->rec_len));
198}
199
200static inline unsigned
201exofs_validate_entry(char *base, unsigned offset, unsigned mask)
202{
203 struct exofs_dir_entry *de = (struct exofs_dir_entry *)(base + offset);
204 struct exofs_dir_entry *p =
205 (struct exofs_dir_entry *)(base + (offset&mask));
206 while ((char *)p < (char *)de) {
207 if (p->rec_len == 0)
208 break;
209 p = exofs_next_entry(p);
210 }
211 return (char *)p - base;
212}
213
214static unsigned char exofs_filetype_table[EXOFS_FT_MAX] = {
215 [EXOFS_FT_UNKNOWN] = DT_UNKNOWN,
216 [EXOFS_FT_REG_FILE] = DT_REG,
217 [EXOFS_FT_DIR] = DT_DIR,
218 [EXOFS_FT_CHRDEV] = DT_CHR,
219 [EXOFS_FT_BLKDEV] = DT_BLK,
220 [EXOFS_FT_FIFO] = DT_FIFO,
221 [EXOFS_FT_SOCK] = DT_SOCK,
222 [EXOFS_FT_SYMLINK] = DT_LNK,
223};
224
225#define S_SHIFT 12
226static unsigned char exofs_type_by_mode[S_IFMT >> S_SHIFT] = {
227 [S_IFREG >> S_SHIFT] = EXOFS_FT_REG_FILE,
228 [S_IFDIR >> S_SHIFT] = EXOFS_FT_DIR,
229 [S_IFCHR >> S_SHIFT] = EXOFS_FT_CHRDEV,
230 [S_IFBLK >> S_SHIFT] = EXOFS_FT_BLKDEV,
231 [S_IFIFO >> S_SHIFT] = EXOFS_FT_FIFO,
232 [S_IFSOCK >> S_SHIFT] = EXOFS_FT_SOCK,
233 [S_IFLNK >> S_SHIFT] = EXOFS_FT_SYMLINK,
234};
235
236static inline
237void exofs_set_de_type(struct exofs_dir_entry *de, struct inode *inode)
238{
239 mode_t mode = inode->i_mode;
240 de->file_type = exofs_type_by_mode[(mode & S_IFMT) >> S_SHIFT];
241}
242
243static int
244exofs_readdir(struct file *filp, void *dirent, filldir_t filldir)
245{
246 loff_t pos = filp->f_pos;
247 struct inode *inode = filp->f_path.dentry->d_inode;
248 unsigned int offset = pos & ~PAGE_CACHE_MASK;
249 unsigned long n = pos >> PAGE_CACHE_SHIFT;
250 unsigned long npages = dir_pages(inode);
251 unsigned chunk_mask = ~(exofs_chunk_size(inode)-1);
252 unsigned char *types = NULL;
253 int need_revalidate = (filp->f_version != inode->i_version);
254
255 if (pos > inode->i_size - EXOFS_DIR_REC_LEN(1))
256 return 0;
257
258 types = exofs_filetype_table;
259
260 for ( ; n < npages; n++, offset = 0) {
261 char *kaddr, *limit;
262 struct exofs_dir_entry *de;
263 struct page *page = exofs_get_page(inode, n);
264
265 if (IS_ERR(page)) {
266 EXOFS_ERR("ERROR: "
267 "bad page in #%lu",
268 inode->i_ino);
269 filp->f_pos += PAGE_CACHE_SIZE - offset;
270 return PTR_ERR(page);
271 }
272 kaddr = page_address(page);
273 if (unlikely(need_revalidate)) {
274 if (offset) {
275 offset = exofs_validate_entry(kaddr, offset,
276 chunk_mask);
277 filp->f_pos = (n<<PAGE_CACHE_SHIFT) + offset;
278 }
279 filp->f_version = inode->i_version;
280 need_revalidate = 0;
281 }
282 de = (struct exofs_dir_entry *)(kaddr + offset);
283 limit = kaddr + exofs_last_byte(inode, n) -
284 EXOFS_DIR_REC_LEN(1);
285 for (; (char *)de <= limit; de = exofs_next_entry(de)) {
286 if (de->rec_len == 0) {
287 EXOFS_ERR("ERROR: "
288 "zero-length directory entry");
289 exofs_put_page(page);
290 return -EIO;
291 }
292 if (de->inode_no) {
293 int over;
294 unsigned char d_type = DT_UNKNOWN;
295
296 if (types && de->file_type < EXOFS_FT_MAX)
297 d_type = types[de->file_type];
298
299 offset = (char *)de - kaddr;
300 over = filldir(dirent, de->name, de->name_len,
301 (n<<PAGE_CACHE_SHIFT) | offset,
302 le64_to_cpu(de->inode_no),
303 d_type);
304 if (over) {
305 exofs_put_page(page);
306 return 0;
307 }
308 }
309 filp->f_pos += le16_to_cpu(de->rec_len);
310 }
311 exofs_put_page(page);
312 }
313
314 return 0;
315}
316
317struct exofs_dir_entry *exofs_find_entry(struct inode *dir,
318 struct dentry *dentry, struct page **res_page)
319{
320 const unsigned char *name = dentry->d_name.name;
321 int namelen = dentry->d_name.len;
322 unsigned reclen = EXOFS_DIR_REC_LEN(namelen);
323 unsigned long start, n;
324 unsigned long npages = dir_pages(dir);
325 struct page *page = NULL;
326 struct exofs_i_info *oi = exofs_i(dir);
327 struct exofs_dir_entry *de;
328
329 if (npages == 0)
330 goto out;
331
332 *res_page = NULL;
333
334 start = oi->i_dir_start_lookup;
335 if (start >= npages)
336 start = 0;
337 n = start;
338 do {
339 char *kaddr;
340 page = exofs_get_page(dir, n);
341 if (!IS_ERR(page)) {
342 kaddr = page_address(page);
343 de = (struct exofs_dir_entry *) kaddr;
344 kaddr += exofs_last_byte(dir, n) - reclen;
345 while ((char *) de <= kaddr) {
346 if (de->rec_len == 0) {
347 EXOFS_ERR(
348 "ERROR: exofs_find_entry: "
349 "zero-length directory entry");
350 exofs_put_page(page);
351 goto out;
352 }
353 if (exofs_match(namelen, name, de))
354 goto found;
355 de = exofs_next_entry(de);
356 }
357 exofs_put_page(page);
358 }
359 if (++n >= npages)
360 n = 0;
361 } while (n != start);
362out:
363 return NULL;
364
365found:
366 *res_page = page;
367 oi->i_dir_start_lookup = n;
368 return de;
369}
370
371struct exofs_dir_entry *exofs_dotdot(struct inode *dir, struct page **p)
372{
373 struct page *page = exofs_get_page(dir, 0);
374 struct exofs_dir_entry *de = NULL;
375
376 if (!IS_ERR(page)) {
377 de = exofs_next_entry(
378 (struct exofs_dir_entry *)page_address(page));
379 *p = page;
380 }
381 return de;
382}
383
384ino_t exofs_parent_ino(struct dentry *child)
385{
386 struct page *page;
387 struct exofs_dir_entry *de;
388 ino_t ino;
389
390 de = exofs_dotdot(child->d_inode, &page);
391 if (!de)
392 return 0;
393
394 ino = le64_to_cpu(de->inode_no);
395 exofs_put_page(page);
396 return ino;
397}
398
399ino_t exofs_inode_by_name(struct inode *dir, struct dentry *dentry)
400{
401 ino_t res = 0;
402 struct exofs_dir_entry *de;
403 struct page *page;
404
405 de = exofs_find_entry(dir, dentry, &page);
406 if (de) {
407 res = le64_to_cpu(de->inode_no);
408 exofs_put_page(page);
409 }
410 return res;
411}
412
413int exofs_set_link(struct inode *dir, struct exofs_dir_entry *de,
414 struct page *page, struct inode *inode)
415{
416 loff_t pos = page_offset(page) +
417 (char *) de - (char *) page_address(page);
418 unsigned len = le16_to_cpu(de->rec_len);
419 int err;
420
421 lock_page(page);
422 err = exofs_write_begin(NULL, page->mapping, pos, len,
423 AOP_FLAG_UNINTERRUPTIBLE, &page, NULL);
424 if (err)
425 EXOFS_ERR("exofs_set_link: exofs_write_begin FAILD => %d\n",
426 err);
427
428 de->inode_no = cpu_to_le64(inode->i_ino);
429 exofs_set_de_type(de, inode);
430 if (likely(!err))
431 err = exofs_commit_chunk(page, pos, len);
432 exofs_put_page(page);
433 dir->i_mtime = dir->i_ctime = CURRENT_TIME;
434 mark_inode_dirty(dir);
435 return err;
436}
437
438int exofs_add_link(struct dentry *dentry, struct inode *inode)
439{
440 struct inode *dir = dentry->d_parent->d_inode;
441 const unsigned char *name = dentry->d_name.name;
442 int namelen = dentry->d_name.len;
443 unsigned chunk_size = exofs_chunk_size(dir);
444 unsigned reclen = EXOFS_DIR_REC_LEN(namelen);
445 unsigned short rec_len, name_len;
446 struct page *page = NULL;
447 struct exofs_sb_info *sbi = inode->i_sb->s_fs_info;
448 struct exofs_dir_entry *de;
449 unsigned long npages = dir_pages(dir);
450 unsigned long n;
451 char *kaddr;
452 loff_t pos;
453 int err;
454
455 for (n = 0; n <= npages; n++) {
456 char *dir_end;
457
458 page = exofs_get_page(dir, n);
459 err = PTR_ERR(page);
460 if (IS_ERR(page))
461 goto out;
462 lock_page(page);
463 kaddr = page_address(page);
464 dir_end = kaddr + exofs_last_byte(dir, n);
465 de = (struct exofs_dir_entry *)kaddr;
466 kaddr += PAGE_CACHE_SIZE - reclen;
467 while ((char *)de <= kaddr) {
468 if ((char *)de == dir_end) {
469 name_len = 0;
470 rec_len = chunk_size;
471 de->rec_len = cpu_to_le16(chunk_size);
472 de->inode_no = 0;
473 goto got_it;
474 }
475 if (de->rec_len == 0) {
476 EXOFS_ERR("ERROR: exofs_add_link: "
477 "zero-length directory entry");
478 err = -EIO;
479 goto out_unlock;
480 }
481 err = -EEXIST;
482 if (exofs_match(namelen, name, de))
483 goto out_unlock;
484 name_len = EXOFS_DIR_REC_LEN(de->name_len);
485 rec_len = le16_to_cpu(de->rec_len);
486 if (!de->inode_no && rec_len >= reclen)
487 goto got_it;
488 if (rec_len >= name_len + reclen)
489 goto got_it;
490 de = (struct exofs_dir_entry *) ((char *) de + rec_len);
491 }
492 unlock_page(page);
493 exofs_put_page(page);
494 }
495
496 EXOFS_ERR("exofs_add_link: BAD dentry=%p or inode=%p", dentry, inode);
497 return -EINVAL;
498
499got_it:
500 pos = page_offset(page) +
501 (char *)de - (char *)page_address(page);
502 err = exofs_write_begin(NULL, page->mapping, pos, rec_len, 0,
503 &page, NULL);
504 if (err)
505 goto out_unlock;
506 if (de->inode_no) {
507 struct exofs_dir_entry *de1 =
508 (struct exofs_dir_entry *)((char *)de + name_len);
509 de1->rec_len = cpu_to_le16(rec_len - name_len);
510 de->rec_len = cpu_to_le16(name_len);
511 de = de1;
512 }
513 de->name_len = namelen;
514 memcpy(de->name, name, namelen);
515 de->inode_no = cpu_to_le64(inode->i_ino);
516 exofs_set_de_type(de, inode);
517 err = exofs_commit_chunk(page, pos, rec_len);
518 dir->i_mtime = dir->i_ctime = CURRENT_TIME;
519 mark_inode_dirty(dir);
520 sbi->s_numfiles++;
521
522out_put:
523 exofs_put_page(page);
524out:
525 return err;
526out_unlock:
527 unlock_page(page);
528 goto out_put;
529}
530
531int exofs_delete_entry(struct exofs_dir_entry *dir, struct page *page)
532{
533 struct address_space *mapping = page->mapping;
534 struct inode *inode = mapping->host;
535 struct exofs_sb_info *sbi = inode->i_sb->s_fs_info;
536 char *kaddr = page_address(page);
537 unsigned from = ((char *)dir - kaddr) & ~(exofs_chunk_size(inode)-1);
538 unsigned to = ((char *)dir - kaddr) + le16_to_cpu(dir->rec_len);
539 loff_t pos;
540 struct exofs_dir_entry *pde = NULL;
541 struct exofs_dir_entry *de = (struct exofs_dir_entry *) (kaddr + from);
542 int err;
543
544 while (de < dir) {
545 if (de->rec_len == 0) {
546 EXOFS_ERR("ERROR: exofs_delete_entry:"
547 "zero-length directory entry");
548 err = -EIO;
549 goto out;
550 }
551 pde = de;
552 de = exofs_next_entry(de);
553 }
554 if (pde)
555 from = (char *)pde - (char *)page_address(page);
556 pos = page_offset(page) + from;
557 lock_page(page);
558 err = exofs_write_begin(NULL, page->mapping, pos, to - from, 0,
559 &page, NULL);
560 if (err)
561 EXOFS_ERR("exofs_delete_entry: exofs_write_begin FAILD => %d\n",
562 err);
563 if (pde)
564 pde->rec_len = cpu_to_le16(to - from);
565 dir->inode_no = 0;
566 if (likely(!err))
567 err = exofs_commit_chunk(page, pos, to - from);
568 inode->i_ctime = inode->i_mtime = CURRENT_TIME;
569 mark_inode_dirty(inode);
570 sbi->s_numfiles--;
571out:
572 exofs_put_page(page);
573 return err;
574}
575
576/* kept aligned on 4 bytes */
577#define THIS_DIR ".\0\0"
578#define PARENT_DIR "..\0"
579
580int exofs_make_empty(struct inode *inode, struct inode *parent)
581{
582 struct address_space *mapping = inode->i_mapping;
583 struct page *page = grab_cache_page(mapping, 0);
584 unsigned chunk_size = exofs_chunk_size(inode);
585 struct exofs_dir_entry *de;
586 int err;
587 void *kaddr;
588
589 if (!page)
590 return -ENOMEM;
591
592 err = exofs_write_begin(NULL, page->mapping, 0, chunk_size, 0,
593 &page, NULL);
594 if (err) {
595 unlock_page(page);
596 goto fail;
597 }
598
599 kaddr = kmap_atomic(page, KM_USER0);
600 de = (struct exofs_dir_entry *)kaddr;
601 de->name_len = 1;
602 de->rec_len = cpu_to_le16(EXOFS_DIR_REC_LEN(1));
603 memcpy(de->name, THIS_DIR, sizeof(THIS_DIR));
604 de->inode_no = cpu_to_le64(inode->i_ino);
605 exofs_set_de_type(de, inode);
606
607 de = (struct exofs_dir_entry *)(kaddr + EXOFS_DIR_REC_LEN(1));
608 de->name_len = 2;
609 de->rec_len = cpu_to_le16(chunk_size - EXOFS_DIR_REC_LEN(1));
610 de->inode_no = cpu_to_le64(parent->i_ino);
611 memcpy(de->name, PARENT_DIR, sizeof(PARENT_DIR));
612 exofs_set_de_type(de, inode);
613 kunmap_atomic(page, KM_USER0);
614 err = exofs_commit_chunk(page, 0, chunk_size);
615fail:
616 page_cache_release(page);
617 return err;
618}
619
620int exofs_empty_dir(struct inode *inode)
621{
622 struct page *page = NULL;
623 unsigned long i, npages = dir_pages(inode);
624
625 for (i = 0; i < npages; i++) {
626 char *kaddr;
627 struct exofs_dir_entry *de;
628 page = exofs_get_page(inode, i);
629
630 if (IS_ERR(page))
631 continue;
632
633 kaddr = page_address(page);
634 de = (struct exofs_dir_entry *)kaddr;
635 kaddr += exofs_last_byte(inode, i) - EXOFS_DIR_REC_LEN(1);
636
637 while ((char *)de <= kaddr) {
638 if (de->rec_len == 0) {
639 EXOFS_ERR("ERROR: exofs_empty_dir: "
640 "zero-length directory entry"
641 "kaddr=%p, de=%p\n", kaddr, de);
642 goto not_empty;
643 }
644 if (de->inode_no != 0) {
645 /* check for . and .. */
646 if (de->name[0] != '.')
