aboutsummaryrefslogtreecommitdiffstats
path: root/fs/exofs
diff options
context:
space:
mode:
Diffstat (limited to 'fs/exofs')
-rw-r--r--fs/exofs/common.h22
-rw-r--r--fs/exofs/dir.c33
-rw-r--r--fs/exofs/exofs.h6
-rw-r--r--fs/exofs/file.c16
-rw-r--r--fs/exofs/inode.c52
-rw-r--r--fs/exofs/super.c190
6 files changed, 252 insertions, 67 deletions
diff --git a/fs/exofs/common.h b/fs/exofs/common.h
index f0d520312d8b..3bbd46956d77 100644
--- a/fs/exofs/common.h
+++ b/fs/exofs/common.h
@@ -53,10 +53,14 @@
53#define EXOFS_ROOT_ID 0x10002 /* object ID for root directory */ 53#define EXOFS_ROOT_ID 0x10002 /* object ID for root directory */
54 54
55/* exofs Application specific page/attribute */ 55/* exofs Application specific page/attribute */
56/* Inode attrs */
56# define EXOFS_APAGE_FS_DATA (OSD_APAGE_APP_DEFINED_FIRST + 3) 57# define EXOFS_APAGE_FS_DATA (OSD_APAGE_APP_DEFINED_FIRST + 3)
57# define EXOFS_ATTR_INODE_DATA 1 58# define EXOFS_ATTR_INODE_DATA 1
58# define EXOFS_ATTR_INODE_FILE_LAYOUT 2 59# define EXOFS_ATTR_INODE_FILE_LAYOUT 2
59# define EXOFS_ATTR_INODE_DIR_LAYOUT 3 60# define EXOFS_ATTR_INODE_DIR_LAYOUT 3
61/* Partition attrs */
62# define EXOFS_APAGE_SB_DATA (0xF0000000U + 3)
63# define EXOFS_ATTR_SB_STATS 1
60 64
61/* 65/*
62 * The maximum number of files we can have is limited by the size of the 66 * The maximum number of files we can have is limited by the size of the
@@ -86,8 +90,8 @@ enum {
86 */ 90 */
87enum {EXOFS_FSCB_VER = 1, EXOFS_DT_VER = 1}; 91enum {EXOFS_FSCB_VER = 1, EXOFS_DT_VER = 1};
88struct exofs_fscb { 92struct exofs_fscb {
89 __le64 s_nextid; /* Highest object ID used */ 93 __le64 s_nextid; /* Only used after mkfs */
90 __le64 s_numfiles; /* Number of files on fs */ 94 __le64 s_numfiles; /* Only used after mkfs */
91 __le32 s_version; /* == EXOFS_FSCB_VER */ 95 __le32 s_version; /* == EXOFS_FSCB_VER */
92 __le16 s_magic; /* Magic signature */ 96 __le16 s_magic; /* Magic signature */
93 __le16 s_newfs; /* Non-zero if this is a new fs */ 97 __le16 s_newfs; /* Non-zero if this is a new fs */
@@ -98,10 +102,20 @@ struct exofs_fscb {
98} __packed; 102} __packed;
99 103
100/* 104/*
105 * This struct is set on the FS partition's attributes.
106 * [EXOFS_APAGE_SB_DATA, EXOFS_ATTR_SB_STATS] and is written together
107 * with the create command, to atomically persist the sb writeable information.
108 */
109struct exofs_sb_stats {
110 __le64 s_nextid; /* Highest object ID used */
111 __le64 s_numfiles; /* Number of files on fs */
112} __packed;
113
114/*
101 * Describes the raid used in the FS. It is part of the device table. 115 * Describes the raid used in the FS. It is part of the device table.
102 * This here is taken from the pNFS-objects definition. In exofs we 116 * This here is taken from the pNFS-objects definition. In exofs we
103 * use one raid policy through-out the filesystem. (NOTE: the funny 117 * use one raid policy through-out the filesystem. (NOTE: the funny
104 * alignment at begining. We take care of it at exofs_device_table. 118 * alignment at beginning. We take care of it at exofs_device_table.
