exofs specific options: Options are separated by commas (,)
pid=<integer> - The partition number to mount/create as
container of the filesystem.
- This option is mandatory.
+ This option is mandatory. integer can be
+ Hex by pre-pending an 0x to the number.
+ osdname=<id> - Mount by a device's osdname.
+ osdname is usually a 36 character uuid of the
+ form "d2683732-c906-4ee1-9dbd-c10c27bb40df".
+ It is one of the device's uuid specified in the
+ mkfs.exofs format command.
+ If this option is specified then the /dev/osdX
+ above can be empty and is ignored.
to=<integer> - Timeout in ticks for a single command.
default is (60 * HZ) [for debugging only]
#define EXOFS_ROOT_ID 0x10002 /* object ID for root directory */
/* exofs Application specific page/attribute */
+/* Inode attrs */
# define EXOFS_APAGE_FS_DATA (OSD_APAGE_APP_DEFINED_FIRST + 3)
# define EXOFS_ATTR_INODE_DATA 1
# define EXOFS_ATTR_INODE_FILE_LAYOUT 2
# define EXOFS_ATTR_INODE_DIR_LAYOUT 3
+/* Partition attrs */
+# define EXOFS_APAGE_SB_DATA (0xF0000000U + 3)
+# define EXOFS_ATTR_SB_STATS 1
/*
* The maximum number of files we can have is limited by the size of the
*/
enum {EXOFS_FSCB_VER = 1, EXOFS_DT_VER = 1};
struct exofs_fscb {
- __le64 s_nextid; /* Highest object ID used */
- __le64 s_numfiles; /* Number of files on fs */
+ __le64 s_nextid; /* Only used after mkfs */
+ __le64 s_numfiles; /* Only used after mkfs */
__le32 s_version; /* == EXOFS_FSCB_VER */
__le16 s_magic; /* Magic signature */
__le16 s_newfs; /* Non-zero if this is a new fs */
__le64 s_dev_table_count; /* == 0 means no dev_table */
} __packed;
+/*
+ * This struct is set on the FS partition's attributes.
+ * [EXOFS_APAGE_SB_DATA, EXOFS_ATTR_SB_STATS] and is written together
+ * with the create command, to atomically persist the sb writeable information.
+ */
+struct exofs_sb_stats {
+ __le64 s_nextid; /* Highest object ID used */
+ __le64 s_numfiles; /* Number of files on fs */
+} __packed;
+
/*
* Describes the raid used in the FS. It is part of the device table.
* This here is taken from the pNFS-objects definition. In exofs we
Ebadsize:
EXOFS_ERR("ERROR [exofs_check_page]: "
- "size of directory #%lu is not a multiple of chunk size",
+ "size of directory(0x%lx) is not a multiple of chunk size\n",
dir->i_ino
);
goto fail;
goto bad_entry;
bad_entry:
EXOFS_ERR(
- "ERROR [exofs_check_page]: bad entry in directory #%lu: %s - "
- "offset=%lu, inode=%llu, rec_len=%d, name_len=%d",
+ "ERROR [exofs_check_page]: bad entry in directory(0x%lx): %s - "
+ "offset=%lu, inode=0x%llu, rec_len=%d, name_len=%d\n",
dir->i_ino, error, (page->index<<PAGE_CACHE_SHIFT)+offs,
_LLU(le64_to_cpu(p->inode_no)),
rec_len, p->name_len);
Eend:
p = (struct exofs_dir_entry *)(kaddr + offs);
EXOFS_ERR("ERROR [exofs_check_page]: "
- "entry in directory #%lu spans the page boundary"
- "offset=%lu, inode=%llu",
+ "entry in directory(0x%lx) spans the page boundary"
+ "offset=%lu, inode=0x%llx\n",
dir->i_ino, (page->index<<PAGE_CACHE_SHIFT)+offs,
_LLU(le64_to_cpu(p->inode_no)));
fail:
struct page *page = exofs_get_page(inode, n);