647 goto not_empty;
648 if (de->name_len > 2)
649 goto not_empty;
650 if (de->name_len < 2) {
651 if (le64_to_cpu(de->inode_no) !=
652 inode->i_ino)
653 goto not_empty;
654 } else if (de->name[1] != '.')
655 goto not_empty;
656 }
657 de = exofs_next_entry(de);
658 }
659 exofs_put_page(page);
660 }
661 return 1;
662
663not_empty:
664 exofs_put_page(page);
665 return 0;
666}
667
668const struct file_operations exofs_dir_operations = {
669 .llseek = generic_file_llseek,
670 .read = generic_read_dir,
671 .readdir = exofs_readdir,
672};
diff --git a/fs/exofs/exofs.h b/fs/exofs/exofs.h
new file mode 100644
index 000000000000..0fd4c7859679
--- /dev/null
+++ b/fs/exofs/exofs.h
@@ -0,0 +1,180 @@
1/*
2 * Copyright (C) 2005, 2006
3 * Avishay Traeger (avishay@gmail.com) (avishay@il.ibm.com)
4 * Copyright (C) 2005, 2006
5 * International Business Machines
6 * Copyright (C) 2008, 2009
7 * Boaz Harrosh <bharrosh@panasas.com>
8 *
9 * Copyrights for code taken from ext2:
10 * Copyright (C) 1992, 1993, 1994, 1995
11 * Remy Card (card@masi.ibp.fr)
12 * Laboratoire MASI - Institut Blaise Pascal
13 * Universite Pierre et Marie Curie (Paris VI)
14 * from
15 * linux/fs/minix/inode.c
16 * Copyright (C) 1991, 1992 Linus Torvalds
17 *
18 * This file is part of exofs.
19 *
20 * exofs is free software; you can redistribute it and/or modify
21 * it under the terms of the GNU General Public License as published by
22 * the Free Software Foundation. Since it is based on ext2, and the only
23 * valid version of GPL for the Linux kernel is version 2, the only valid
24 * version of GPL for exofs is version 2.
25 *
26 * exofs is distributed in the hope that it will be useful,
27 * but WITHOUT ANY WARRANTY; without even the implied warranty of
28 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
29 * GNU General Public License for more details.
30 *
31 * You should have received a copy of the GNU General Public License
32 * along with exofs; if not, write to the Free Software
33 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
34 */
35
36#include <linux/fs.h>
37#include <linux/time.h>
38#include "common.h"
39
40#ifndef __EXOFS_H__
41#define __EXOFS_H__
42
43#define EXOFS_ERR(fmt, a...) printk(KERN_ERR "exofs: " fmt, ##a)
44
45#ifdef CONFIG_EXOFS_DEBUG
46#define EXOFS_DBGMSG(fmt, a...) \
47 printk(KERN_NOTICE "exofs @%s:%d: " fmt, __func__, __LINE__, ##a)
48#else
49#define EXOFS_DBGMSG(fmt, a...) \
50 do { if (0) printk(fmt, ##a); } while (0)
51#endif
52
53/* u64 has problems with printk this will cast it to unsigned long long */
54#define _LLU(x) (unsigned long long)(x)
55
56/*
57 * our extension to the in-memory superblock
58 */
59struct exofs_sb_info {
60 struct osd_dev *s_dev; /* returned by get_osd_dev */
61 osd_id s_pid; /* partition ID of file system*/
62 int s_timeout; /* timeout for OSD operations */
63 uint64_t s_nextid; /* highest object ID used */
64 uint32_t s_numfiles; /* number of files on fs */
65 spinlock_t s_next_gen_lock; /* spinlock for gen # update */
66 u32 s_next_generation; /* next gen # to use */
67 atomic_t s_curr_pending; /* number of pending commands */
68 uint8_t s_cred[OSD_CAP_LEN]; /* all-powerful credential */
69};
70
71/*
72 * our extension to the in-memory inode
73 */
74struct exofs_i_info {
75 unsigned long i_flags; /* various atomic flags */
76 uint32_t i_data[EXOFS_IDATA];/*short symlink names and device #s*/
77 uint32_t i_dir_start_lookup; /* which page to start lookup */
78 wait_queue_head_t i_wq; /* wait queue for inode */
79 uint64_t i_commit_size; /* the object's written length */
80 uint8_t i_cred[OSD_CAP_LEN];/* all-powerful credential */
81 struct inode vfs_inode; /* normal in-memory inode */
82};
83
84/*
85 * our inode flags
86 */
87#define OBJ_2BCREATED 0 /* object will be created soon*/
88#define OBJ_CREATED 1 /* object has been created on the osd*/
89
90static inline int obj_2bcreated(struct exofs_i_info *oi)
91{
92 return test_bit(OBJ_2BCREATED, &oi->i_flags);
93}
94
95static inline void set_obj_2bcreated(struct exofs_i_info *oi)
96{
97 set_bit(OBJ_2BCREATED, &oi->i_flags);
98}
99
100static inline int obj_created(struct exofs_i_info *oi)
101{
102 return test_bit(OBJ_CREATED, &oi->i_flags);
103}
104
105static inline void set_obj_created(struct exofs_i_info *oi)
106{
107 set_bit(OBJ_CREATED, &oi->i_flags);
108}
109
110int __exofs_wait_obj_created(struct exofs_i_info *oi);
111static inline int wait_obj_created(struct exofs_i_info *oi)
112{
113 if (likely(obj_created(oi)))
114 return 0;
115
116 return __exofs_wait_obj_created(oi);
117}
118
119/*
120 * get to our inode from the vfs inode
121 */
122static inline struct exofs_i_info *exofs_i(struct inode *inode)
123{
124 return container_of(inode, struct exofs_i_info, vfs_inode);
125}
126
127/*
128 * Maximum count of links to a file
129 */
130#define EXOFS_LINK_MAX 32000
131
132/*************************
133 * function declarations *
134 *************************/
135/* inode.c */
136void exofs_truncate(struct inode *inode);
137int exofs_setattr(struct dentry *, struct iattr *);
138int exofs_write_begin(struct file *file, struct address_space *mapping,
139 loff_t pos, unsigned len, unsigned flags,
140 struct page **pagep, void **fsdata);
141extern struct inode *exofs_iget(struct super_block *, unsigned long);
142struct inode *exofs_new_inode(struct inode *, int);
143extern int exofs_write_inode(struct inode *, int);
144extern void exofs_delete_inode(struct inode *);
145
146/* dir.c: */
147int exofs_add_link(struct dentry *, struct inode *);
148ino_t exofs_inode_by_name(struct inode *, struct dentry *);
149int exofs_delete_entry(struct exofs_dir_entry *, struct page *);
150int exofs_make_empty(struct inode *, struct inode *);
151struct exofs_dir_entry *exofs_find_entry(struct inode *, struct dentry *,
152 struct page **);
153int exofs_empty_dir(struct inode *);
154struct exofs_dir_entry *exofs_dotdot(struct inode *, struct page **);
155ino_t exofs_parent_ino(struct dentry *child);
156int exofs_set_link(struct inode *, struct exofs_dir_entry *, struct page *,
157 struct inode *);
158
159/*********************
160 * operation vectors *
161 *********************/
162/* dir.c: */
163extern const struct file_operations exofs_dir_operations;
164
165/* file.c */
166extern const struct inode_operations exofs_file_inode_operations;
167extern const struct file_operations exofs_file_operations;
168
169/* inode.c */
170extern const struct address_space_operations exofs_aops;
171
172/* namei.c */
173extern const struct inode_operations exofs_dir_inode_operations;
174extern const struct inode_operations exofs_special_inode_operations;
175
176/* symlink.c */
177extern const struct inode_operations exofs_symlink_inode_operations;
178extern const struct inode_operations exofs_fast_symlink_inode_operations;
179
180#endif
diff --git a/fs/exofs/file.c b/fs/exofs/file.c
new file mode 100644
index 000000000000..6ed7fe484752
--- /dev/null
+++ b/fs/exofs/file.c
@@ -0,0 +1,87 @@
1/*
2 * Copyright (C) 2005, 2006
3 * Avishay Traeger (avishay@gmail.com) (avishay@il.ibm.com)
4 * Copyright (C) 2005, 2006
5 * International Business Machines
6 * Copyright (C) 2008, 2009
7 * Boaz Harrosh <bharrosh@panasas.com>
8 *
9 * Copyrights for code taken from ext2:
10 * Copyright (C) 1992, 1993, 1994, 1995
11 * Remy Card (card@masi.ibp.fr)
12 * Laboratoire MASI - Institut Blaise Pascal
13 * Universite Pierre et Marie Curie (Paris VI)
14 * from
15 * linux/fs/minix/inode.c
16 * Copyright (C) 1991, 1992 Linus Torvalds
17 *
18 * This file is part of exofs.
19 *
20 * exofs is free software; you can redistribute it and/or modify
21 * it under the terms of the GNU General Public License as published by
22 * the Free Software Foundation. Since it is based on ext2, and the only
23 * valid version of GPL for the Linux kernel is version 2, the only valid
24 * version of GPL for exofs is version 2.
25 *
26 * exofs is distributed in the hope that it will be useful,
27 * but WITHOUT ANY WARRANTY; without even the implied warranty of
28 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
29 * GNU General Public License for more details.
30 *
31 * You should have received a copy of the GNU General Public License
32 * along with exofs; if not, write to the Free Software
33 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
34 */
35
36#include <linux/buffer_head.h>
37
38#include "exofs.h"
39
40static int exofs_release_file(struct inode *inode, struct file *filp)
41{
42 return 0;
43}
44
45static int exofs_file_fsync(struct file *filp, struct dentry *dentry,
46 int datasync)
47{
48 int ret;
49 struct address_space *mapping = filp->f_mapping;
50
51 ret = filemap_write_and_wait(mapping);
52 if (ret)
53 return ret;
54
55 /*Note: file_fsync below also calles sync_blockdev, which is a no-op
56 * for exofs, but other then that it does sync_inode and
57 * sync_superblock which is what we need here.
58 */
59 return file_fsync(filp, dentry, datasync);
60}
61
62static int exofs_flush(struct file *file, fl_owner_t id)
63{
64 exofs_file_fsync(file, file->f_path.dentry, 1);
65 /* TODO: Flush the OSD target */
66 return 0;
67}
68
69const struct file_operations exofs_file_operations = {
70 .llseek = generic_file_llseek,
71 .read = do_sync_read,
72 .write = do_sync_write,
73 .aio_read = generic_file_aio_read,
74 .aio_write = generic_file_aio_write,
75 .mmap = generic_file_mmap,
76 .open = generic_file_open,
77 .release = exofs_release_file,
78 .fsync = exofs_file_fsync,
79 .flush = exofs_flush,
80 .splice_read = generic_file_splice_read,
81 .splice_write = generic_file_splice_write,
82};
83
84const struct inode_operations exofs_file_inode_operations = {
85 .truncate = exofs_truncate,
86 .setattr = exofs_setattr,
87};
diff --git a/fs/exofs/inode.c b/fs/exofs/inode.c
new file mode 100644
index 000000000000..ba8d9fab4693
--- /dev/null
+++ b/fs/exofs/inode.c
@@ -0,0 +1,1303 @@
1/*
2 * Copyright (C) 2005, 2006
3 * Avishay Traeger (avishay@gmail.com) (avishay@il.ibm.com)
4 * Copyright (C) 2005, 2006
5 * International Business Machines
6 * Copyright (C) 2008, 2009
7 * Boaz Harrosh <bharrosh@panasas.com>
8 *
9 * Copyrights for code taken from ext2:
10 * Copyright (C) 1992, 1993, 1994, 1995
11 * Remy Card (card@masi.ibp.fr)
12 * Laboratoire MASI - Institut Blaise Pascal
13 * Universite Pierre et Marie Curie (Paris VI)
14 * from
15 * linux/fs/minix/inode.c
16 * Copyright (C) 1991, 1992 Linus Torvalds
17 *
18 * This file is part of exofs.