105 */ 119 */
106struct exofs_dt_data_map { 120struct exofs_dt_data_map {
107 __le32 cb_num_comps; 121 __le32 cb_num_comps;
@@ -122,7 +136,7 @@ struct exofs_dt_device_info {
122 u8 systemid[OSD_SYSTEMID_LEN]; 136 u8 systemid[OSD_SYSTEMID_LEN];
123 __le64 long_name_offset; /* If !0 then offset-in-file */ 137 __le64 long_name_offset; /* If !0 then offset-in-file */
124 __le32 osdname_len; /* */ 138 __le32 osdname_len; /* */
125 u8 osdname[44]; /* Embbeded, Ususally an asci uuid */ 139 u8 osdname[44]; /* Embbeded, Usually an asci uuid */
126} __packed; 140} __packed;
127 141
128/* 142/*
diff --git a/fs/exofs/dir.c b/fs/exofs/dir.c
index dcc941d82d67..d0941c6a1f72 100644
--- a/fs/exofs/dir.c
+++ b/fs/exofs/dir.c
@@ -124,7 +124,7 @@ out:
124 124
125Ebadsize: 125Ebadsize:
126 EXOFS_ERR("ERROR [exofs_check_page]: " 126 EXOFS_ERR("ERROR [exofs_check_page]: "
127 "size of directory #%lu is not a multiple of chunk size", 127 "size of directory(0x%lx) is not a multiple of chunk size\n",
128 dir->i_ino 128 dir->i_ino
129 ); 129 );
130 goto fail; 130 goto fail;
@@ -142,8 +142,8 @@ Espan:
142 goto bad_entry; 142 goto bad_entry;
143bad_entry: 143bad_entry:
144 EXOFS_ERR( 144 EXOFS_ERR(
145 "ERROR [exofs_check_page]: bad entry in directory #%lu: %s - " 145 "ERROR [exofs_check_page]: bad entry in directory(0x%lx): %s - "
146 "offset=%lu, inode=%llu, rec_len=%d, name_len=%d", 146 "offset=%lu, inode=0x%llu, rec_len=%d, name_len=%d\n",
147 dir->i_ino, error, (page->index<<PAGE_CACHE_SHIFT)+offs, 147 dir->i_ino, error, (page->index<<PAGE_CACHE_SHIFT)+offs,
148 _LLU(le64_to_cpu(p->inode_no)), 148 _LLU(le64_to_cpu(p->inode_no)),
149 rec_len, p->name_len); 149 rec_len, p->name_len);
@@ -151,8 +151,8 @@ bad_entry:
151Eend: 151Eend:
152 p = (struct exofs_dir_entry *)(kaddr + offs); 152 p = (struct exofs_dir_entry *)(kaddr + offs);
153 EXOFS_ERR("ERROR [exofs_check_page]: " 153 EXOFS_ERR("ERROR [exofs_check_page]: "
154 "entry in directory #%lu spans the page boundary" 154 "entry in directory(0x%lx) spans the page boundary"
155 "offset=%lu, inode=%llu", 155 "offset=%lu, inode=0x%llx\n",
156 dir->i_ino, (page->index<<PAGE_CACHE_SHIFT)+offs, 156 dir->i_ino, (page->index<<PAGE_CACHE_SHIFT)+offs,
157 _LLU(le64_to_cpu(p->inode_no))); 157 _LLU(le64_to_cpu(p->inode_no)));
158fail: 158fail:
@@ -261,9 +261,8 @@ exofs_readdir(struct file *filp, void *dirent, filldir_t filldir)
261 struct page *page = exofs_get_page(inode, n); 261 struct page *page = exofs_get_page(inode, n);
262 262
263 if (IS_ERR(page)) { 263 if (IS_ERR(page)) {
264 EXOFS_ERR("ERROR: " 264 EXOFS_ERR("ERROR: bad page in directory(0x%lx)\n",
265 "bad page in #%lu", 265 inode->i_ino);
266 inode->i_ino);
267 filp->f_pos += PAGE_CACHE_SIZE - offset; 266 filp->f_pos += PAGE_CACHE_SIZE - offset;
268 return PTR_ERR(page); 267 return PTR_ERR(page);
269 } 268 }
@@ -283,7 +282,8 @@ exofs_readdir(struct file *filp, void *dirent, filldir_t filldir)
283 for (; (char *)de <= limit; de = exofs_next_entry(de)) { 282 for (; (char *)de <= limit; de = exofs_next_entry(de)) {
284 if (de->rec_len == 0) { 283 if (de->rec_len == 0) {
285 EXOFS_ERR("ERROR: " 284 EXOFS_ERR("ERROR: "
286 "zero-length directory entry"); 285 "zero-length entry in directory(0x%lx)\n",
286 inode->i_ino);
287 exofs_put_page(page); 287 exofs_put_page(page);
288 return -EIO; 288 return -EIO;
289 } 289 }
@@ -342,9 +342,9 @@ struct exofs_dir_entry *exofs_find_entry(struct inode *dir,
342 kaddr += exofs_last_byte(dir, n) - reclen; 342 kaddr += exofs_last_byte(dir, n) - reclen;
343 while ((char *) de <= kaddr) { 343 while ((char *) de <= kaddr) {
344 if (de->rec_len == 0) { 344 if (de->rec_len == 0) {
345 EXOFS_ERR( 345 EXOFS_ERR("ERROR: zero-length entry in "
346 "ERROR: exofs_find_entry: " 346 "directory(0x%lx)\n",
347 "zero-length directory entry"); 347 dir->i_ino);