if (IS_ERR(page)) {
- EXOFS_ERR("ERROR: "
- "bad page in #%lu",
- inode->i_ino);
+ EXOFS_ERR("ERROR: bad page in directory(0x%lx)\n",
+ inode->i_ino);
filp->f_pos += PAGE_CACHE_SIZE - offset;
return PTR_ERR(page);
}
for (; (char *)de <= limit; de = exofs_next_entry(de)) {
if (de->rec_len == 0) {
EXOFS_ERR("ERROR: "
- "zero-length directory entry");
+ "zero-length entry in directory(0x%lx)\n",
+ inode->i_ino);
exofs_put_page(page);
return -EIO;
}
kaddr += exofs_last_byte(dir, n) - reclen;
while ((char *) de <= kaddr) {
if (de->rec_len == 0) {
- EXOFS_ERR(
- "ERROR: exofs_find_entry: "
- "zero-length directory entry");
+ EXOFS_ERR("ERROR: zero-length entry in "
+ "directory(0x%lx)\n",
+ dir->i_ino);
exofs_put_page(page);
goto out;
}
}
if (de->rec_len == 0) {
EXOFS_ERR("ERROR: exofs_add_link: "
- "zero-length directory entry");
+ "zero-length entry in directory(0x%lx)\n",
+ inode->i_ino);
err = -EIO;
goto out_unlock;
}
exofs_put_page(page);
}
- EXOFS_ERR("exofs_add_link: BAD dentry=%p or inode=%p", dentry, inode);
+ EXOFS_ERR("exofs_add_link: BAD dentry=%p or inode=0x%lx\n",
+ dentry, inode->i_ino);
return -EINVAL;
got_it:
while (de < dir) {
if (de->rec_len == 0) {
EXOFS_ERR("ERROR: exofs_delete_entry:"
- "zero-length directory entry");
+ "zero-length entry in directory(0x%lx)\n",
+ inode->i_ino);
err = -EIO;
goto out;
}
* our extension to the in-memory superblock
*/
struct exofs_sb_info {
- struct exofs_fscb s_fscb; /* Written often, pre-allocate*/
+ struct exofs_sb_stats s_ess; /* Written often, pre-allocate*/
int s_timeout; /* timeout for OSD operations */
uint64_t s_nextid; /* highest object ID used */
uint32_t s_numfiles; /* number of files on fs */
}
/* inode.c */
+unsigned exofs_max_io_pages(struct exofs_layout *layout,
+ unsigned expected_pages);
int exofs_setattr(struct dentry *, struct iattr *);
int exofs_write_begin(struct file *file, struct address_space *mapping,
loff_t pos, unsigned len, unsigned flags,
struct inode *);
/* super.c */
-int exofs_sync_fs(struct super_block *sb, int wait);
+int exofs_sbi_write_stats(struct exofs_sb_info *sbi);
/*********************
* operation vectors *
static int exofs_file_fsync(struct file *filp, int datasync)
{
int ret;
- struct inode *inode = filp->f_mapping->host;
- struct super_block *sb;
-
- if (!(inode->i_state & I_DIRTY))
- return 0;
- if (datasync && !(inode->i_state & I_DIRTY_DATASYNC))
- return 0;
-
- ret = sync_inode_metadata(inode, 1);
-
- /* This is a good place to write the sb */
- /* TODO: Sechedule an sb-sync on create */
- sb = inode->i_sb;
- if (sb->s_dirt)
- exofs_sync_fs(sb, 1);
+ ret = sync_inode_metadata(filp->f_mapping->host, 1);
return ret;
}
PAGE_SIZE / sizeof(struct page *),
};
+unsigned exofs_max_io_pages(struct exofs_layout *layout,
+ unsigned expected_pages)
+{
+ unsigned pages = min_t(unsigned, expected_pages, MAX_PAGES_KMALLOC);
+
+ /* TODO: easily support bio chaining */
+ pages = min_t(unsigned, pages,
+ layout->group_width * BIO_MAX_PAGES_KMALLOC);
+ return pages;
+}
+
struct page_collect {
struct exofs_sb_info *sbi;