19 *
20 * exofs is free software; you can redistribute it and/or modify
21 * it under the terms of the GNU General Public License as published by
22 * the Free Software Foundation. Since it is based on ext2, and the only
23 * valid version of GPL for the Linux kernel is version 2, the only valid
24 * version of GPL for exofs is version 2.
25 *
26 * exofs is distributed in the hope that it will be useful,
27 * but WITHOUT ANY WARRANTY; without even the implied warranty of
28 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
29 * GNU General Public License for more details.
30 *
31 * You should have received a copy of the GNU General Public License
32 * along with exofs; if not, write to the Free Software
33 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
34 */
35
36#include <linux/writeback.h>
37#include <linux/buffer_head.h>
38#include <scsi/scsi_device.h>
39
40#include "exofs.h"
41
42#ifdef CONFIG_EXOFS_DEBUG
43# define EXOFS_DEBUG_OBJ_ISIZE 1
44#endif
45
46struct page_collect {
47 struct exofs_sb_info *sbi;
48 struct request_queue *req_q;
49 struct inode *inode;
50 unsigned expected_pages;
51
52 struct bio *bio;
53 unsigned nr_pages;
54 unsigned long length;
55 loff_t pg_first; /* keep 64bit also in 32-arches */
56};
57
58static void _pcol_init(struct page_collect *pcol, unsigned expected_pages,
59 struct inode *inode)
60{
61 struct exofs_sb_info *sbi = inode->i_sb->s_fs_info;
62 struct request_queue *req_q = sbi->s_dev->scsi_device->request_queue;
63
64 pcol->sbi = sbi;
65 pcol->req_q = req_q;
66 pcol->inode = inode;
67 pcol->expected_pages = expected_pages;
68
69 pcol->bio = NULL;
70 pcol->nr_pages = 0;
71 pcol->length = 0;
72 pcol->pg_first = -1;
73
74 EXOFS_DBGMSG("_pcol_init ino=0x%lx expected_pages=%u\n", inode->i_ino,
75 expected_pages);
76}
77
78static void _pcol_reset(struct page_collect *pcol)
79{
80 pcol->expected_pages -= min(pcol->nr_pages, pcol->expected_pages);
81
82 pcol->bio = NULL;
83 pcol->nr_pages = 0;
84 pcol->length = 0;
85 pcol->pg_first = -1;
86 EXOFS_DBGMSG("_pcol_reset ino=0x%lx expected_pages=%u\n",
87 pcol->inode->i_ino, pcol->expected_pages);
88
89 /* this is probably the end of the loop but in writes
90 * it might not end here. don't be left with nothing
91 */
92 if (!pcol->expected_pages)
93 pcol->expected_pages = 128;
94}
95
96static int pcol_try_alloc(struct page_collect *pcol)
97{
98 int pages = min_t(unsigned, pcol->expected_pages, BIO_MAX_PAGES);
99
100 for (; pages; pages >>= 1) {
101 pcol->bio = bio_alloc(GFP_KERNEL, pages);
102 if (likely(pcol->bio))
103 return 0;
104 }
105
106 EXOFS_ERR("Failed to kcalloc expected_pages=%u\n",
107 pcol->expected_pages);
108 return -ENOMEM;
109}
110
111static void pcol_free(struct page_collect *pcol)
112{
113 bio_put(pcol->bio);
114 pcol->bio = NULL;
115}
116
117static int pcol_add_page(struct page_collect *pcol, struct page *page,
118 unsigned len)
119{
120 int added_len = bio_add_pc_page(pcol->req_q, pcol->bio, page, len, 0);
121 if (unlikely(len != added_len))
122 return -ENOMEM;
123
124 ++pcol->nr_pages;
125 pcol->length += len;
126 return 0;
127}
128
129static int update_read_page(struct page *page, int ret)
130{
131 if (ret == 0) {
132 /* Everything is OK */
133 SetPageUptodate(page);
134 if (PageError(page))
135 ClearPageError(page);
136 } else if (ret == -EFAULT) {
137 /* In this case we were trying to read something that wasn't on
138 * disk yet - return a page full of zeroes. This should be OK,
139 * because the object should be empty (if there was a write
140 * before this read, the read would be waiting with the page
141 * locked */
142 clear_highpage(page);
143
144 SetPageUptodate(page);
145 if (PageError(page))
146 ClearPageError(page);
147 ret = 0; /* recovered error */
148 EXOFS_DBGMSG("recovered read error\n");
149 } else /* Error */
150 SetPageError(page);
151
152 return ret;
153}
154
155static void update_write_page(struct page *page, int ret)
156{
157 if (ret) {
158 mapping_set_error(page->mapping, ret);
159 SetPageError(page);
160 }
161 end_page_writeback(page);
162}
163
164/* Called at the end of reads, to optionally unlock pages and update their
165 * status.
166 */
167static int __readpages_done(struct osd_request *or, struct page_collect *pcol,
168 bool do_unlock)
169{
170 struct bio_vec *bvec;
171 int i;
172 u64 resid;
173 u64 good_bytes;
174 u64 length = 0;
175 int ret = exofs_check_ok_resid(or, &resid, NULL);
176
177 osd_end_request(or);
178
179 if (likely(!ret))
180 good_bytes = pcol->length;
181 else if (!resid)
182 good_bytes = 0;
183 else
184 good_bytes = pcol->length - resid;
185
186 EXOFS_DBGMSG("readpages_done(0x%lx) good_bytes=0x%llx"
187 " length=0x%lx nr_pages=%u\n",
188 pcol->inode->i_ino, _LLU(good_bytes), pcol->length,
189 pcol->nr_pages);
190
191 __bio_for_each_segment(bvec, pcol->bio, i, 0) {
192 struct page *page = bvec->bv_page;
193 struct inode *inode = page->mapping->host;
194 int page_stat;
195
196 if (inode != pcol->inode)
197 continue; /* osd might add more pages at end */
198
199 if (likely(length < good_bytes))
200 page_stat = 0;
201 else
202 page_stat = ret;
203
204 EXOFS_DBGMSG(" readpages_done(0x%lx, 0x%lx) %s\n",
205 inode->i_ino, page->index,
206 page_stat ? "bad_bytes" : "good_bytes");
207
208 ret = update_read_page(page, page_stat);
209 if (do_unlock)
210 unlock_page(page);
211 length += bvec->bv_len;
212 }
213
214 pcol_free(pcol);
215 EXOFS_DBGMSG("readpages_done END\n");
216 return ret;
217}
218
219/* callback of async reads */
220static void readpages_done(struct osd_request *or, void *p)
221{
222 struct page_collect *pcol = p;
223
224 __readpages_done(or, pcol, true);
225 atomic_dec(&pcol->sbi->s_curr_pending);
226 kfree(p);
227}
228
229static void _unlock_pcol_pages(struct page_collect *pcol, int ret, int rw)
230{
231 struct bio_vec *bvec;
232 int i;
233
234 __bio_for_each_segment(bvec, pcol->bio, i, 0) {
235 struct page *page = bvec->bv_page;
236
237 if (rw == READ)
238 update_read_page(page, ret);
239 else
240 update_write_page(page, ret);
241
242 unlock_page(page);
243 }
244 pcol_free(pcol);
245}
246
247static int read_exec(struct page_collect *pcol, bool is_sync)
248{
249 struct exofs_i_info *oi = exofs_i(pcol->inode);
250 struct osd_obj_id obj = {pcol->sbi->s_pid,
251 pcol->inode->i_ino + EXOFS_OBJ_OFF};
252 struct osd_request *or = NULL;
253 struct page_collect *pcol_copy = NULL;
254 loff_t i_start = pcol->pg_first << PAGE_CACHE_SHIFT;
255 int ret;
256
257 if (!pcol->bio)
258 return 0;
259
260 /* see comment in _readpage() about sync reads */
261 WARN_ON(is_sync && (pcol->nr_pages != 1));
262
263 or = osd_start_request(pcol->sbi->s_dev, GFP_KERNEL);
264 if (unlikely(!or)) {
265 ret = -ENOMEM;
266 goto err;
267 }
268
269 osd_req_read(or, &obj, pcol->bio, i_start);
270
271 if (is_sync) {
272 exofs_sync_op(or, pcol->sbi->s_timeout, oi->i_cred);
273 return __readpages_done(or, pcol, false);
274 }
275
276 pcol_copy = kmalloc(sizeof(*pcol_copy), GFP_KERNEL);
277 if (!pcol_copy) {
278 ret = -ENOMEM;
279 goto err;
280 }
281
282 *pcol_copy = *pcol;
283 ret = exofs_async_op(or, readpages_done, pcol_copy, oi->i_cred);
284 if (unlikely(ret))
285 goto err;
286
287 atomic_inc(&pcol->sbi->s_curr_pending);
288
289 EXOFS_DBGMSG("read_exec obj=0x%llx start=0x%llx length=0x%lx\n",
290 obj.id, _LLU(i_start), pcol->length);
291
292 /* pages ownership was passed to pcol_copy */
293 _pcol_reset(pcol);
294 return 0;
295
296err:
297 if (!is_sync)
298 _unlock_pcol_pages(pcol, ret, READ);
299 kfree(pcol_copy);
300 if (or)
301 osd_end_request(or);
302 return ret;
303}
304
305/* readpage_strip is called either directly from readpage() or by the VFS from
306 * within read_cache_pages(), to add one more page to be read. It will try to
307 * collect as many contiguous pages as posible. If a discontinuity is
308 * encountered, or it runs out of resources, it will submit the previous segment
309 * and will start a new collection. Eventually caller must submit the last
310 * segment if present.
311 */
312static int readpage_strip(void *data, struct page *page)
313{
314 struct page_collect *pcol = data;
315 struct inode *inode = pcol->inode;
316 struct exofs_i_info *oi = exofs_i(inode);
317 loff_t i_size = i_size_read(inode);
318 pgoff_t end_index = i_size >> PAGE_CACHE_SHIFT;
319 size_t len;
320 int ret;
321
322 /* FIXME: Just for debugging, will be removed */
323 if (PageUptodate(page))
324 EXOFS_ERR("PageUptodate(0x%lx, 0x%lx)\n", pcol->inode->i_ino,
325 page->index);
326
327 if (page->index < end_index)
328 len = PAGE_CACHE_SIZE;
329 else if (page->index == end_index)
330 len = i_size & ~PAGE_CACHE_MASK;
331 else
332 len = 0;
333
334 if (!len || !obj_created(oi)) {
335 /* this will be out of bounds, or doesn't exist yet.
336 * Current page is cleared and the request is split
337 */
338 clear_highpage(page);
339
340 SetPageUptodate(page);
341 if (PageError(page))
342 ClearPageError(page);
343
344 unlock_page(page);
345 EXOFS_DBGMSG("readpage_strip(0x%lx, 0x%lx) empty page,"
346 " splitting\n", inode->i_ino, page->index);
347
348 return read_exec(pcol, false);
349 }
350
351try_again:
352
353 if (unlikely(pcol->pg_first == -1)) {
354 pcol->pg_first = page->index;
355 } else if (unlikely((pcol->pg_first + pcol->nr_pages) !=
356 page->index)) {
357 /* Discontinuity detected, split the request */
358 ret = read_exec(pcol, false);
359 if (unlikely(ret))
360 goto fail;
361 goto try_again;
362 }
363
364 if (!pcol->bio) {
365 ret = pcol_try_alloc(pcol);
366 if (unlikely(ret))
367 goto fail;
368 }
369
370 if (len != PAGE_CACHE_SIZE)
371 zero_user(page, len, PAGE_CACHE_SIZE - len);
372
373 EXOFS_DBGMSG(" readpage_strip(0x%lx, 0x%lx) len=0x%zx\n",
374 inode->i_ino, page->index, len);
375
376 ret = pcol_add_page(pcol, page, len);
377 if (ret) {
378 EXOFS_DBGMSG("Failed pcol_add_page pages[i]=%p "
379 "this_len=0x%zx nr_pages=%u length=0x%lx\n",
380 page, len, pcol->nr_pages, pcol->length);
381
382 /* split the request, and start again with current page */
383 ret = read_exec(pcol, false);
384 if (unlikely(ret))
385 goto fail;
386
387 goto try_again;
388 }
389
390 return 0;
391
392fail:
393 /* SetPageError(page); ??? */
394 unlock_page(page);
395 return ret;
396}
397
398static int exofs_readpages(struct file *file, struct address_space *mapping,
399 struct list_head *pages, unsigned nr_pages)
400{
401 struct page_collect pcol;
402 int ret;
403
404 _pcol_init(&pcol, nr_pages, mapping->host);
405
406 ret = read_cache_pages(mapping, pages, readpage_strip, &pcol);
407 if (ret) {
408 EXOFS_ERR("read_cache_pages => %d\n", ret);
409 return ret;
410 }
411
412 return read_exec(&pcol, false);
413}
414
415static int _readpage(struct page *page, bool is_sync)
416{
417 struct page_collect pcol;
418 int ret;
419
420 _pcol_init(&pcol, 1, page->mapping->host);
421
422 /* readpage_strip might call read_exec(,async) inside at several places
423 * but this is safe for is_async=0 since read_exec will not do anything
424 * when we have a single page.