348 exofs_put_page(page); 348 exofs_put_page(page);
349 goto out; 349 goto out;
350 } 350 }
@@ -472,7 +472,8 @@ int exofs_add_link(struct dentry *dentry, struct inode *inode)
472 } 472 }
473 if (de->rec_len == 0) { 473 if (de->rec_len == 0) {
474 EXOFS_ERR("ERROR: exofs_add_link: " 474 EXOFS_ERR("ERROR: exofs_add_link: "
475 "zero-length directory entry"); 475 "zero-length entry in directory(0x%lx)\n",
476 inode->i_ino);
476 err = -EIO; 477 err = -EIO;
477 goto out_unlock; 478 goto out_unlock;
478 } 479 }
@@ -491,7 +492,8 @@ int exofs_add_link(struct dentry *dentry, struct inode *inode)
491 exofs_put_page(page); 492 exofs_put_page(page);
492 } 493 }
493 494
494 EXOFS_ERR("exofs_add_link: BAD dentry=%p or inode=%p", dentry, inode); 495 EXOFS_ERR("exofs_add_link: BAD dentry=%p or inode=0x%lx\n",
496 dentry, inode->i_ino);
495 return -EINVAL; 497 return -EINVAL;
496 498
497got_it: 499got_it:
@@ -542,7 +544,8 @@ int exofs_delete_entry(struct exofs_dir_entry *dir, struct page *page)
542 while (de < dir) { 544 while (de < dir) {
543 if (de->rec_len == 0) { 545 if (de->rec_len == 0) {
544 EXOFS_ERR("ERROR: exofs_delete_entry:" 546 EXOFS_ERR("ERROR: exofs_delete_entry:"
545 "zero-length directory entry"); 547 "zero-length entry in directory(0x%lx)\n",
548 inode->i_ino);
546 err = -EIO; 549 err = -EIO;
547 goto out; 550 goto out;
548 } 551 }
diff --git a/fs/exofs/exofs.h b/fs/exofs/exofs.h
index 2dc925fa1010..c965806c2821 100644
--- a/fs/exofs/exofs.h
+++ b/fs/exofs/exofs.h
@@ -77,7 +77,7 @@ struct exofs_layout {
77 * our extension to the in-memory superblock 77 * our extension to the in-memory superblock
78 */ 78 */
79struct exofs_sb_info { 79struct exofs_sb_info {
80 struct exofs_fscb s_fscb; /* Written often, pre-allocate*/ 80 struct exofs_sb_stats s_ess; /* Written often, pre-allocate*/
81 int s_timeout; /* timeout for OSD operations */ 81 int s_timeout; /* timeout for OSD operations */
82 uint64_t s_nextid; /* highest object ID used */ 82 uint64_t s_nextid; /* highest object ID used */
83 uint32_t s_numfiles; /* number of files on fs */ 83 uint32_t s_numfiles; /* number of files on fs */
@@ -256,6 +256,8 @@ static inline int exofs_oi_read(struct exofs_i_info *oi,
256} 256}
257 257
258/* inode.c */ 258/* inode.c */
259unsigned exofs_max_io_pages(struct exofs_layout *layout,
260 unsigned expected_pages);
259int exofs_setattr(struct dentry *, struct iattr *); 261int exofs_setattr(struct dentry *, struct iattr *);
260int exofs_write_begin(struct file *file, struct address_space *mapping, 262int exofs_write_begin(struct file *file, struct address_space *mapping,
261 loff_t pos, unsigned len, unsigned flags, 263 loff_t pos, unsigned len, unsigned flags,
@@ -279,7 +281,7 @@ int exofs_set_link(struct inode *, struct exofs_dir_entry *, struct page *,
279 struct inode *); 281 struct inode *);
280 282
281/* super.c */ 283/* super.c */
282int exofs_sync_fs(struct super_block *sb, int wait); 284int exofs_sbi_write_stats(struct exofs_sb_info *sbi);
283 285
284/********************* 286/*********************
285 * operation vectors * 287 * operation vectors *
diff --git a/fs/exofs/file.c b/fs/exofs/file.c
index b905c79b4f0a..45ca323d8363 100644
--- a/fs/exofs/file.c
+++ b/fs/exofs/file.c
@@ -45,22 +45,8 @@ static int exofs_release_file(struct inode *inode, struct file *filp)
45static int exofs_file_fsync(struct file *filp, int datasync) 45static int exofs_file_fsync(struct file *filp, int datasync)
46{ 46{
47 int ret; 47 int ret;
48 struct inode *inode = filp->f_mapping->host;
49 struct super_block *sb;
50
51 if (!(inode->i_state & I_DIRTY))
52 return 0;
53 if (datasync && !(inode->i_state & I_DIRTY_DATASYNC))
54 return 0;
55
56 ret = sync_inode_metadata(inode, 1);
57
58 /* This is a good place to write the sb */
59 /* TODO: Sechedule an sb-sync on create */
60 sb = inode->i_sb;