struct inode *inode;
static int pcol_try_alloc(struct page_collect *pcol)
{
- unsigned pages = min_t(unsigned, pcol->expected_pages,
- MAX_PAGES_KMALLOC);
+ unsigned pages;
if (!pcol->ios) { /* First time allocate io_state */
int ret = exofs_get_io_state(&pcol->sbi->layout, &pcol->ios);
}
/* TODO: easily support bio chaining */
- pages = min_t(unsigned, pages,
- pcol->sbi->layout.group_width * BIO_MAX_PAGES_KMALLOC);
+ pages = exofs_max_io_pages(&pcol->sbi->layout, pcol->expected_pages);
for (; pages; pages >>= 1) {
pcol->pages = kmalloc(pages * sizeof(struct page *),
if (!pcol->read_4_write)
unlock_page(page);
- EXOFS_DBGMSG("readpage_strip(0x%lx, 0x%lx) empty page,"
- " splitting\n", inode->i_ino, page->index);
+ EXOFS_DBGMSG("readpage_strip(0x%lx) empty page len=%zx "
+ "read_4_write=%d index=0x%lx end_index=0x%lx "
+ "splitting\n", inode->i_ino, len,
+ pcol->read_4_write, page->index, end_index);
return read_exec(pcol);
}
/* read modify write */
if (!PageUptodate(page) && (len != PAGE_CACHE_SIZE)) {
+ loff_t i_size = i_size_read(mapping->host);
+ pgoff_t end_index = i_size >> PAGE_CACHE_SHIFT;
+ size_t rlen;
+
+ if (page->index < end_index)
+ rlen = PAGE_CACHE_SIZE;
+ else if (page->index == end_index)
+ rlen = i_size & ~PAGE_CACHE_MASK;
+ else
+ rlen = 0;
+
+ if (!rlen) {
+ clear_highpage(page);
+ SetPageUptodate(page);
+ goto out;
+ }
+
ret = _readpage(page, true);
if (ret) {
/*SetPageError was done by _readpage. Is it ok?*/
unlock_page(page);
- EXOFS_DBGMSG("__readpage_filler failed\n");
+ EXOFS_DBGMSG("__readpage failed\n");
}
}
out:
memcpy(oi->i_data, fcb.i_data, sizeof(fcb.i_data));
}
+ inode->i_mapping->backing_dev_info = sb->s_bdi;
if (S_ISREG(inode->i_mode)) {
inode->i_op = &exofs_file_inode_operations;
inode->i_fop = &exofs_file_operations;
}
return unlikely(is_bad_inode(&oi->vfs_inode)) ? -EIO : 0;
}
+
/*
* Callback function from exofs_new_inode(). The important thing is that we
* set the obj_created flag so that other methods know that the object exists on
sbi = sb->s_fs_info;
- sb->s_dirt = 1;
+ inode->i_mapping->backing_dev_info = sb->s_bdi;
inode_init_owner(inode, dir, mode);
inode->i_ino = sbi->s_nextid++;
inode->i_blkbits = EXOFS_BLKSHIFT;
spin_unlock(&sbi->s_next_gen_lock);
insert_inode_hash(inode);
+ exofs_sbi_write_stats(sbi); /* Make sure new sbi->s_nextid is on disk */
+
mark_inode_dirty(inode);
ret = exofs_get_io_state(&sbi->layout, &ios);
int exofs_write_inode(struct inode *inode, struct writeback_control *wbc)
{
- return exofs_update_inode(inode, wbc->sync_mode == WB_SYNC_ALL);
+ /* FIXME: fix fsync and use wbc->sync_mode == WB_SYNC_ALL */
+ return exofs_update_inode(inode, 1);
}
/*
* struct to hold what we get from mount options
*/
struct exofs_mountopt {
+ bool is_osdname;
const char *dev_name;
uint64_t pid;
int timeout;
/*
* exofs-specific mount-time options.
*/
-enum { Opt_pid, Opt_to, Opt_mkfs, Opt_format, Opt_err };
+enum { Opt_name, Opt_pid, Opt_to, Opt_err };
/*
* Our mount-time options. These should ideally be 64-bit unsigned, but the
* sufficient for most applications now.