425 */
426 ret = readpage_strip(&pcol, page);
427 if (ret) {
428 EXOFS_ERR("_readpage => %d\n", ret);
429 return ret;
430 }
431
432 return read_exec(&pcol, is_sync);
433}
434
435/*
436 * We don't need the file
437 */
438static int exofs_readpage(struct file *file, struct page *page)
439{
440 return _readpage(page, false);
441}
442
443/* Callback for osd_write. All writes are asynchronouse */
444static void writepages_done(struct osd_request *or, void *p)
445{
446 struct page_collect *pcol = p;
447 struct bio_vec *bvec;
448 int i;
449 u64 resid;
450 u64 good_bytes;
451 u64 length = 0;
452
453 int ret = exofs_check_ok_resid(or, NULL, &resid);
454
455 osd_end_request(or);
456 atomic_dec(&pcol->sbi->s_curr_pending);
457
458 if (likely(!ret))
459 good_bytes = pcol->length;
460 else if (!resid)
461 good_bytes = 0;
462 else
463 good_bytes = pcol->length - resid;
464
465 EXOFS_DBGMSG("writepages_done(0x%lx) good_bytes=0x%llx"
466 " length=0x%lx nr_pages=%u\n",
467 pcol->inode->i_ino, _LLU(good_bytes), pcol->length,
468 pcol->nr_pages);
469
470 __bio_for_each_segment(bvec, pcol->bio, i, 0) {
471 struct page *page = bvec->bv_page;
472 struct inode *inode = page->mapping->host;
473 int page_stat;
474
475 if (inode != pcol->inode)
476 continue; /* osd might add more pages to a bio */
477
478 if (likely(length < good_bytes))
479 page_stat = 0;
480 else
481 page_stat = ret;
482
483 update_write_page(page, page_stat);
484 unlock_page(page);
485 EXOFS_DBGMSG(" writepages_done(0x%lx, 0x%lx) status=%d\n",
486 inode->i_ino, page->index, page_stat);
487
488 length += bvec->bv_len;
489 }
490
491 pcol_free(pcol);
492 kfree(pcol);
493 EXOFS_DBGMSG("writepages_done END\n");
494}
495
496static int write_exec(struct page_collect *pcol)
497{
498 struct exofs_i_info *oi = exofs_i(pcol->inode);
499 struct osd_obj_id obj = {pcol->sbi->s_pid,
500 pcol->inode->i_ino + EXOFS_OBJ_OFF};
501 struct osd_request *or = NULL;
502 struct page_collect *pcol_copy = NULL;
503 loff_t i_start = pcol->pg_first << PAGE_CACHE_SHIFT;
504 int ret;
505
506 if (!pcol->bio)
507 return 0;
508
509 or = osd_start_request(pcol->sbi->s_dev, GFP_KERNEL);
510 if (unlikely(!or)) {
511 EXOFS_ERR("write_exec: Faild to osd_start_request()\n");
512 ret = -ENOMEM;
513 goto err;
514 }
515
516 pcol_copy = kmalloc(sizeof(*pcol_copy), GFP_KERNEL);
517 if (!pcol_copy) {
518 EXOFS_ERR("write_exec: Faild to kmalloc(pcol)\n");
519 ret = -ENOMEM;
520 goto err;
521 }
522
523 *pcol_copy = *pcol;
524
525 osd_req_write(or, &obj, pcol_copy->bio, i_start);
526 ret = exofs_async_op(or, writepages_done, pcol_copy, oi->i_cred);
527 if (unlikely(ret)) {
528 EXOFS_ERR("write_exec: exofs_async_op() Faild\n");
529 goto err;
530 }
531
532 atomic_inc(&pcol->sbi->s_curr_pending);
533 EXOFS_DBGMSG("write_exec(0x%lx, 0x%llx) start=0x%llx length=0x%lx\n",
534 pcol->inode->i_ino, pcol->pg_first, _LLU(i_start),
535 pcol->length);
536 /* pages ownership was passed to pcol_copy */
537 _pcol_reset(pcol);
538 return 0;
539
540err:
541 _unlock_pcol_pages(pcol, ret, WRITE);
542 kfree(pcol_copy);
543 if (or)
544 osd_end_request(or);
545 return ret;
546}
547
548/* writepage_strip is called either directly from writepage() or by the VFS from
549 * within write_cache_pages(), to add one more page to be written to storage.
550 * It will try to collect as many contiguous pages as possible. If a
551 * discontinuity is encountered or it runs out of resources it will submit the
552 * previous segment and will start a new collection.
553 * Eventually caller must submit the last segment if present.
554 */
555static int writepage_strip(struct page *page,
556 struct writeback_control *wbc_unused, void *data)
557{
558 struct page_collect *pcol = data;
559 struct inode *inode = pcol->inode;
560 struct exofs_i_info *oi = exofs_i(inode);
561 loff_t i_size = i_size_read(inode);
562 pgoff_t end_index = i_size >> PAGE_CACHE_SHIFT;
563 size_t len;
564 int ret;
565
566 BUG_ON(!PageLocked(page));
567
568 ret = wait_obj_created(oi);
569 if (unlikely(ret))
570 goto fail;
571
572 if (page->index < end_index)
573 /* in this case, the page is within the limits of the file */
574 len = PAGE_CACHE_SIZE;
575 else {
576 len = i_size & ~PAGE_CACHE_MASK;
577
578 if (page->index > end_index || !len) {
579 /* in this case, the page is outside the limits
580 * (truncate in progress)
581 */
582 ret = write_exec(pcol);
583 if (unlikely(ret))
584 goto fail;
585 if (PageError(page))
586 ClearPageError(page);
587 unlock_page(page);
588 return 0;
589 }
590 }
591
592try_again:
593
594 if (unlikely(pcol->pg_first == -1)) {
595 pcol->pg_first = page->index;
596 } else if (unlikely((pcol->pg_first + pcol->nr_pages) !=
597 page->index)) {
598 /* Discontinuity detected, split the request */
599 ret = write_exec(pcol);
600 if (unlikely(ret))
601 goto fail;
602 goto try_again;
603 }
604
605 if (!pcol->bio) {
606 ret = pcol_try_alloc(pcol);
607 if (unlikely(ret))
608 goto fail;
609 }
610
611 EXOFS_DBGMSG(" writepage_strip(0x%lx, 0x%lx) len=0x%zx\n",
612 inode->i_ino, page->index, len);
613
614 ret = pcol_add_page(pcol, page, len);
615 if (unlikely(ret)) {
616 EXOFS_DBGMSG("Failed pcol_add_page "
617 "nr_pages=%u total_length=0x%lx\n",
618 pcol->nr_pages, pcol->length);
619
620 /* split the request, next loop will start again */
621 ret = write_exec(pcol);
622 if (unlikely(ret)) {
623 EXOFS_DBGMSG("write_exec faild => %d", ret);
624 goto fail;
625 }
626
627 goto try_again;
628 }
629
630 BUG_ON(PageWriteback(page));
631 set_page_writeback(page);
632
633 return 0;
634
635fail:
636 set_bit(AS_EIO, &page->mapping->flags);
637 unlock_page(page);
638 return ret;
639}
640
641static int exofs_writepages(struct address_space *mapping,
642 struct writeback_control *wbc)
643{
644 struct page_collect pcol;
645 long start, end, expected_pages;
646 int ret;
647
648 start = wbc->range_start >> PAGE_CACHE_SHIFT;
649 end = (wbc->range_end == LLONG_MAX) ?
650 start + mapping->nrpages :
651 wbc->range_end >> PAGE_CACHE_SHIFT;
652
653 if (start || end)
654 expected_pages = min(end - start + 1, 32L);
655 else
656 expected_pages = mapping->nrpages;
657
658 EXOFS_DBGMSG("inode(0x%lx) wbc->start=0x%llx wbc->end=0x%llx"
659 " m->nrpages=%lu start=0x%lx end=0x%lx\n",
660 mapping->host->i_ino, wbc->range_start, wbc->range_end,
661 mapping->nrpages, start, end);
662
663 _pcol_init(&pcol, expected_pages, mapping->host);
664
665 ret = write_cache_pages(mapping, wbc, writepage_strip, &pcol);
666 if (ret) {
667 EXOFS_ERR("write_cache_pages => %d\n", ret);
668 return ret;
669 }
670
671 return write_exec(&pcol);
672}
673
674static int exofs_writepage(struct page *page, struct writeback_control *wbc)
675{
676 struct page_collect pcol;
677 int ret;
678
679 _pcol_init(&pcol, 1, page->mapping->host);
680
681 ret = writepage_strip(page, NULL, &pcol);
682 if (ret) {
683 EXOFS_ERR("exofs_writepage => %d\n", ret);
684 return ret;
685 }
686
687 return write_exec(&pcol);
688}
689
690int exofs_write_begin(struct file *file, struct address_space *mapping,
691 loff_t pos, unsigned len, unsigned flags,
692 struct page **pagep, void **fsdata)
693{
694 int ret = 0;
695 struct page *page;
696
697 page = *pagep;
698 if (page == NULL) {
699 ret = simple_write_begin(file, mapping, pos, len, flags, pagep,
700 fsdata);
701 if (ret) {
702 EXOFS_DBGMSG("simple_write_begin faild\n");
703 return ret;
704 }
705
706 page = *pagep;
707 }
708
709 /* read modify write */
710 if (!PageUptodate(page) && (len != PAGE_CACHE_SIZE)) {
711 ret = _readpage(page, true);
712 if (ret) {
713 /*SetPageError was done by _readpage. Is it ok?*/
714 unlock_page(page);
715 EXOFS_DBGMSG("__readpage_filler faild\n");
716 }
717 }
718
719 return ret;
720}
721
722static int exofs_write_begin_export(struct file *file,
723 struct address_space *mapping,
724 loff_t pos, unsigned len, unsigned flags,
725 struct page **pagep, void **fsdata)
726{
727 *pagep = NULL;
728
729 return exofs_write_begin(file, mapping, pos, len, flags, pagep,
730 fsdata);
731}
732
733const struct address_space_operations exofs_aops = {
734 .readpage = exofs_readpage,
735 .readpages = exofs_readpages,
736 .writepage = exofs_writepage,
737 .writepages = exofs_writepages,
738 .write_begin = exofs_write_begin_export,
739 .write_end = simple_write_end,
740};
741
742/******************************************************************************
743 * INODE OPERATIONS
744 *****************************************************************************/
745
746/*
747 * Test whether an inode is a fast symlink.
748 */
749static inline int exofs_inode_is_fast_symlink(struct inode *inode)
750{
751 struct exofs_i_info *oi = exofs_i(inode);
752
753 return S_ISLNK(inode->i_mode) && (oi->i_data[0] != 0);
754}
755
756/*
757 * get_block_t - Fill in a buffer_head
758 * An OSD takes care of block allocation so we just fake an allocation by
759 * putting in the inode's sector_t in the buffer_head.
760 * TODO: What about the case of create==0 and @iblock does not exist in the
761 * object?
762 */
763static int exofs_get_block(struct inode *inode, sector_t iblock,
764 struct buffer_head *bh_result, int create)
765{
766 map_bh(bh_result, inode->i_sb, iblock);
767 return 0;
768}
769
770const struct osd_attr g_attr_logical_length = ATTR_DEF(
771 OSD_APAGE_OBJECT_INFORMATION, OSD_ATTR_OI_LOGICAL_LENGTH, 8);
772
773/*
774 * Truncate a file to the specified size - all we have to do is set the size
775 * attribute. We make sure the object exists first.
776 */
777void exofs_truncate(struct inode *inode)
778{
779 struct exofs_sb_info *sbi = inode->i_sb->s_fs_info;
780 struct exofs_i_info *oi = exofs_i(inode);
781 struct osd_obj_id obj = {sbi->s_pid, inode->i_ino + EXOFS_OBJ_OFF};
782 struct osd_request *or;
783 struct osd_attr attr;
784 loff_t isize = i_size_read(inode);
785 __be64 newsize;
786 int ret;
787
788 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)
789 || S_ISLNK(inode->i_mode)))
790 return;
791 if (exofs_inode_is_fast_symlink(inode))
792 return;
793 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
794 return;
795 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
796
797 nobh_truncate_page(inode->i_mapping, isize, exofs_get_block);
798
799 or = osd_start_request(sbi->s_dev, GFP_KERNEL);
800 if (unlikely(!or)) {
801 EXOFS_ERR("ERROR: exofs_truncate: osd_start_request failed\n");
802 goto fail;
803 }
804
805 osd_req_set_attributes(or, &obj);
806
807 newsize = cpu_to_be64((u64)isize);
808 attr = g_attr_logical_length;
809 attr.val_ptr = &newsize;
810 osd_req_add_set_attr_list(or, &attr, 1);
811
812 /* if we are about to truncate an object, and it hasn't been
813 * created yet, wait
814 */
815 if (unlikely(wait_obj_created(oi)))
816 goto fail;
817
818 ret = exofs_sync_op(or, sbi->s_timeout, oi->i_cred);
819 osd_end_request(or);
820 if (ret)
821 goto fail;
822
823out:
824 mark_inode_dirty(inode);
825 return;
826fail:
827 make_bad_inode(inode);
828 goto out;
829}
830
831/*
832 * Set inode attributes - just call generic functions.
833 */
834int exofs_setattr(struct dentry *dentry, struct iattr *iattr)
835{
836 struct inode *inode = dentry->d_inode;
837 int error;
838
839 error = inode_change_ok(inode, iattr);
840 if (error)
841 return error;
842
843 error = inode_setattr(inode, iattr);
844 return error;
845}
846
847/*
848 * Read an inode from the OSD, and return it as is. We also return the size
849 * attribute in the 'sanity' argument if we got compiled with debugging turned
850 * on.