61 if (sb->s_dirt)
62 exofs_sync_fs(sb, 1);
63 48
49 ret = sync_inode_metadata(filp->f_mapping->host, 1);
64 return ret; 50 return ret;
65} 51}
66 52
diff --git a/fs/exofs/inode.c b/fs/exofs/inode.c
index a7555238c41a..8472c098445d 100644
--- a/fs/exofs/inode.c
+++ b/fs/exofs/inode.c
@@ -43,6 +43,17 @@ enum { BIO_MAX_PAGES_KMALLOC =
43 PAGE_SIZE / sizeof(struct page *), 43 PAGE_SIZE / sizeof(struct page *),
44}; 44};
45 45
46unsigned exofs_max_io_pages(struct exofs_layout *layout,
47 unsigned expected_pages)
48{
49 unsigned pages = min_t(unsigned, expected_pages, MAX_PAGES_KMALLOC);
50
51 /* TODO: easily support bio chaining */
52 pages = min_t(unsigned, pages,
53 layout->group_width * BIO_MAX_PAGES_KMALLOC);
54 return pages;
55}
56
46struct page_collect { 57struct page_collect {
47 struct exofs_sb_info *sbi; 58 struct exofs_sb_info *sbi;
48 struct inode *inode; 59 struct inode *inode;
@@ -97,8 +108,7 @@ static void _pcol_reset(struct page_collect *pcol)
97 108
98static int pcol_try_alloc(struct page_collect *pcol) 109static int pcol_try_alloc(struct page_collect *pcol)
99{ 110{
100 unsigned pages = min_t(unsigned, pcol->expected_pages, 111 unsigned pages;
101 MAX_PAGES_KMALLOC);
102 112
103 if (!pcol->ios) { /* First time allocate io_state */ 113 if (!pcol->ios) { /* First time allocate io_state */
104 int ret = exofs_get_io_state(&pcol->sbi->layout, &pcol->ios); 114 int ret = exofs_get_io_state(&pcol->sbi->layout, &pcol->ios);
@@ -108,8 +118,7 @@ static int pcol_try_alloc(struct page_collect *pcol)
108 } 118 }
109 119
110 /* TODO: easily support bio chaining */ 120 /* TODO: easily support bio chaining */
111 pages = min_t(unsigned, pages, 121 pages = exofs_max_io_pages(&pcol->sbi->layout, pcol->expected_pages);
112 pcol->sbi->layout.group_width * BIO_MAX_PAGES_KMALLOC);
113 122
114 for (; pages; pages >>= 1) { 123 for (; pages; pages >>= 1) {
115 pcol->pages = kmalloc(pages * sizeof(struct page *), 124 pcol->pages = kmalloc(pages * sizeof(struct page *),
@@ -350,8 +359,10 @@ static int readpage_strip(void *data, struct page *page)
350 359
351 if (!pcol->read_4_write) 360 if (!pcol->read_4_write)
352 unlock_page(page); 361 unlock_page(page);
353 EXOFS_DBGMSG("readpage_strip(0x%lx, 0x%lx) empty page," 362 EXOFS_DBGMSG("readpage_strip(0x%lx) empty page len=%zx "
354 " splitting\n", inode->i_ino, page->index); 363 "read_4_write=%d index=0x%lx end_index=0x%lx "
364 "splitting\n", inode->i_ino, len,
365 pcol->read_4_write, page->index, end_index);
355 366
356 return read_exec(pcol); 367 return read_exec(pcol);
357 } 368 }
@@ -722,11 +733,28 @@ int exofs_write_begin(struct file *file, struct address_space *mapping,
722 733
723 /* read modify write */ 734 /* read modify write */
724 if (!PageUptodate(page) && (len != PAGE_CACHE_SIZE)) { 735 if (!PageUptodate(page) && (len != PAGE_CACHE_SIZE)) {
736 loff_t i_size = i_size_read(mapping->host);
737 pgoff_t end_index = i_size >> PAGE_CACHE_SHIFT;
738 size_t rlen;
739
740 if (page->index < end_index)
741 rlen = PAGE_CACHE_SIZE;
742 else if (page->index == end_index)
743 rlen = i_size & ~PAGE_CACHE_MASK;
744 else
745 rlen = 0;
746
747 if (!rlen) {
748 clear_highpage(page);
749 SetPageUptodate(page);
750 goto out;
751 }
752
725 ret = _readpage(page, true); 753 ret = _readpage(page, true);
726 if (ret) { 754 if (ret) {
727 /*SetPageError was done by _readpage. Is it ok?*/ 755 /*SetPageError was done by _readpage. Is it ok?*/
728 unlock_page(page); 756 unlock_page(page);
729 EXOFS_DBGMSG("__readpage_filler failed\n"); 757 EXOFS_DBGMSG("__readpage failed\n");
730 } 758 }
731 } 759 }
732out: 760out:
@@ -795,7 +823,6 @@ const struct address_space_operations exofs_aops = {
795 .direct_IO = NULL, /* TODO: Should be trivial to do */ 823 .direct_IO = NULL, /* TODO: Should be trivial to do */
796 824