*/
static match_table_t tokens = {
+ {Opt_name, "osdname=%s"},
{Opt_pid, "pid=%u"},
{Opt_to, "to=%u"},
{Opt_err, NULL}
token = match_token(p, tokens, args);
switch (token) {
+ case Opt_name:
+ opts->dev_name = match_strdup(&args[0]);
+ if (unlikely(!opts->dev_name)) {
+ EXOFS_ERR("Error allocating dev_name");
+ return -ENOMEM;
+ }
+ opts->is_osdname = true;
+ break;
case Opt_pid:
if (0 == match_strlcpy(str, &args[0], sizeof(str)))
return -EINVAL;
static const struct super_operations exofs_sops;
static const struct export_operations exofs_export_ops;
+static const struct osd_attr g_attr_sb_stats = ATTR_DEF(
+ EXOFS_APAGE_SB_DATA,
+ EXOFS_ATTR_SB_STATS,
+ sizeof(struct exofs_sb_stats));
+
+static int __sbi_read_stats(struct exofs_sb_info *sbi)
+{
+ struct osd_attr attrs[] = {
+ [0] = g_attr_sb_stats,
+ };
+ struct exofs_io_state *ios;
+ int ret;
+
+ ret = exofs_get_io_state(&sbi->layout, &ios);
+ if (unlikely(ret)) {
+ EXOFS_ERR("%s: exofs_get_io_state failed.\n", __func__);
+ return ret;
+ }
+
+ ios->cred = sbi->s_cred;
+
+ ios->in_attr = attrs;
+ ios->in_attr_len = ARRAY_SIZE(attrs);
+
+ ret = exofs_sbi_read(ios);
+ if (unlikely(ret)) {
+ EXOFS_ERR("Error reading super_block stats => %d\n", ret);
+ goto out;
+ }
+
+ ret = extract_attr_from_ios(ios, &attrs[0]);
+ if (ret) {
+ EXOFS_ERR("%s: extract_attr of sb_stats failed\n", __func__);
+ goto out;
+ }
+ if (attrs[0].len) {
+ struct exofs_sb_stats *ess;
+
+ if (unlikely(attrs[0].len != sizeof(*ess))) {
+ EXOFS_ERR("%s: Wrong version of exofs_sb_stats "
+ "size(%d) != expected(%zd)\n",
+ __func__, attrs[0].len, sizeof(*ess));
+ goto out;
+ }
+
+ ess = attrs[0].val_ptr;
+ sbi->s_nextid = le64_to_cpu(ess->s_nextid);
+ sbi->s_numfiles = le32_to_cpu(ess->s_numfiles);
+ }
+
+out:
+ exofs_put_io_state(ios);
+ return ret;
+}
+
+static void stats_done(struct exofs_io_state *ios, void *p)
+{
+ exofs_put_io_state(ios);
+ /* Good thanks nothing to do anymore */
+}
+
+/* Asynchronously write the stats attribute */
+int exofs_sbi_write_stats(struct exofs_sb_info *sbi)
+{
+ struct osd_attr attrs[] = {
+ [0] = g_attr_sb_stats,
+ };
+ struct exofs_io_state *ios;
+ int ret;
+
+ ret = exofs_get_io_state(&sbi->layout, &ios);
+ if (unlikely(ret)) {
+ EXOFS_ERR("%s: exofs_get_io_state failed.\n", __func__);
+ return ret;
+ }
+
+ sbi->s_ess.s_nextid = cpu_to_le64(sbi->s_nextid);
+ sbi->s_ess.s_numfiles = cpu_to_le64(sbi->s_numfiles);
+ attrs[0].val_ptr = &sbi->s_ess;
+
+ ios->cred = sbi->s_cred;
+ ios->done = stats_done;
+ ios->private = sbi;
+ ios->out_attr = attrs;
+ ios->out_attr_len = ARRAY_SIZE(attrs);
+
+ ret = exofs_sbi_write(ios);
+ if (unlikely(ret)) {
+ EXOFS_ERR("%s: exofs_sbi_write failed.\n", __func__);
+ exofs_put_io_state(ios);
+ }
+
+ return ret;
+}
+
/*
* Write the superblock to the OSD
*/
struct exofs_io_state *ios;
int ret = -ENOMEM;
- lock_super(sb);
+ fscb = kmalloc(sizeof(*fscb), GFP_KERNEL);
+ if (unlikely(!fscb))
+ return -ENOMEM;
+
sbi = sb->s_fs_info;
- fscb = &sbi->s_fscb;
+ /* NOTE: We no longer dirty the super_block anywhere in exofs. The
+ * reason we write the fscb here on unmount is so we can stay backwards
+ * compatible with fscb->s_version == 1. (What we are not compatible
+ * with is if a new version FS crashed and then we try to mount an old
+ * version). Otherwise the exofs_fscb is read-only from mkfs time. All
+ * the writeable info is set in exofs_sbi_write_stats() above.
+ */
ret = exofs_get_io_state(&sbi->layout, &ios);
- if (ret)
+ if (unlikely(ret))
goto out;
- /* Note: We only write the changing part of the fscb. .i.e upto the
- * the fscb->s_dev_table_oid member. There is no read-modify-write
- * here.