851 */
852static int exofs_get_inode(struct super_block *sb, struct exofs_i_info *oi,
853 struct exofs_fcb *inode, uint64_t *sanity)
854{
855 struct exofs_sb_info *sbi = sb->s_fs_info;
856 struct osd_request *or;
857 struct osd_attr attr;
858 struct osd_obj_id obj = {sbi->s_pid,
859 oi->vfs_inode.i_ino + EXOFS_OBJ_OFF};
860 int ret;
861
862 exofs_make_credential(oi->i_cred, &obj);
863
864 or = osd_start_request(sbi->s_dev, GFP_KERNEL);
865 if (unlikely(!or)) {
866 EXOFS_ERR("exofs_get_inode: osd_start_request failed.\n");
867 return -ENOMEM;
868 }
869 osd_req_get_attributes(or, &obj);
870
871 /* we need the inode attribute */
872 osd_req_add_get_attr_list(or, &g_attr_inode_data, 1);
873
874#ifdef EXOFS_DEBUG_OBJ_ISIZE
875 /* we get the size attributes to do a sanity check */
876 osd_req_add_get_attr_list(or, &g_attr_logical_length, 1);
877#endif
878
879 ret = exofs_sync_op(or, sbi->s_timeout, oi->i_cred);
880 if (ret)
881 goto out;
882
883 attr = g_attr_inode_data;
884 ret = extract_attr_from_req(or, &attr);
885 if (ret) {
886 EXOFS_ERR("exofs_get_inode: extract_attr_from_req failed\n");
887 goto out;
888 }
889
890 WARN_ON(attr.len != EXOFS_INO_ATTR_SIZE);
891 memcpy(inode, attr.val_ptr, EXOFS_INO_ATTR_SIZE);
892
893#ifdef EXOFS_DEBUG_OBJ_ISIZE
894 attr = g_attr_logical_length;
895 ret = extract_attr_from_req(or, &attr);
896 if (ret) {
897 EXOFS_ERR("ERROR: extract attr from or failed\n");
898 goto out;
899 }
900 *sanity = get_unaligned_be64(attr.val_ptr);
901#endif
902
903out:
904 osd_end_request(or);
905 return ret;
906}
907
908/*
909 * Fill in an inode read from the OSD and set it up for use
910 */
911struct inode *exofs_iget(struct super_block *sb, unsigned long ino)
912{
913 struct exofs_i_info *oi;
914 struct exofs_fcb fcb;
915 struct inode *inode;
916 uint64_t uninitialized_var(sanity);
917 int ret;
918
919 inode = iget_locked(sb, ino);
920 if (!inode)
921 return ERR_PTR(-ENOMEM);
922 if (!(inode->i_state & I_NEW))
923 return inode;
924 oi = exofs_i(inode);
925
926 /* read the inode from the osd */
927 ret = exofs_get_inode(sb, oi, &fcb, &sanity);
928 if (ret)
929 goto bad_inode;
930
931 init_waitqueue_head(&oi->i_wq);
932 set_obj_created(oi);
933
934 /* copy stuff from on-disk struct to in-memory struct */
935 inode->i_mode = le16_to_cpu(fcb.i_mode);
936 inode->i_uid = le32_to_cpu(fcb.i_uid);
937 inode->i_gid = le32_to_cpu(fcb.i_gid);
938 inode->i_nlink = le16_to_cpu(fcb.i_links_count);
939 inode->i_ctime.tv_sec = (signed)le32_to_cpu(fcb.i_ctime);
940 inode->i_atime.tv_sec = (signed)le32_to_cpu(fcb.i_atime);
941 inode->i_mtime.tv_sec = (signed)le32_to_cpu(fcb.i_mtime);
942 inode->i_ctime.tv_nsec =
943 inode->i_atime.tv_nsec = inode->i_mtime.tv_nsec = 0;
944 oi->i_commit_size = le64_to_cpu(fcb.i_size);
945 i_size_write(inode, oi->i_commit_size);
946 inode->i_blkbits = EXOFS_BLKSHIFT;
947 inode->i_generation = le32_to_cpu(fcb.i_generation);
948
949#ifdef EXOFS_DEBUG_OBJ_ISIZE
950 if ((inode->i_size != sanity) &&
951 (!exofs_inode_is_fast_symlink(inode))) {
952 EXOFS_ERR("WARNING: Size of object from inode and "
953 "attributes differ (%lld != %llu)\n",
954 inode->i_size, _LLU(sanity));
955 }
956#endif
957
958 oi->i_dir_start_lookup = 0;
959
960 if ((inode->i_nlink == 0) && (inode->i_mode == 0)) {
961 ret = -ESTALE;
962 goto bad_inode;
963 }
964
965 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode)) {
966 if (fcb.i_data[0])
967 inode->i_rdev =
968 old_decode_dev(le32_to_cpu(fcb.i_data[0]));
969 else
970 inode->i_rdev =
971 new_decode_dev(le32_to_cpu(fcb.i_data[1]));
972 } else {
973 memcpy(oi->i_data, fcb.i_data, sizeof(fcb.i_data));
974 }
975
976 if (S_ISREG(inode->i_mode)) {
977 inode->i_op = &exofs_file_inode_operations;
978 inode->i_fop = &exofs_file_operations;
979 inode->i_mapping->a_ops = &exofs_aops;
980 } else if (S_ISDIR(inode->i_mode)) {
981 inode->i_op = &exofs_dir_inode_operations;
982 inode->i_fop = &exofs_dir_operations;
983 inode->i_mapping->a_ops = &exofs_aops;
984 } else if (S_ISLNK(inode->i_mode)) {
985 if (exofs_inode_is_fast_symlink(inode))
986 inode->i_op = &exofs_fast_symlink_inode_operations;
987 else {
988 inode->i_op = &exofs_symlink_inode_operations;
989 inode->i_mapping->a_ops = &exofs_aops;
990 }
991 } else {
992 inode->i_op = &exofs_special_inode_operations;
993 if (fcb.i_data[0])
994 init_special_inode(inode, inode->i_mode,
995 old_decode_dev(le32_to_cpu(fcb.i_data[0])));
996 else
997 init_special_inode(inode, inode->i_mode,
998 new_decode_dev(le32_to_cpu(fcb.i_data[1])));
999 }
1000
1001 unlock_new_inode(inode);
1002 return inode;
1003
1004bad_inode:
1005 iget_failed(inode);
1006 return ERR_PTR(ret);
1007}
1008
1009int __exofs_wait_obj_created(struct exofs_i_info *oi)
1010{
1011 if (!obj_created(oi)) {
1012 BUG_ON(!obj_2bcreated(oi));
1013 wait_event(oi->i_wq, obj_created(oi));
1014 }
1015 return unlikely(is_bad_inode(&oi->vfs_inode)) ? -EIO : 0;
1016}
1017/*
1018 * Callback function from exofs_new_inode(). The important thing is that we
1019 * set the obj_created flag so that other methods know that the object exists on
1020 * the OSD.
1021 */
1022static void create_done(struct osd_request *or, void *p)
1023{
1024 struct inode *inode = p;
1025 struct exofs_i_info *oi = exofs_i(inode);
1026 struct exofs_sb_info *sbi = inode->i_sb->s_fs_info;
1027 int ret;
1028
1029 ret = exofs_check_ok(or);
1030 osd_end_request(or);
1031 atomic_dec(&sbi->s_curr_pending);
1032
1033 if (unlikely(ret)) {
1034 EXOFS_ERR("object=0x%llx creation faild in pid=0x%llx",
1035 _LLU(sbi->s_pid), _LLU(inode->i_ino + EXOFS_OBJ_OFF));
1036 make_bad_inode(inode);
1037 } else
1038 set_obj_created(oi);
1039
1040 atomic_dec(&inode->i_count);
1041 wake_up(&oi->i_wq);
1042}
1043
1044/*
1045 * Set up a new inode and create an object for it on the OSD
1046 */
1047struct inode *exofs_new_inode(struct inode *dir, int mode)
1048{
1049 struct super_block *sb;
1050 struct inode *inode;
1051 struct exofs_i_info *oi;
1052 struct exofs_sb_info *sbi;
1053 struct osd_request *or;
1054 struct osd_obj_id obj;
1055 int ret;
1056
1057 sb = dir->i_sb;
1058 inode = new_inode(sb);
1059 if (!inode)
1060 return ERR_PTR(-ENOMEM);
1061
1062 oi = exofs_i(inode);
1063
1064 init_waitqueue_head(&oi->i_wq);
1065 set_obj_2bcreated(oi);
1066
1067 sbi = sb->s_fs_info;
1068
1069 sb->s_dirt = 1;
1070 inode->i_uid = current->cred->fsuid;
1071 if (dir->i_mode & S_ISGID) {
1072 inode->i_gid = dir->i_gid;
1073 if (S_ISDIR(mode))
1074 mode |= S_ISGID;
1075 } else {
1076 inode->i_gid = current->cred->fsgid;
1077 }
1078 inode->i_mode = mode;
1079
1080 inode->i_ino = sbi->s_nextid++;
1081 inode->i_blkbits = EXOFS_BLKSHIFT;
1082 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1083 oi->i_commit_size = inode->i_size = 0;
1084 spin_lock(&sbi->s_next_gen_lock);
1085 inode->i_generation = sbi->s_next_generation++;
1086 spin_unlock(&sbi->s_next_gen_lock);
1087 insert_inode_hash(inode);
1088
1089 mark_inode_dirty(inode);
1090
1091 obj.partition = sbi->s_pid;
1092 obj.id = inode->i_ino + EXOFS_OBJ_OFF;
1093 exofs_make_credential(oi->i_cred, &obj);
1094
1095 or = osd_start_request(sbi->s_dev, GFP_KERNEL);
1096 if (unlikely(!or)) {
1097 EXOFS_ERR("exofs_new_inode: osd_start_request failed\n");
1098 return ERR_PTR(-ENOMEM);
1099 }
1100
1101 osd_req_create_object(or, &obj);
1102
1103 /* increment the refcount so that the inode will still be around when we
1104 * reach the callback
1105 */
1106 atomic_inc(&inode->i_count);
1107
1108 ret = exofs_async_op(or, create_done, inode, oi->i_cred);
1109 if (ret) {
1110 atomic_dec(&inode->i_count);
1111 osd_end_request(or);
1112 return ERR_PTR(-EIO);
1113 }
1114 atomic_inc(&sbi->s_curr_pending);
1115
1116 return inode;
1117}
1118
1119/*
1120 * struct to pass two arguments to update_inode's callback
1121 */
1122struct updatei_args {
1123 struct exofs_sb_info *sbi;
1124 struct exofs_fcb fcb;
1125};
1126
1127/*
1128 * Callback function from exofs_update_inode().
1129 */
1130static void updatei_done(struct osd_request *or, void *p)
1131{
1132 struct updatei_args *args = p;
1133
1134 osd_end_request(or);
1135
1136 atomic_dec(&args->sbi->s_curr_pending);
1137
1138 kfree(args);
1139}
1140
1141/*
1142 * Write the inode to the OSD. Just fill up the struct, and set the attribute
1143 * synchronously or asynchronously depending on the do_sync flag.
1144 */
1145static int exofs_update_inode(struct inode *inode, int do_sync)
1146{
1147 struct exofs_i_info *oi = exofs_i(inode);
1148 struct super_block *sb = inode->i_sb;
1149 struct exofs_sb_info *sbi = sb->s_fs_info;
1150 struct osd_obj_id obj = {sbi->s_pid, inode->i_ino + EXOFS_OBJ_OFF};
1151 struct osd_request *or;
1152 struct osd_attr attr;
1153 struct exofs_fcb *fcb;
1154 struct updatei_args *args;
1155 int ret;
1156
1157 args = kzalloc(sizeof(*args), GFP_KERNEL);
1158 if (!args)
1159 return -ENOMEM;
1160
1161 fcb = &args->fcb;
1162
1163 fcb->i_mode = cpu_to_le16(inode->i_mode);
1164 fcb->i_uid = cpu_to_le32(inode->i_uid);
1165 fcb->i_gid = cpu_to_le32(inode->i_gid);
1166 fcb->i_links_count = cpu_to_le16(inode->i_nlink);
1167 fcb->i_ctime = cpu_to_le32(inode->i_ctime.tv_sec);
1168 fcb->i_atime = cpu_to_le32(inode->i_atime.tv_sec);
1169 fcb->i_mtime = cpu_to_le32(inode->i_mtime.tv_sec);
1170 oi->i_commit_size = i_size_read(inode);
1171 fcb->i_size = cpu_to_le64(oi->i_commit_size);
1172 fcb->i_generation = cpu_to_le32(inode->i_generation);
1173
1174 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode)) {
1175 if (old_valid_dev(inode->i_rdev)) {
1176 fcb->i_data[0] =
1177 cpu_to_le32(old_encode_dev(inode->i_rdev));
1178 fcb->i_data[1] = 0;
1179 } else {
1180 fcb->i_data[0] = 0;
1181 fcb->i_data[1] =
1182 cpu_to_le32(new_encode_dev(inode->i_rdev));
1183 fcb->i_data[2] = 0;
1184 }
1185 } else
1186 memcpy(fcb->i_data, oi->i_data, sizeof(fcb->i_data));
1187
1188 or = osd_start_request(sbi->s_dev, GFP_KERNEL);
1189 if (unlikely(!or)) {
1190 EXOFS_ERR("exofs_update_inode: osd_start_request failed.\n");
1191 ret = -ENOMEM;
1192 goto free_args;
1193 }
1194
1195 osd_req_set_attributes(or, &obj);
1196
1197 attr = g_attr_inode_data;
1198 attr.val_ptr = fcb;
1199 osd_req_add_set_attr_list(or, &attr, 1);
1200
1201 if (!obj_created(oi)) {
1202 EXOFS_DBGMSG("!obj_created\n");
1203 BUG_ON(!obj_2bcreated(oi));
1204 wait_event(oi->i_wq, obj_created(oi));
1205 EXOFS_DBGMSG("wait_event done\n");
1206 }
1207
1208 if (do_sync) {
1209 ret = exofs_sync_op(or, sbi->s_timeout, oi->i_cred);
1210 osd_end_request(or);
1211 goto free_args;
1212 } else {
1213 args->sbi = sbi;
1214
1215 ret = exofs_async_op(or, updatei_done, args, oi->i_cred);
1216 if (ret) {
1217 osd_end_request(or);
1218 goto free_args;
1219 }
1220 atomic_inc(&sbi->s_curr_pending);
1221 goto out; /* deallocation in updatei_done */
1222 }
1223
1224free_args:
1225 kfree(args);
1226out:
1227 EXOFS_DBGMSG("ret=>%d\n", ret);
1228 return ret;
1229}
1230
1231int exofs_write_inode(struct inode *inode, int wait)
1232{
1233 return exofs_update_inode(inode, wait);
1234}
1235
1236/*
1237 * Callback function from exofs_delete_inode() - don't have much cleaning up to
1238 * do.