797 /* With these NULL has special meaning or default is not exported */ 825 /* With these NULL has special meaning or default is not exported */
798 .sync_page = NULL,
799 .get_xip_mem = NULL, 826 .get_xip_mem = NULL,
800 .migratepage = NULL, 827 .migratepage = NULL,
801 .launder_page = NULL, 828 .launder_page = NULL,
@@ -1030,6 +1057,7 @@ struct inode *exofs_iget(struct super_block *sb, unsigned long ino)
1030 memcpy(oi->i_data, fcb.i_data, sizeof(fcb.i_data)); 1057 memcpy(oi->i_data, fcb.i_data, sizeof(fcb.i_data));
1031 } 1058 }
1032 1059
1060 inode->i_mapping->backing_dev_info = sb->s_bdi;
1033 if (S_ISREG(inode->i_mode)) { 1061 if (S_ISREG(inode->i_mode)) {
1034 inode->i_op = &exofs_file_inode_operations; 1062 inode->i_op = &exofs_file_inode_operations;
1035 inode->i_fop = &exofs_file_operations; 1063 inode->i_fop = &exofs_file_operations;
@@ -1073,6 +1101,7 @@ int __exofs_wait_obj_created(struct exofs_i_info *oi)
1073 } 1101 }
1074 return unlikely(is_bad_inode(&oi->vfs_inode)) ? -EIO : 0; 1102 return unlikely(is_bad_inode(&oi->vfs_inode)) ? -EIO : 0;
1075} 1103}
1104
1076/* 1105/*
1077 * Callback function from exofs_new_inode(). The important thing is that we 1106 * Callback function from exofs_new_inode(). The important thing is that we
1078 * set the obj_created flag so that other methods know that the object exists on 1107 * set the obj_created flag so that other methods know that the object exists on
@@ -1130,7 +1159,7 @@ struct inode *exofs_new_inode(struct inode *dir, int mode)
1130 1159
1131 sbi = sb->s_fs_info; 1160 sbi = sb->s_fs_info;
1132 1161
1133 sb->s_dirt = 1; 1162 inode->i_mapping->backing_dev_info = sb->s_bdi;
1134 inode_init_owner(inode, dir, mode); 1163 inode_init_owner(inode, dir, mode);
1135 inode->i_ino = sbi->s_nextid++; 1164 inode->i_ino = sbi->s_nextid++;
1136 inode->i_blkbits = EXOFS_BLKSHIFT; 1165 inode->i_blkbits = EXOFS_BLKSHIFT;
@@ -1141,6 +1170,8 @@ struct inode *exofs_new_inode(struct inode *dir, int mode)
1141 spin_unlock(&sbi->s_next_gen_lock); 1170 spin_unlock(&sbi->s_next_gen_lock);
1142 insert_inode_hash(inode); 1171 insert_inode_hash(inode);
1143 1172
1173 exofs_sbi_write_stats(sbi); /* Make sure new sbi->s_nextid is on disk */
1174
1144 mark_inode_dirty(inode); 1175 mark_inode_dirty(inode);
1145 1176
1146 ret = exofs_get_io_state(&sbi->layout, &ios); 1177 ret = exofs_get_io_state(&sbi->layout, &ios);
@@ -1271,7 +1302,8 @@ out:
1271 1302
1272int exofs_write_inode(struct inode *inode, struct writeback_control *wbc) 1303int exofs_write_inode(struct inode *inode, struct writeback_control *wbc)
1273{ 1304{
1274 return exofs_update_inode(inode, wbc->sync_mode == WB_SYNC_ALL); 1305 /* FIXME: fix fsync and use wbc->sync_mode == WB_SYNC_ALL */
1306 return exofs_update_inode(inode, 1);
1275} 1307}
1276 1308
1277/* 1309/*
diff --git a/fs/exofs/super.c b/fs/exofs/super.c
index 8c6c4669b381..06065bd37fc3 100644
--- a/fs/exofs/super.c
+++ b/fs/exofs/super.c
@@ -48,6 +48,7 @@
48 * struct to hold what we get from mount options 48 * struct to hold what we get from mount options
49 */ 49 */
50struct exofs_mountopt { 50struct exofs_mountopt {
51 bool is_osdname;
51 const char *dev_name; 52 const char *dev_name;
52 uint64_t pid; 53 uint64_t pid;
53 int timeout; 54 int timeout;
@@ -56,7 +57,7 @@ struct exofs_mountopt {
56/* 57/*
57 * exofs-specific mount-time options. 58 * exofs-specific mount-time options.
58 */ 59 */
59enum { Opt_pid, Opt_to, Opt_mkfs, Opt_format, Opt_err }; 60enum { Opt_name, Opt_pid, Opt_to, Opt_err };
60 61
61/* 62/*
62 * Our mount-time options. These should ideally be 64-bit unsigned, but the 63 * Our mount-time options. These should ideally be 64-bit unsigned, but the
@@ -64,6 +65,7 @@ enum { Opt_pid, Opt_to, Opt_mkfs, Opt_format, Opt_err };
64 * sufficient for most applications now. 65 * sufficient for most applications now.