- */
+ lock_super(sb);
+
ios->length = offsetof(struct exofs_fscb, s_dev_table_oid);
memset(fscb, 0, ios->length);
fscb->s_nextid = cpu_to_le64(sbi->s_nextid);
ios->cred = sbi->s_cred;
ret = exofs_sbi_write(ios);
- if (unlikely(ret)) {
+ if (unlikely(ret))
EXOFS_ERR("%s: exofs_sbi_write failed.\n", __func__);
- goto out;
- }
- sb->s_dirt = 0;
+ else
+ sb->s_dirt = 0;
+
+ unlock_super(sb);
out:
EXOFS_DBGMSG("s_nextid=0x%llx ret=%d\n", _LLU(sbi->s_nextid), ret);
exofs_put_io_state(ios);
- unlock_super(sb);
+ kfree(fscb);
return ret;
}
int num_pend;
struct exofs_sb_info *sbi = sb->s_fs_info;
- if (sb->s_dirt)
- exofs_write_super(sb);
-
/* make sure there are no pending commands */
for (num_pend = atomic_read(&sbi->s_curr_pending); num_pend > 0;
num_pend = atomic_read(&sbi->s_curr_pending)) {
wait_queue_head_t wq;
+
+ printk(KERN_NOTICE "%s: !!Pending operations in flight. "
+ "This is a BUG. please report to osd-dev@open-osd.org\n",
+ __func__);
init_waitqueue_head(&wq);
wait_event_timeout(wq,
(atomic_read(&sbi->s_curr_pending) == 0),
return 0;
}
+static unsigned __ra_pages(struct exofs_layout *layout)
+{
+ const unsigned _MIN_RA = 32; /* min 128K read-ahead */
+ unsigned ra_pages = layout->group_width * layout->stripe_unit /
+ PAGE_SIZE;
+ unsigned max_io_pages = exofs_max_io_pages(layout, ~0);
+
+ ra_pages *= 2; /* two stripes */
+ if (ra_pages < _MIN_RA)
+ ra_pages = roundup(_MIN_RA, ra_pages / 2);
+
+ if (ra_pages > max_io_pages)
+ ra_pages = max_io_pages;
+
+ return ra_pages;
+}
+
/* @odi is valid only as long as @fscb_dev is valid */
static int exofs_devs_2_odi(struct exofs_dt_device_info *dt_dev,
struct osd_dev_info *odi)
}
od = osduld_info_lookup(&odi);
- if (unlikely(IS_ERR(od))) {
+ if (IS_ERR(od)) {
ret = PTR_ERR(od);
EXOFS_ERR("ERROR: device requested is not found "
"osd_name-%s =>%d\n", odi.osdname, ret);
goto free_bdi;
/* use mount options to fill superblock */
- od = osduld_path_lookup(opts->dev_name);
+ if (opts->is_osdname) {
+ struct osd_dev_info odi = {.systemid_len = 0};
+
+ odi.osdname_len = strlen(opts->dev_name);
+ odi.osdname = (u8 *)opts->dev_name;
+ od = osduld_info_lookup(&odi);
+ } else {
+ od = osduld_path_lookup(opts->dev_name);
+ }
if (IS_ERR(od)) {
- ret = PTR_ERR(od);
+ ret = -EINVAL;
goto free_sbi;
}
goto free_sbi;
sb->s_magic = le16_to_cpu(fscb.s_magic);
+ /* NOTE: we read below to be backward compatible with old versions */
sbi->s_nextid = le64_to_cpu(fscb.s_nextid);
sbi->s_numfiles = le32_to_cpu(fscb.s_numfiles);
ret = -EINVAL;
goto free_sbi;
}
- if (le32_to_cpu(fscb.s_version) != EXOFS_FSCB_VER) {
+ if (le32_to_cpu(fscb.s_version) > EXOFS_FSCB_VER) {
EXOFS_ERR("ERROR: Bad FSCB version expected-%d got-%d\n",
EXOFS_FSCB_VER, le32_to_cpu(fscb.s_version));
ret = -EINVAL;
goto free_sbi;
}
+ __sbi_read_stats(sbi);
+
/* set up operation vectors */
+ sbi->bdi.ra_pages = __ra_pages(&sbi->layout);
sb->s_bdi = &sbi->bdi;
sb->s_fs_info = sbi;
sb->s_op = &exofs_sops;
_exofs_print_device("Mounting", opts->dev_name, sbi->layout.s_ods[0],
sbi->layout.s_pid);
+ if (opts->is_osdname)
+ kfree(opts->dev_name);
return 0;
free_sbi:
EXOFS_ERR("Unable to mount exofs on %s pid=0x%llx err=%d\n",
opts->dev_name, sbi->layout.s_pid, ret);
exofs_free_sbi(sbi);
+ if (opts->is_osdname)
+ kfree(opts->dev_name);
return ret;
}
if (ret)
return ERR_PTR(ret);
- opts.dev_name = dev_name;
+ if (!opts.dev_name)
+ opts.dev_name = dev_name;
return mount_nodev(type, flags, &opts, exofs_fill_super);
}
projects / cascardo / linux.git / commitdiff