1239 */
1240static void delete_done(struct osd_request *or, void *p)
1241{
1242 struct exofs_sb_info *sbi;
1243 osd_end_request(or);
1244 sbi = p;
1245 atomic_dec(&sbi->s_curr_pending);
1246}
1247
1248/*
1249 * Called when the refcount of an inode reaches zero. We remove the object
1250 * from the OSD here. We make sure the object was created before we try and
1251 * delete it.
1252 */
1253void exofs_delete_inode(struct inode *inode)
1254{
1255 struct exofs_i_info *oi = exofs_i(inode);
1256 struct super_block *sb = inode->i_sb;
1257 struct exofs_sb_info *sbi = sb->s_fs_info;
1258 struct osd_obj_id obj = {sbi->s_pid, inode->i_ino + EXOFS_OBJ_OFF};
1259 struct osd_request *or;
1260 int ret;
1261
1262 truncate_inode_pages(&inode->i_data, 0);
1263
1264 if (is_bad_inode(inode))
1265 goto no_delete;
1266
1267 mark_inode_dirty(inode);
1268 exofs_update_inode(inode, inode_needs_sync(inode));
1269
1270 inode->i_size = 0;
1271 if (inode->i_blocks)
1272 exofs_truncate(inode);
1273
1274 clear_inode(inode);
1275
1276 or = osd_start_request(sbi->s_dev, GFP_KERNEL);
1277 if (unlikely(!or)) {
1278 EXOFS_ERR("exofs_delete_inode: osd_start_request failed\n");
1279 return;
1280 }
1281
1282 osd_req_remove_object(or, &obj);
1283
1284 /* if we are deleting an obj that hasn't been created yet, wait */
1285 if (!obj_created(oi)) {
1286 BUG_ON(!obj_2bcreated(oi));
1287 wait_event(oi->i_wq, obj_created(oi));
1288 }
1289
1290 ret = exofs_async_op(or, delete_done, sbi, oi->i_cred);
1291 if (ret) {
1292 EXOFS_ERR(
1293 "ERROR: @exofs_delete_inode exofs_async_op failed\n");
1294 osd_end_request(or);
1295 return;
1296 }
1297 atomic_inc(&sbi->s_curr_pending);
1298
1299 return;
1300
1301no_delete:
1302 clear_inode(inode);
1303}
diff --git a/fs/exofs/namei.c b/fs/exofs/namei.c
new file mode 100644
index 000000000000..77fdd765e76d
--- /dev/null
+++ b/fs/exofs/namei.c
@@ -0,0 +1,342 @@
1/*
2 * Copyright (C) 2005, 2006
3 * Avishay Traeger (avishay@gmail.com) (avishay@il.ibm.com)
4 * Copyright (C) 2005, 2006
5 * International Business Machines
6 * Copyright (C) 2008, 2009
7 * Boaz Harrosh <bharrosh@panasas.com>
8 *
9 * Copyrights for code taken from ext2:
10 * Copyright (C) 1992, 1993, 1994, 1995
11 * Remy Card (card@masi.ibp.fr)
12 * Laboratoire MASI - Institut Blaise Pascal
13 * Universite Pierre et Marie Curie (Paris VI)
14 * from
15 * linux/fs/minix/inode.c
16 * Copyright (C) 1991, 1992 Linus Torvalds
17 *
18 * This file is part of exofs.
19 *
20 * exofs is free software; you can redistribute it and/or modify
21 * it under the terms of the GNU General Public License as published by
22 * the Free Software Foundation. Since it is based on ext2, and the only
23 * valid version of GPL for the Linux kernel is version 2, the only valid
24 * version of GPL for exofs is version 2.
25 *
26 * exofs is distributed in the hope that it will be useful,
27 * but WITHOUT ANY WARRANTY; without even the implied warranty of
28 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
29 * GNU General Public License for more details.
30 *
31 * You should have received a copy of the GNU General Public License
32 * along with exofs; if not, write to the Free Software
33 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
34 */
35
36#include "exofs.h"
37
38static inline int exofs_add_nondir(struct dentry *dentry, struct inode *inode)
39{
40 int err = exofs_add_link(dentry, inode);
41 if (!err) {
42 d_instantiate(dentry, inode);
43 return 0;
44 }
45 inode_dec_link_count(inode);
46 iput(inode);
47 return err;
48}
49
50static struct dentry *exofs_lookup(struct inode *dir, struct dentry *dentry,
51 struct nameidata *nd)
52{
53 struct inode *inode;
54 ino_t ino;
55
56 if (dentry->d_name.len > EXOFS_NAME_LEN)
57 return ERR_PTR(-ENAMETOOLONG);
58
59 ino = exofs_inode_by_name(dir, dentry);
60 inode = NULL;
61 if (ino) {
62 inode = exofs_iget(dir->i_sb, ino);
63 if (IS_ERR(inode))
64 return ERR_CAST(inode);
65 }
66 return d_splice_alias(inode, dentry);
67}
68
69static int exofs_create(struct inode *dir, struct dentry *dentry, int mode,
70 struct nameidata *nd)
71{
72 struct inode *inode = exofs_new_inode(dir, mode);
73 int err = PTR_ERR(inode);
74 if (!IS_ERR(inode)) {
75 inode->i_op = &exofs_file_inode_operations;
76 inode->i_fop = &exofs_file_operations;
77 inode->i_mapping->a_ops = &exofs_aops;
78 mark_inode_dirty(inode);
79 err = exofs_add_nondir(dentry, inode);
80 }
81 return err;
82}
83
84static int exofs_mknod(struct inode *dir, struct dentry *dentry, int mode,
85 dev_t rdev)
86{
87 struct inode *inode;
88 int err;
89
90 if (!new_valid_dev(rdev))
91 return -EINVAL;
92
93 inode = exofs_new_inode(dir, mode);
94 err = PTR_ERR(inode);
95 if (!IS_ERR(inode)) {
96 init_special_inode(inode, inode->i_mode, rdev);
97 mark_inode_dirty(inode);
98 err = exofs_add_nondir(dentry, inode);
99 }
100 return err;
101}
102
103static int exofs_symlink(struct inode *dir, struct dentry *dentry,
104 const char *symname)
105{
106 struct super_block *sb = dir->i_sb;
107 int err = -ENAMETOOLONG;
108 unsigned l = strlen(symname)+1;
109 struct inode *inode;
110 struct exofs_i_info *oi;
111
112 if (l > sb->s_blocksize)
113 goto out;
114
115 inode = exofs_new_inode(dir, S_IFLNK | S_IRWXUGO);
116 err = PTR_ERR(inode);
117 if (IS_ERR(inode))
118 goto out;
119
120 oi = exofs_i(inode);
121 if (l > sizeof(oi->i_data)) {
122 /* slow symlink */
123 inode->i_op = &exofs_symlink_inode_operations;
124 inode->i_mapping->a_ops = &exofs_aops;
125 memset(oi->i_data, 0, sizeof(oi->i_data));
126
127 err = page_symlink(inode, symname, l);
128 if (err)
129 goto out_fail;
130 } else {
131 /* fast symlink */
132 inode->i_op = &exofs_fast_symlink_inode_operations;
133 memcpy(oi->i_data, symname, l);
134 inode->i_size = l-1;
135 }
136 mark_inode_dirty(inode);
137
138 err = exofs_add_nondir(dentry, inode);
139out:
140 return err;
141
142out_fail:
143 inode_dec_link_count(inode);
144 iput(inode);
145 goto out;
146}
147
148static int exofs_link(struct dentry *old_dentry, struct inode *dir,
149 struct dentry *dentry)
150{
151 struct inode *inode = old_dentry->d_inode;
152
153 if (inode->i_nlink >= EXOFS_LINK_MAX)
154 return -EMLINK;
155
156 inode->i_ctime = CURRENT_TIME;
157 inode_inc_link_count(inode);
158 atomic_inc(&inode->i_count);
159
160 return exofs_add_nondir(dentry, inode);
161}
162
163static int exofs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
164{
165 struct inode *inode;
166 int err = -EMLINK;
167
168 if (dir->i_nlink >= EXOFS_LINK_MAX)
169 goto out;
170
171 inode_inc_link_count(dir);
172
173 inode = exofs_new_inode(dir, S_IFDIR | mode);
174 err = PTR_ERR(inode);
175 if (IS_ERR(inode))
176 goto out_dir;
177
178 inode->i_op = &exofs_dir_inode_operations;
179 inode->i_fop = &exofs_dir_operations;
180 inode->i_mapping->a_ops = &exofs_aops;
181
182 inode_inc_link_count(inode);
183
184 err = exofs_make_empty(inode, dir);
185 if (err)
186 goto out_fail;
187
188 err = exofs_add_link(dentry, inode);
189 if (err)
190 goto out_fail;
191
192 d_instantiate(dentry, inode);
193out:
194 return err;
195
196out_fail:
197 inode_dec_link_count(inode);
198 inode_dec_link_count(inode);
199 iput(inode);
200out_dir:
201 inode_dec_link_count(dir);
202 goto out;
203}
204
205static int exofs_unlink(struct inode *dir, struct dentry *dentry)
206{
207 struct inode *inode = dentry->d_inode;
208 struct exofs_dir_entry *de;
209 struct page *page;
210 int err = -ENOENT;
211
212 de = exofs_find_entry(dir, dentry, &page);
213 if (!de)
214 goto out;
215
216 err = exofs_delete_entry(de, page);
217 if (err)
218 goto out;
219
220 inode->i_ctime = dir->i_ctime;
221 inode_dec_link_count(inode);
222 err = 0;
223out:
224 return err;
225}
226
227static int exofs_rmdir(struct inode *dir, struct dentry *dentry)
228{
229 struct inode *inode = dentry->d_inode;
230 int err = -ENOTEMPTY;
231
232 if (exofs_empty_dir(inode)) {
233 err = exofs_unlink(dir, dentry);
234 if (!err) {
235 inode->i_size = 0;
236 inode_dec_link_count(inode);
237 inode_dec_link_count(dir);
238 }
239 }
240 return err;
241}
242
243static int exofs_rename(struct inode *old_dir, struct dentry *old_dentry,
244 struct inode *new_dir, struct dentry *new_dentry)
245{
246 struct inode *old_inode = old_dentry->d_inode;
247 struct inode *new_inode = new_dentry->d_inode;
248 struct page *dir_page = NULL;
249 struct exofs_dir_entry *dir_de = NULL;
250 struct page *old_page;
251 struct exofs_dir_entry *old_de;
252 int err = -ENOENT;
253
254 old_de = exofs_find_entry(old_dir, old_dentry, &old_page);
255 if (!old_de)
256 goto out;
257
258 if (S_ISDIR(old_inode->i_mode)) {
259 err = -EIO;
260 dir_de = exofs_dotdot(old_inode, &dir_page);
261 if (!dir_de)
262 goto out_old;
263 }
264
265 if (new_inode) {
266 struct page *new_page;
267 struct exofs_dir_entry *new_de;
268
269 err = -ENOTEMPTY;
270 if (dir_de && !exofs_empty_dir(new_inode))
271 goto out_dir;
272
273 err = -ENOENT;
274 new_de = exofs_find_entry(new_dir, new_dentry, &new_page);
275 if (!new_de)
276 goto out_dir;
277 inode_inc_link_count(old_inode);
278 err = exofs_set_link(new_dir, new_de, new_page, old_inode);
279 new_inode->i_ctime = CURRENT_TIME;
280 if (dir_de)
281 drop_nlink(new_inode);
282 inode_dec_link_count(new_inode);
283 if (err)
284 goto out_dir;
285 } else {
286 if (dir_de) {
287 err = -EMLINK;
288 if (new_dir->i_nlink >= EXOFS_LINK_MAX)
289 goto out_dir;
290 }
291 inode_inc_link_count(old_inode);
292 err = exofs_add_link(new_dentry, old_inode);
293 if (err) {
294 inode_dec_link_count(old_inode);
295 goto out_dir;
296 }
297 if (dir_de)
298 inode_inc_link_count(new_dir);
299 }
300
301 old_inode->i_ctime = CURRENT_TIME;
302
303 exofs_delete_entry(old_de, old_page);
304 inode_dec_link_count(old_inode);
305
306 if (dir_de) {
307 err = exofs_set_link(old_inode, dir_de, dir_page, new_dir);
308 inode_dec_link_count(old_dir);
309 if (err)
310 goto out_dir;
311 }
312 return 0;
313
314
315out_dir:
316 if (dir_de) {
317 kunmap(dir_page);
318 page_cache_release(dir_page);
319 }
320out_old:
321 kunmap(old_page);
322 page_cache_release(old_page);
323out:
324 return err;
325}
326
327const struct inode_operations exofs_dir_inode_operations = {
328 .create = exofs_create,
329 .lookup = exofs_lookup,
330 .link = exofs_link,
331 .unlink = exofs_unlink,
332 .symlink = exofs_symlink,
333 .mkdir = exofs_mkdir,
334 .rmdir = exofs_rmdir,
335 .mknod = exofs_mknod,
336 .rename = exofs_rename,
337 .setattr = exofs_setattr,
338};
339
340const struct inode_operations exofs_special_inode_operations = {
341 .setattr = exofs_setattr,
342};
diff --git a/fs/exofs/osd.c b/fs/exofs/osd.c
new file mode 100644
index 000000000000..b249ae97fb15
--- /dev/null
+++ b/fs/exofs/osd.c
@@ -0,0 +1,153 @@
1/*
2 * Copyright (C) 2005, 2006
3 * Avishay Traeger (avishay@gmail.com) (avishay@il.ibm.com)
4 * Copyright (C) 2005, 2006
5 * International Business Machines
6 * Copyright (C) 2008, 2009
7 * Boaz Harrosh <bharrosh@panasas.com>
8 *
9 * This file is part of exofs.