65 */ 66 */
66static match_table_t tokens = { 67static match_table_t tokens = {
68 {Opt_name, "osdname=%s"},
67 {Opt_pid, "pid=%u"}, 69 {Opt_pid, "pid=%u"},
68 {Opt_to, "to=%u"}, 70 {Opt_to, "to=%u"},
69 {Opt_err, NULL} 71 {Opt_err, NULL}
@@ -94,6 +96,14 @@ static int parse_options(char *options, struct exofs_mountopt *opts)
94 96
95 token = match_token(p, tokens, args); 97 token = match_token(p, tokens, args);
96 switch (token) { 98 switch (token) {
99 case Opt_name:
100 opts->dev_name = match_strdup(&args[0]);
101 if (unlikely(!opts->dev_name)) {
102 EXOFS_ERR("Error allocating dev_name");
103 return -ENOMEM;
104 }
105 opts->is_osdname = true;
106 break;
97 case Opt_pid: 107 case Opt_pid:
98 if (0 == match_strlcpy(str, &args[0], sizeof(str))) 108 if (0 == match_strlcpy(str, &args[0], sizeof(str)))
99 return -EINVAL; 109 return -EINVAL;
@@ -203,6 +213,101 @@ static void destroy_inodecache(void)
203static const struct super_operations exofs_sops; 213static const struct super_operations exofs_sops;
204static const struct export_operations exofs_export_ops; 214static const struct export_operations exofs_export_ops;
205 215
216static const struct osd_attr g_attr_sb_stats = ATTR_DEF(
217 EXOFS_APAGE_SB_DATA,
218 EXOFS_ATTR_SB_STATS,
219 sizeof(struct exofs_sb_stats));
220
221static int __sbi_read_stats(struct exofs_sb_info *sbi)
222{
223 struct osd_attr attrs[] = {
224 [0] = g_attr_sb_stats,
225 };
226 struct exofs_io_state *ios;
227 int ret;
228
229 ret = exofs_get_io_state(&sbi->layout, &ios);
230 if (unlikely(ret)) {
231 EXOFS_ERR("%s: exofs_get_io_state failed.\n", __func__);
232 return ret;
233 }
234
235 ios->cred = sbi->s_cred;
236
237 ios->in_attr = attrs;
238 ios->in_attr_len = ARRAY_SIZE(attrs);
239
240 ret = exofs_sbi_read(ios);
241 if (unlikely(ret)) {
242 EXOFS_ERR("Error reading super_block stats => %d\n", ret);
243 goto out;
244 }
245
246 ret = extract_attr_from_ios(ios, &attrs[0]);
247 if (ret) {
248 EXOFS_ERR("%s: extract_attr of sb_stats failed\n", __func__);
249 goto out;
250 }
251 if (attrs[0].len) {
252 struct exofs_sb_stats *ess;
253
254 if (unlikely(attrs[0].len != sizeof(*ess))) {
255 EXOFS_ERR("%s: Wrong version of exofs_sb_stats "
256 "size(%d) != expected(%zd)\n",
257 __func__, attrs[0].len, sizeof(*ess));
258 goto out;
259 }
260
261 ess = attrs[0].val_ptr;
262 sbi->s_nextid = le64_to_cpu(ess->s_nextid);
263 sbi->s_numfiles = le32_to_cpu(ess->s_numfiles);
264 }
265
266out:
267 exofs_put_io_state(ios);
268 return ret;
269}
270
271static void stats_done(struct exofs_io_state *ios, void *p)
272{
273 exofs_put_io_state(ios);
274 /* Good thanks nothing to do anymore */
275}
276
277/* Asynchronously write the stats attribute */
278int exofs_sbi_write_stats(struct exofs_sb_info *sbi)
279{
280 struct osd_attr attrs[] = {
281 [0] = g_attr_sb_stats,
282 };
283 struct exofs_io_state *ios;
284 int ret;
285
286 ret = exofs_get_io_state(&sbi->layout, &ios);
287 if (unlikely(ret)) {
288 EXOFS_ERR("%s: exofs_get_io_state failed.\n", __func__);
289 return ret;
290 }
291
292 sbi->s_ess.s_nextid = cpu_to_le64(sbi->s_nextid);
293 sbi->s_ess.s_numfiles = cpu_to_le64(sbi->s_numfiles);
294 attrs[0].val_ptr = &sbi->s_ess;
295
296 ios->cred = sbi->s_cred;
297 ios->done = stats_done;
298 ios->private = sbi;
299 ios->out_attr = attrs;
300 ios->out_attr_len = ARRAY_SIZE(attrs);
301
302 ret = exofs_sbi_write(ios);
303 if (unlikely(ret)) {
304 EXOFS_ERR("%s: exofs_sbi_write failed.\n", __func__);
305 exofs_put_io_state(ios);
306 }
307
308 return ret;
309}
310
206/* 311/*
207 * Write the superblock to the OSD 312 * Write the superblock to the OSD
208 */ 313 */
@@ -213,18 +318,25 @@ int exofs_sync_fs(struct super_block *sb, int wait)
213 struct exofs_io_state *ios; 318 struct exofs_io_state *ios;
214 int ret = -ENOMEM; 319 int ret = -ENOMEM;
215 320
216 lock_super(sb); 321 fscb = kmalloc(sizeof(*fscb), GFP_KERNEL);
322 if (unlikely(!fscb))
323 return -ENOMEM;
324
217 sbi = sb->s_fs_info; 325 sbi = sb->s_fs_info;
218 fscb = &sbi->s_fscb;
219 326
327 /* NOTE: We no longer dirty the super_block anywhere in exofs. The
328 * reason we write the fscb here on unmount is so we can stay backwards
329 * compatible with fscb->s_version == 1. (What we are not compatible
330 * with is if a new version FS crashed and then we try to mount an old
331 * version). Otherwise the exofs_fscb is read-only from mkfs time. All
332 * the writeable info is set in exofs_sbi_write_stats() above.