10 *
11 * exofs is free software; you can redistribute it and/or modify
12 * it under the terms of the GNU General Public License as published by
13 * the Free Software Foundation. Since it is based on ext2, and the only
14 * valid version of GPL for the Linux kernel is version 2, the only valid
15 * version of GPL for exofs is version 2.
16 *
17 * exofs is distributed in the hope that it will be useful,
18 * but WITHOUT ANY WARRANTY; without even the implied warranty of
19 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
20 * GNU General Public License for more details.
21 *
22 * You should have received a copy of the GNU General Public License
23 * along with exofs; if not, write to the Free Software
24 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
25 */
26
27#include <scsi/scsi_device.h>
28#include <scsi/osd_sense.h>
29
30#include "exofs.h"
31
32int exofs_check_ok_resid(struct osd_request *or, u64 *in_resid, u64 *out_resid)
33{
34 struct osd_sense_info osi;
35 int ret = osd_req_decode_sense(or, &osi);
36
37 if (ret) { /* translate to Linux codes */
38 if (osi.additional_code == scsi_invalid_field_in_cdb) {
39 if (osi.cdb_field_offset == OSD_CFO_STARTING_BYTE)
40 ret = -EFAULT;
41 if (osi.cdb_field_offset == OSD_CFO_OBJECT_ID)
42 ret = -ENOENT;
43 else
44 ret = -EINVAL;
45 } else if (osi.additional_code == osd_quota_error)
46 ret = -ENOSPC;
47 else
48 ret = -EIO;
49 }
50
51 /* FIXME: should be include in osd_sense_info */
52 if (in_resid)
53 *in_resid = or->in.req ? or->in.req->data_len : 0;
54
55 if (out_resid)
56 *out_resid = or->out.req ? or->out.req->data_len : 0;
57
58 return ret;
59}
60
61void exofs_make_credential(u8 cred_a[OSD_CAP_LEN], const struct osd_obj_id *obj)
62{
63 osd_sec_init_nosec_doall_caps(cred_a, obj, false, true);
64}
65
66/*
67 * Perform a synchronous OSD operation.
68 */
69int exofs_sync_op(struct osd_request *or, int timeout, uint8_t *credential)
70{
71 int ret;
72
73 or->timeout = timeout;
74 ret = osd_finalize_request(or, 0, credential, NULL);
75 if (ret) {
76 EXOFS_DBGMSG("Faild to osd_finalize_request() => %d\n", ret);
77 return ret;
78 }
79
80 ret = osd_execute_request(or);
81
82 if (ret)
83 EXOFS_DBGMSG("osd_execute_request() => %d\n", ret);
84 /* osd_req_decode_sense(or, ret); */
85 return ret;
86}
87
88/*
89 * Perform an asynchronous OSD operation.
90 */
91int exofs_async_op(struct osd_request *or, osd_req_done_fn *async_done,
92 void *caller_context, u8 *cred)
93{
94 int ret;
95
96 ret = osd_finalize_request(or, 0, cred, NULL);
97 if (ret) {
98 EXOFS_DBGMSG("Faild to osd_finalize_request() => %d\n", ret);
99 return ret;
100 }
101
102 ret = osd_execute_request_async(or, async_done, caller_context);
103
104 if (ret)
105 EXOFS_DBGMSG("osd_execute_request_async() => %d\n", ret);
106 return ret;
107}
108
109int extract_attr_from_req(struct osd_request *or, struct osd_attr *attr)
110{
111 struct osd_attr cur_attr = {.attr_page = 0}; /* start with zeros */
112 void *iter = NULL;
113 int nelem;
114
115 do {
116 nelem = 1;
117 osd_req_decode_get_attr_list(or, &cur_attr, &nelem, &iter);
118 if ((cur_attr.attr_page == attr->attr_page) &&
119 (cur_attr.attr_id == attr->attr_id)) {
120 attr->len = cur_attr.len;
121 attr->val_ptr = cur_attr.val_ptr;
122 return 0;
123 }
124 } while (iter);
125
126 return -EIO;
127}
128
129int osd_req_read_kern(struct osd_request *or,
130 const struct osd_obj_id *obj, u64 offset, void* buff, u64 len)
131{
132 struct request_queue *req_q = or->osd_dev->scsi_device->request_queue;
133 struct bio *bio = bio_map_kern(req_q, buff, len, GFP_KERNEL);
134
135 if (!bio)
136 return -ENOMEM;
137
138 osd_req_read(or, obj, bio, offset);
139 return 0;
140}
141
142int osd_req_write_kern(struct osd_request *or,
143 const struct osd_obj_id *obj, u64 offset, void* buff, u64 len)
144{
145 struct request_queue *req_q = or->osd_dev->scsi_device->request_queue;
146 struct bio *bio = bio_map_kern(req_q, buff, len, GFP_KERNEL);
147
148 if (!bio)
149 return -ENOMEM;
150
151 osd_req_write(or, obj, bio, offset);
152 return 0;
153}
diff --git a/fs/exofs/super.c b/fs/exofs/super.c
new file mode 100644
index 000000000000..9f1985e857e2
--- /dev/null
+++ b/fs/exofs/super.c
@@ -0,0 +1,584 @@
1/*
2 * Copyright (C) 2005, 2006
3 * Avishay Traeger (avishay@gmail.com) (avishay@il.ibm.com)
4 * Copyright (C) 2005, 2006
5 * International Business Machines
6 * Copyright (C) 2008, 2009
7 * Boaz Harrosh <bharrosh@panasas.com>
8 *
9 * Copyrights for code taken from ext2:
10 * Copyright (C) 1992, 1993, 1994, 1995
11 * Remy Card (card@masi.ibp.fr)
12 * Laboratoire MASI - Institut Blaise Pascal
13 * Universite Pierre et Marie Curie (Paris VI)
14 * from
15 * linux/fs/minix/inode.c
16 * Copyright (C) 1991, 1992 Linus Torvalds
17 *
18 * This file is part of exofs.
19 *
20 * exofs is free software; you can redistribute it and/or modify
21 * it under the terms of the GNU General Public License as published by
22 * the Free Software Foundation. Since it is based on ext2, and the only
23 * valid version of GPL for the Linux kernel is version 2, the only valid
24 * version of GPL for exofs is version 2.
25 *
26 * exofs is distributed in the hope that it will be useful,
27 * but WITHOUT ANY WARRANTY; without even the implied warranty of
28 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
29 * GNU General Public License for more details.
30 *
31 * You should have received a copy of the GNU General Public License
32 * along with exofs; if not, write to the Free Software
33 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
34 */
35
36#include <linux/string.h>
37#include <linux/parser.h>
38#include <linux/vfs.h>
39#include <linux/random.h>
40#include <linux/exportfs.h>
41
42#include "exofs.h"
43
44/******************************************************************************
45 * MOUNT OPTIONS
46 *****************************************************************************/
47
48/*
49 * struct to hold what we get from mount options
50 */
51struct exofs_mountopt {
52 const char *dev_name;
53 uint64_t pid;
54 int timeout;
55};
56
57/*
58 * exofs-specific mount-time options.
59 */
60enum { Opt_pid, Opt_to, Opt_mkfs, Opt_format, Opt_err };
61
62/*
63 * Our mount-time options. These should ideally be 64-bit unsigned, but the
64 * kernel's parsing functions do not currently support that. 32-bit should be
65 * sufficient for most applications now.
66 */
67static match_table_t tokens = {
68 {Opt_pid, "pid=%u"},
69 {Opt_to, "to=%u"},
70 {Opt_err, NULL}
71};
72
73/*
74 * The main option parsing method. Also makes sure that all of the mandatory
75 * mount options were set.
76 */
77static int parse_options(char *options, struct exofs_mountopt *opts)
78{
79 char *p;
80 substring_t args[MAX_OPT_ARGS];
81 int option;
82 bool s_pid = false;
83
84 EXOFS_DBGMSG("parse_options %s\n", options);
85 /* defaults */
86 memset(opts, 0, sizeof(*opts));
87 opts->timeout = BLK_DEFAULT_SG_TIMEOUT;
88
89 while ((p = strsep(&options, ",")) != NULL) {
90 int token;
91 char str[32];
92
93 if (!*p)
94 continue;
95
96 token = match_token(p, tokens, args);
97 switch (token) {
98 case Opt_pid:
99 if (0 == match_strlcpy(str, &args[0], sizeof(str)))
100 return -EINVAL;
101 opts->pid = simple_strtoull(str, NULL, 0);
102 if (opts->pid < EXOFS_MIN_PID) {
103 EXOFS_ERR("Partition ID must be >= %u",
104 EXOFS_MIN_PID);
105 return -EINVAL;
106 }
107 s_pid = 1;
108 break;
109 case Opt_to:
110 if (match_int(&args[0], &option))
111 return -EINVAL;
112 if (option <= 0) {
113 EXOFS_ERR("Timout must be > 0");
114 return -EINVAL;
115 }
116 opts->timeout = option * HZ;
117 break;
118 }
119 }
120
121 if (!s_pid) {
122 EXOFS_ERR("Need to specify the following options:\n");
123 EXOFS_ERR(" -o pid=pid_no_to_use\n");
124 return -EINVAL;
125 }
126
127 return 0;
128}
129
130/******************************************************************************
131 * INODE CACHE
132 *****************************************************************************/
133
134/*
135 * Our inode cache. Isn't it pretty?
136 */
137static struct kmem_cache *exofs_inode_cachep;
138
139/*
140 * Allocate an inode in the cache
141 */
142static struct inode *exofs_alloc_inode(struct super_block *sb)
143{
144 struct exofs_i_info *oi;
145
146 oi = kmem_cache_alloc(exofs_inode_cachep, GFP_KERNEL);
147 if (!oi)
148 return NULL;
149
150 oi->vfs_inode.i_version = 1;
151 return &oi->vfs_inode;
152}
153
154/*
155 * Remove an inode from the cache
156 */
157static void exofs_destroy_inode(struct inode *inode)
158{
159 kmem_cache_free(exofs_inode_cachep, exofs_i(inode));
160}
161
162/*
163 * Initialize the inode
164 */
165static void exofs_init_once(void *foo)
166{
167 struct exofs_i_info *oi = foo;
168
169 inode_init_once(&oi->vfs_inode);
170}
171
172/*
173 * Create and initialize the inode cache
174 */
175static int init_inodecache(void)
176{
177 exofs_inode_cachep = kmem_cache_create("exofs_inode_cache",
178 sizeof(struct exofs_i_info), 0,
179 SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD,
180 exofs_init_once);
181 if (exofs_inode_cachep == NULL)
182 return -ENOMEM;
183 return 0;
184}
185
186/*
187 * Destroy the inode cache
188 */
189static void destroy_inodecache(void)
190{
191 kmem_cache_destroy(exofs_inode_cachep);
192}
193
194/******************************************************************************
195 * SUPERBLOCK FUNCTIONS
196 *****************************************************************************/
197static const struct super_operations exofs_sops;
198static const struct export_operations exofs_export_ops;
199
200/*
201 * Write the superblock to the OSD
202 */
203static void exofs_write_super(struct super_block *sb)
204{
205 struct exofs_sb_info *sbi;
206 struct exofs_fscb *fscb;
207 struct osd_request *or;
208 struct osd_obj_id obj;
209 int ret;
210
211 fscb = kzalloc(sizeof(struct exofs_fscb), GFP_KERNEL);
212 if (!fscb) {
213 EXOFS_ERR("exofs_write_super: memory allocation failed.\n");
214 return;
215 }
216
217 lock_kernel();
218 sbi = sb->s_fs_info;
219 fscb->s_nextid = cpu_to_le64(sbi->s_nextid);
220 fscb->s_numfiles = cpu_to_le32(sbi->s_numfiles);
221 fscb->s_magic = cpu_to_le16(sb->s_magic);
222 fscb->s_newfs = 0;
223
224 or = osd_start_request(sbi->s_dev, GFP_KERNEL);
225 if (unlikely(!or)) {
226 EXOFS_ERR("exofs_write_super: osd_start_request failed.\n");
227 goto out;
228 }
229
230 obj.partition = sbi->s_pid;
231 obj.id = EXOFS_SUPER_ID;
232 ret = osd_req_write_kern(or, &obj, 0, fscb, sizeof(*fscb));
233 if (unlikely(ret)) {
234 EXOFS_ERR("exofs_write_super: osd_req_write_kern failed.\n");
235 goto out;
236 }
237
238 ret = exofs_sync_op(or, sbi->s_timeout, sbi->s_cred);
239 if (unlikely(ret)) {
240 EXOFS_ERR("exofs_write_super: exofs_sync_op failed.\n");
241 goto out;
242 }
243 sb->s_dirt = 0;
244
245out:
246 if (or)
247 osd_end_request(or);
248 unlock_kernel();
249 kfree(fscb);
250}
251
252/*
253 * This function is called when the vfs is freeing the superblock. We just
254 * need to free our own part.
255 */
256static void exofs_put_super(struct super_block *sb)
257{
258 int num_pend;
259 struct exofs_sb_info *sbi = sb->s_fs_info;
260
261 /* make sure there are no pending commands */
262 for (num_pend = atomic_read(&sbi->s_curr_pending); num_pend > 0;
263 num_pend = atomic_read(&sbi->s_curr_pending)) {
264 wait_queue_head_t wq;
265 init_waitqueue_head(&wq);
266 wait_event_timeout(wq,
267 (atomic_read(&sbi->s_curr_pending) == 0),
268 msecs_to_jiffies(100));
269 }
270
271 osduld_put_device(sbi->s_dev);
272 kfree(sb->s_fs_info);
273 sb->s_fs_info = NULL;
274}
275
276/*
277 * Read the superblock from the OSD and fill in the fields
278 */
279static int exofs_fill_super(struct super_block *sb, void *data, int silent)
280{
281 struct inode *root;
282 struct exofs_mountopt *opts = data;
283 struct exofs_sb_info *sbi; /*extended info */
284 struct exofs_fscb fscb; /*on-disk superblock info */
285 struct osd_request *or = NULL;
286 struct osd_obj_id obj;
287 int ret;
288
289 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
290 if (!sbi)
291 return -ENOMEM;
292 sb->s_fs_info = sbi;
293
294 /* use mount options to fill superblock */
295 sbi->s_dev = osduld_path_lookup(opts->dev_name);
296 if (IS_ERR(sbi->s_dev)) {
297 ret = PTR_ERR(sbi->s_dev);
298 sbi->s_dev = NULL;
299 goto free_sbi;
300 }
301
302 sbi->s_pid = opts->pid;
303 sbi->s_timeout = opts->timeout;
304
305 /* fill in some other data by hand */
306 memset(sb->s_id, 0, sizeof(sb->s_id));
307 strcpy(sb->s_id, "exofs");
308 sb->s_blocksize = EXOFS_BLKSIZE;
309 sb->s_blocksize_bits = EXOFS_BLKSHIFT;
310 sb->s_maxbytes = MAX_LFS_FILESIZE;
311 atomic_set(&sbi->s_curr_pending, 0);
312 sb->s_bdev = NULL;
313 sb->s_dev = 0;
314
315 /* read data from on-disk superblock object */
316 obj.partition = sbi->s_pid;
317 obj.id = EXOFS_SUPER_ID;
318 exofs_make_credential(sbi->s_cred, &obj);
319
320 or = osd_start_request(sbi->s_dev, GFP_KERNEL);
321 if (unlikely(!or)) {
322 if (!silent)
323 EXOFS_ERR(
324 "exofs_fill_super: osd_start_request failed.\n");
325 ret = -ENOMEM;
326 goto free_sbi;
327 }
328 ret = osd_req_read_kern(or, &obj, 0, &fscb, sizeof(fscb));
329 if (unlikely(ret)) {
330 if (!silent)
331 EXOFS_ERR(
332 "exofs_fill_super: osd_req_read_kern failed.\n");
333 ret = -ENOMEM;
334 goto free_sbi;
335 }
336
337 ret = exofs_sync_op(or, sbi->s_timeout, sbi->s_cred);
338 if (unlikely(ret)) {
339 if (!silent)
340 EXOFS_ERR("exofs_fill_super: exofs_sync_op failed.\n");
341 ret = -EIO;
342 goto free_sbi;
343 }
344
345 sb->s_magic = le16_to_cpu(fscb.s_magic);
346 sbi->s_nextid = le64_to_cpu(fscb.s_nextid);
347 sbi->s_numfiles = le32_to_cpu(fscb.s_numfiles);
348
349 /* make sure what we read from the object store is correct */
350 if (sb->s_magic != EXOFS_SUPER_MAGIC) {
351 if (!silent)
352 EXOFS_ERR("ERROR: Bad magic value\n");
353 ret = -EINVAL;
354 goto free_sbi;
355 }
356
357 /* start generation numbers from a random point */
358 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
359 spin_lock_init(&sbi->s_next_gen_lock);
360
361 /* set up operation vectors */
362 sb->s_op = &exofs_sops;
363 sb->s_export_op = &exofs_export_ops;
364 root = exofs_iget(sb, EXOFS_ROOT_ID - EXOFS_OBJ_OFF);
365 if (IS_ERR(root)) {
366 EXOFS_ERR("ERROR: exofs_iget failed\n");
367 ret = PTR_ERR(root);
368 goto free_sbi;
369 }
370 sb->s_root = d_alloc_root(root);
371 if (!sb->s_root) {
372 iput(root);
373 EXOFS_ERR("ERROR: get root inode failed\n");
374 ret = -ENOMEM;
375 goto free_sbi;
376 }
377
378 if (!S_ISDIR(root->i_mode)) {
379 dput(sb->s_root);
380 sb->s_root = NULL;
381 EXOFS_ERR("ERROR: corrupt root inode (mode = %hd)\n",
382 root->i_mode);
383 ret = -EINVAL;
384 goto free_sbi;
385 }
386
387 ret = 0;
388out:
389 if (or)
390 osd_end_request(or);
391 return ret;
392
393free_sbi:
394 osduld_put_device(sbi->s_dev); /* NULL safe */
395 kfree(sbi);
396 goto out;
397}
398
399/*
400 * Set up the superblock (calls exofs_fill_super eventually)
401 */
402static int exofs_get_sb(struct file_system_type *type,
403 int flags, const char *dev_name,
404 void *data, struct vfsmount *mnt)
405{
406 struct exofs_mountopt opts;
407 int ret;
408
409 ret = parse_options(data, &opts);
410 if (ret)
411 return ret;
412
413 opts.dev_name = dev_name;
414 return get_sb_nodev(type, flags, &opts, exofs_fill_super, mnt);
415}
416
417/*
418 * Return information about the file system state in the buffer. This is used
419 * by the 'df' command, for example.
420 */
421static int exofs_statfs(struct dentry *dentry, struct kstatfs *buf)
422{
423 struct super_block *sb = dentry->d_sb;
424 struct exofs_sb_info *sbi = sb->s_fs_info;
425 struct osd_obj_id obj = {sbi->s_pid, 0};
426 struct osd_attr attrs[] = {
427 ATTR_DEF(OSD_APAGE_PARTITION_QUOTAS,
428 OSD_ATTR_PQ_CAPACITY_QUOTA, sizeof(__be64)),
429 ATTR_DEF(OSD_APAGE_PARTITION_INFORMATION,
430 OSD_ATTR_PI_USED_CAPACITY, sizeof(__be64)),
431 };
432 uint64_t capacity = ULLONG_MAX;
433 uint64_t used = ULLONG_MAX;
434 struct osd_request *or;
435 uint8_t cred_a[OSD_CAP_LEN];
436 int ret;
437
438 /* get used/capacity attributes */
439 exofs_make_credential(cred_a, &obj);
440
441 or = osd_start_request(sbi->s_dev, GFP_KERNEL);
442 if (unlikely(!or)) {
443 EXOFS_DBGMSG("exofs_statfs: osd_start_request failed.\n");
444 return -ENOMEM;
445 }
446
447 osd_req_get_attributes(or, &obj);
448 osd_req_add_get_attr_list(or, attrs, ARRAY_SIZE(attrs));
449 ret = exofs_sync_op(or, sbi->s_timeout, cred_a);
450 if (unlikely(ret))
451 goto out;
452
453 ret = extract_attr_from_req(or, &attrs[0]);
454 if (likely(!ret))
455 capacity = get_unaligned_be64(attrs[0].val_ptr);
456 else
457 EXOFS_DBGMSG("exofs_statfs: get capacity failed.\n");
458
459 ret = extract_attr_from_req(or, &attrs[1]);
460 if (likely(!ret))
461 used = get_unaligned_be64(attrs[1].val_ptr);
462 else
463 EXOFS_DBGMSG("exofs_statfs: get used-space failed.\n");
464
465 /* fill in the stats buffer */
466 buf->f_type = EXOFS_SUPER_MAGIC;
467 buf->f_bsize = EXOFS_BLKSIZE;
468 buf->f_blocks = (capacity >> EXOFS_BLKSHIFT);
469 buf->f_bfree = ((capacity - used) >> EXOFS_BLKSHIFT);
470 buf->f_bavail = buf->f_bfree;
471 buf->f_files = sbi->s_numfiles;
472 buf->f_ffree = EXOFS_MAX_ID - sbi->s_numfiles;
473 buf->f_namelen = EXOFS_NAME_LEN;
474
475out:
476 osd_end_request(or);
477 return ret;
478}
479
480static const struct super_operations exofs_sops = {
481 .alloc_inode = exofs_alloc_inode,
482 .destroy_inode = exofs_destroy_inode,
483 .write_inode = exofs_write_inode,
484 .delete_inode = exofs_delete_inode,
485 .put_super = exofs_put_super,
486 .write_super = exofs_write_super,
487 .statfs = exofs_statfs,
488};
489
490/******************************************************************************
491 * EXPORT OPERATIONS
492 *****************************************************************************/
493
494struct dentry *exofs_get_parent(struct dentry *child)
495{
496 unsigned long ino = exofs_parent_ino(child);
497
498 if (!ino)
499 return NULL;
500
501 return d_obtain_alias(exofs_iget(child->d_inode->i_sb, ino));
502}
503
504static struct inode *exofs_nfs_get_inode(struct super_block *sb,
505 u64 ino, u32 generation)
506{
507 struct inode *inode;
508
509 inode = exofs_iget(sb, ino);
510 if (IS_ERR(inode))
511 return ERR_CAST(inode);
512 if (generation && inode->i_generation != generation) {
513 /* we didn't find the right inode.. */
514 iput(inode);
515 return ERR_PTR(-ESTALE);
516 }
517 return inode;
518}
519
520static struct dentry *exofs_fh_to_dentry(struct super_block *sb,
521 struct fid *fid, int fh_len, int fh_type)
522{
523 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
524 exofs_nfs_get_inode);
525}
526
527static struct dentry *exofs_fh_to_parent(struct super_block *sb,
528 struct fid *fid, int fh_len, int fh_type)
529{
530 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
531 exofs_nfs_get_inode);
532}
533
534static const struct export_operations exofs_export_ops = {
535 .fh_to_dentry = exofs_fh_to_dentry,
536 .fh_to_parent = exofs_fh_to_parent,
537 .get_parent = exofs_get_parent,
538};
539
540/******************************************************************************
541 * INSMOD/RMMOD
542 *****************************************************************************/
543
544/*
545 * struct that describes this file system
546 */
547static struct file_system_type exofs_type = {
548 .owner = THIS_MODULE,
549 .name = "exofs",
550 .get_sb = exofs_get_sb,
551 .kill_sb = generic_shutdown_super,
552};
553
554static int __init init_exofs(void)
555{
556 int err;
557
558 err = init_inodecache();
559 if (err)
560 goto out;
561
562 err = register_filesystem(&exofs_type);
563 if (err)
564 goto out_d;
565
566 return 0;
567out_d:
568 destroy_inodecache();
569out:
570 return err;
571}
572
573static void __exit exit_exofs(void)
574{
575 unregister_filesystem(&exofs_type);
576 destroy_inodecache();
577}
578
579MODULE_AUTHOR("Avishay Traeger <avishay@gmail.com>");
580MODULE_DESCRIPTION("exofs");
581MODULE_LICENSE("GPL");
582
583module_init(init_exofs)
584module_exit(exit_exofs)
diff --git a/fs/exofs/symlink.c b/fs/exofs/symlink.c
new file mode 100644
index 000000000000..36e2d7bc7f7b
--- /dev/null
+++ b/fs/exofs/symlink.c
@@ -0,0 +1,57 @@
1/*
2 * Copyright (C) 2005, 2006
3 * Avishay Traeger (avishay@gmail.com) (avishay@il.ibm.com)
4 * Copyright (C) 2005, 2006
5 * International Business Machines
6 * Copyright (C) 2008, 2009
7 * Boaz Harrosh <bharrosh@panasas.com>
8 *
9 * Copyrights for code taken from ext2:
10 * Copyright (C) 1992, 1993, 1994, 1995
11 * Remy Card (card@masi.ibp.fr)
12 * Laboratoire MASI - Institut Blaise Pascal
13 * Universite Pierre et Marie Curie (Paris VI)
14 * from
15 * linux/fs/minix/inode.c
16 * Copyright (C) 1991, 1992 Linus Torvalds
17 *
18 * This file is part of exofs.
19 *
20 * exofs is free software; you can redistribute it and/or modify
21 * it under the terms of the GNU General Public License as published by
22 * the Free Software Foundation. Since it is based on ext2, and the only
23 * valid version of GPL for the Linux kernel is version 2, the only valid
24 * version of GPL for exofs is version 2.
25 *
26 * exofs is distributed in the hope that it will be useful,
27 * but WITHOUT ANY WARRANTY; without even the implied warranty of
28 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
29 * GNU General Public License for more details.
30 *
31 * You should have received a copy of the GNU General Public License
32 * along with exofs; if not, write to the Free Software
33 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
34 */
35
36#include <linux/namei.h>
37
38#include "exofs.h"
39
40static void *exofs_follow_link(struct dentry *dentry, struct nameidata *nd)
41{
42 struct exofs_i_info *oi = exofs_i(dentry->d_inode);
43
44 nd_set_link(nd, (char *)oi->i_data);
45 return NULL;
46}
47
48const struct inode_operations exofs_symlink_inode_operations = {
49 .readlink = generic_readlink,
50 .follow_link = page_follow_link_light,
51 .put_link = page_put_link,
52};
53
54const struct inode_operations exofs_fast_symlink_inode_operations = {
55 .readlink = generic_readlink,
56 .follow_link = exofs_follow_link,
57};
generated by cgit v1.2.2 (git 2.25.0) at 2024-10-23 09:39:01 -0400