333 */
220 ret = exofs_get_io_state(&sbi->layout, &ios); 334 ret = exofs_get_io_state(&sbi->layout, &ios);
221 if (ret) 335 if (unlikely(ret))
222 goto out; 336 goto out;
223 337
224 /* Note: We only write the changing part of the fscb. .i.e upto the 338 lock_super(sb);
225 * the fscb->s_dev_table_oid member. There is no read-modify-write 339
226 * here.
227 */
228 ios->length = offsetof(struct exofs_fscb, s_dev_table_oid); 340 ios->length = offsetof(struct exofs_fscb, s_dev_table_oid);
229 memset(fscb, 0, ios->length); 341 memset(fscb, 0, ios->length);
230 fscb->s_nextid = cpu_to_le64(sbi->s_nextid); 342 fscb->s_nextid = cpu_to_le64(sbi->s_nextid);
@@ -239,16 +351,17 @@ int exofs_sync_fs(struct super_block *sb, int wait)
239 ios->cred = sbi->s_cred; 351 ios->cred = sbi->s_cred;
240 352
241 ret = exofs_sbi_write(ios); 353 ret = exofs_sbi_write(ios);
242 if (unlikely(ret)) { 354 if (unlikely(ret))
243 EXOFS_ERR("%s: exofs_sbi_write failed.\n", __func__); 355 EXOFS_ERR("%s: exofs_sbi_write failed.\n", __func__);
244 goto out; 356 else
245 } 357 sb->s_dirt = 0;
246 sb->s_dirt = 0; 358
247 359
360 unlock_super(sb);
248out: 361out:
249 EXOFS_DBGMSG("s_nextid=0x%llx ret=%d\n", _LLU(sbi->s_nextid), ret); 362 EXOFS_DBGMSG("s_nextid=0x%llx ret=%d\n", _LLU(sbi->s_nextid), ret);
250 exofs_put_io_state(ios); 363 exofs_put_io_state(ios);
251 unlock_super(sb); 364 kfree(fscb);
252 return ret; 365 return ret;
253} 366}
254 367
@@ -292,13 +405,14 @@ static void exofs_put_super(struct super_block *sb)
292 int num_pend; 405 int num_pend;
293 struct exofs_sb_info *sbi = sb->s_fs_info; 406 struct exofs_sb_info *sbi = sb->s_fs_info;
294 407
295 if (sb->s_dirt)
296 exofs_write_super(sb);
297
298 /* make sure there are no pending commands */ 408 /* make sure there are no pending commands */
299 for (num_pend = atomic_read(&sbi->s_curr_pending); num_pend > 0; 409 for (num_pend = atomic_read(&sbi->s_curr_pending); num_pend > 0;
300 num_pend = atomic_read(&sbi->s_curr_pending)) { 410 num_pend = atomic_read(&sbi->s_curr_pending)) {
301 wait_queue_head_t wq; 411 wait_queue_head_t wq;
412
413 printk(KERN_NOTICE "%s: !!Pending operations in flight. "
414 "This is a BUG. please report to osd-dev@open-osd.org\n",
415 __func__);
302 init_waitqueue_head(&wq); 416 init_waitqueue_head(&wq);
303 wait_event_timeout(wq, 417 wait_event_timeout(wq,
304 (atomic_read(&sbi->s_curr_pending) == 0), 418 (atomic_read(&sbi->s_curr_pending) == 0),
@@ -390,6 +504,23 @@ static int _read_and_match_data_map(struct exofs_sb_info *sbi, unsigned numdevs,
390 return 0; 504 return 0;
391} 505}
392 506
507static unsigned __ra_pages(struct exofs_layout *layout)
508{
509 const unsigned _MIN_RA = 32; /* min 128K read-ahead */
510 unsigned ra_pages = layout->group_width * layout->stripe_unit /
511 PAGE_SIZE;
512 unsigned max_io_pages = exofs_max_io_pages(layout, ~0);
513
514 ra_pages *= 2; /* two stripes */
515 if (ra_pages < _MIN_RA)
516 ra_pages = roundup(_MIN_RA, ra_pages / 2);
517
518 if (ra_pages > max_io_pages)
519 ra_pages = max_io_pages;
520
521 return ra_pages;
522}
523
393/* @odi is valid only as long as @fscb_dev is valid */ 524/* @odi is valid only as long as @fscb_dev is valid */
394static int exofs_devs_2_odi(struct exofs_dt_device_info *dt_dev, 525static int exofs_devs_2_odi(struct exofs_dt_device_info *dt_dev,
395 struct osd_dev_info *odi) 526 struct osd_dev_info *odi)
@@ -495,7 +626,7 @@ static int exofs_read_lookup_dev_table(struct exofs_sb_info **psbi,
495 } 626 }
496 627
497 od = osduld_info_lookup(&odi); 628 od = osduld_info_lookup(&odi);
498 if (unlikely(IS_ERR(od))) { 629 if (IS_ERR(od)) {
499 ret = PTR_ERR(od); 630 ret = PTR_ERR(od);
500 EXOFS_ERR("ERROR: device requested is not found " 631 EXOFS_ERR("ERROR: device requested is not found "
501 "osd_name-%s =>%d\n", odi.osdname, ret); 632 "osd_name-%s =>%d\n", odi.osdname, ret);
@@ -558,9 +689,17 @@ static int exofs_fill_super(struct super_block *sb, void *data, int silent)
558 goto free_bdi; 689 goto free_bdi;
559 690
560 /* use mount options to fill superblock */ 691 /* use mount options to fill superblock */
561 od = osduld_path_lookup(opts->dev_name); 692 if (opts->is_osdname) {
693 struct osd_dev_info odi = {.systemid_len = 0};
694
695 odi.osdname_len = strlen(opts->dev_name);
696 odi.osdname = (u8 *)opts->dev_name;
697 od = osduld_info_lookup(&odi);
698 } else {
699 od = osduld_path_lookup(opts->dev_name);
700 }
562 if (IS_ERR(od)) { 701 if (IS_ERR(od)) {
563 ret = PTR_ERR(od); 702 ret = -EINVAL;
564 goto free_sbi; 703 goto free_sbi;
565 } 704 }
566 705
@@ -594,6 +733,7 @@ static int exofs_fill_super(struct super_block *sb, void *data, int silent)
594 goto free_sbi; 733 goto free_sbi;
595 734
596 sb->s_magic = le16_to_cpu(fscb.s_magic); 735 sb->s_magic = le16_to_cpu(fscb.s_magic);
736 /* NOTE: we read below to be backward compatible with old versions */
597 sbi->s_nextid = le64_to_cpu(fscb.s_nextid); 737 sbi->s_nextid = le64_to_cpu(fscb.s_nextid);
598 sbi->s_numfiles = le32_to_cpu(fscb.s_numfiles); 738 sbi->s_numfiles = le32_to_cpu(fscb.s_numfiles);
599 739
@@ -604,7 +744,7 @@ static int exofs_fill_super(struct super_block *sb, void *data, int silent)
604 ret = -EINVAL; 744 ret = -EINVAL;
605 goto free_sbi; 745 goto free_sbi;
606 } 746 }
607 if (le32_to_cpu(fscb.s_version) != EXOFS_FSCB_VER) { 747 if (le32_to_cpu(fscb.s_version) > EXOFS_FSCB_VER) {
608 EXOFS_ERR("ERROR: Bad FSCB version expected-%d got-%d\n", 748 EXOFS_ERR("ERROR: Bad FSCB version expected-%d got-%d\n",
609 EXOFS_FSCB_VER, le32_to_cpu(fscb.s_version)); 749 EXOFS_FSCB_VER, le32_to_cpu(fscb.s_version));
610 ret = -EINVAL; 750 ret = -EINVAL;
@@ -622,7 +762,10 @@ static int exofs_fill_super(struct super_block *sb, void *data, int silent)
622 goto free_sbi; 762 goto free_sbi;
623 } 763 }
624 764
765 __sbi_read_stats(sbi);
766
625 /* set up operation vectors */ 767 /* set up operation vectors */
768 sbi->bdi.ra_pages = __ra_pages(&sbi->layout);
626 sb->s_bdi = &sbi->bdi; 769 sb->s_bdi = &sbi->bdi;
627 sb->s_fs_info = sbi; 770 sb->s_fs_info = sbi;
628 sb->s_op = &exofs_sops; 771 sb->s_op = &exofs_sops;
@@ -652,6 +795,8 @@ static int exofs_fill_super(struct super_block *sb, void *data, int silent)
652 795
653 _exofs_print_device("Mounting", opts->dev_name, sbi->layout.s_ods[0], 796 _exofs_print_device("Mounting", opts->dev_name, sbi->layout.s_ods[0],
654 sbi->layout.s_pid); 797 sbi->layout.s_pid);
798 if (opts->is_osdname)
799 kfree(opts->dev_name);
655 return 0; 800 return 0;
656 801
657free_sbi: 802free_sbi:
@@ -660,6 +805,8 @@ free_bdi:
660 EXOFS_ERR("Unable to mount exofs on %s pid=0x%llx err=%d\n", 805 EXOFS_ERR("Unable to mount exofs on %s pid=0x%llx err=%d\n",
661 opts->dev_name, sbi->layout.s_pid, ret); 806 opts->dev_name, sbi->layout.s_pid, ret);
662 exofs_free_sbi(sbi); 807 exofs_free_sbi(sbi);
808 if (opts->is_osdname)
809 kfree(opts->dev_name);
663 return ret; 810 return ret;
664} 811}
665 812
@@ -677,7 +824,8 @@ static struct dentry *exofs_mount(struct file_system_type *type,
677 if (ret) 824 if (ret)
678 return ERR_PTR(ret); 825 return ERR_PTR(ret);
679 826
680 opts.dev_name = dev_name; 827 if (!opts.dev_name)
828 opts.dev_name = dev_name;
681 return mount_nodev(type, flags, &opts, exofs_fill_super); 829 return mount_nodev(type, flags, &opts, exofs_fill_super);
682} 830}
683 831