#include "bcachefs.h"
#include "acl.h"
-#include "bkey_on_stack.h"
+#include "bkey_buf.h"
#include "btree_update.h"
#include "buckets.h"
#include "chardev.h"
#include <linux/aio.h>
#include <linux/backing-dev.h>
#include <linux/exportfs.h>
+#include <linux/fiemap.h>
#include <linux/module.h>
+#include <linux/pagemap.h>
#include <linux/posix_acl.h>
#include <linux/random.h>
#include <linux/statfs.h>
+#include <linux/string.h>
#include <linux/xattr.h>
static struct kmem_cache *bch2_inode_cache;
struct bch_inode_info *,
struct bch_inode_unpacked *);
-static void journal_seq_copy(struct bch_inode_info *dst,
+static void journal_seq_copy(struct bch_fs *c,
+ struct bch_inode_info *dst,
u64 journal_seq)
{
+ /*
+ * atomic64_cmpxchg has a fallback for archs that don't support it,
+ * cmpxchg does not:
+ */
+ atomic64_t *dst_seq = (void *) &dst->ei_journal_seq;
u64 old, v = READ_ONCE(dst->ei_journal_seq);
do {
if (old >= journal_seq)
break;
- } while ((v = cmpxchg(&dst->ei_journal_seq, old, journal_seq)) != old);
+ } while ((v = atomic64_cmpxchg(dst_seq, old, journal_seq)) != old);
+
+ bch2_journal_set_has_inum(&c->journal, dst->v.i_ino, journal_seq);
}
static void __pagecache_lock_put(struct pagecache_lock *lock, long i)
__pagecache_lock_put(lock, 1);
}
+bool bch2_pagecache_add_tryget(struct pagecache_lock *lock)
+{
+ return __pagecache_lock_tryget(lock, 1);
+}
+
void bch2_pagecache_add_get(struct pagecache_lock *lock)
{
__pagecache_lock_get(lock, 1);
struct bch_inode_unpacked inode_u;
int ret;
- bch2_trans_init(&trans, c, 0, 0);
+ bch2_trans_init(&trans, c, 0, 512);
retry:
bch2_trans_begin(&trans);
bch2_inode_write(&trans, iter, &inode_u) ?:
bch2_trans_commit(&trans, NULL,
&inode->ei_journal_seq,
- BTREE_INSERT_NOUNLOCK|
BTREE_INSERT_NOFAIL);
- if (ret == -EINTR)
- goto retry;
/*
* the btree node lock protects inode->ei_inode, not ei_update_lock;
if (!ret)
bch2_inode_update_after_write(c, inode, &inode_u, fields);
+ bch2_trans_iter_put(&trans, iter);
+
+ if (ret == -EINTR)
+ goto retry;
+
bch2_trans_exit(&trans);
return ret < 0 ? ret : 0;
}
return &inode->v;
}
+static int inum_test(struct inode *inode, void *p)
+{
+ unsigned long *ino = p;
+
+ return *ino == inode->i_ino;
+}
+
static struct bch_inode_info *
-__bch2_create(struct bch_inode_info *dir, struct dentry *dentry,
+__bch2_create(struct user_namespace *mnt_userns,
+ struct bch_inode_info *dir, struct dentry *dentry,
umode_t mode, dev_t rdev, bool tmpfile)
{
struct bch_fs *c = dir->v.i_sb->s_fs_info;
- struct user_namespace *ns = dir->v.i_sb->s_user_ns;
struct btree_trans trans;
struct bch_inode_unpacked dir_u;
struct bch_inode_info *inode, *old;
if (!tmpfile)
mutex_lock(&dir->ei_update_lock);
- bch2_trans_init(&trans, c, 8, 1024);
+ bch2_trans_init(&trans, c, 8,
+ 2048 + (!tmpfile ? dentry->d_name.len : 0));
retry:
bch2_trans_begin(&trans);
ret = bch2_create_trans(&trans, dir->v.i_ino, &dir_u, &inode_u,
!tmpfile ? &dentry->d_name : NULL,
- from_kuid(ns, current_fsuid()),
- from_kgid(ns, current_fsgid()),
+ from_kuid(mnt_userns, current_fsuid()),
+ from_kgid(mnt_userns, current_fsgid()),
mode, rdev,
default_acl, acl) ?:
bch2_quota_acct(c, bch_qid(&inode_u), Q_INO, 1,
if (unlikely(ret))
goto err_before_quota;
- ret = bch2_trans_commit(&trans, NULL, &journal_seq,
- BTREE_INSERT_NOUNLOCK);
+ ret = bch2_trans_commit(&trans, NULL, &journal_seq, 0);
if (unlikely(ret)) {
bch2_quota_acct(c, bch_qid(&inode_u), Q_INO, -1,
KEY_TYPE_QUOTA_WARN);
if (!tmpfile) {
bch2_inode_update_after_write(c, dir, &dir_u,
ATTR_MTIME|ATTR_CTIME);
- journal_seq_copy(dir, journal_seq);
+ journal_seq_copy(c, dir, journal_seq);
mutex_unlock(&dir->ei_update_lock);
}
bch2_vfs_inode_init(c, inode, &inode_u);
- journal_seq_copy(inode, journal_seq);
+ journal_seq_copy(c, inode, journal_seq);
set_cached_acl(&inode->v, ACL_TYPE_ACCESS, acl);
set_cached_acl(&inode->v, ACL_TYPE_DEFAULT, default_acl);
* thread pulling the inode in and modifying it:
*/
- old = to_bch_ei(insert_inode_locked2(&inode->v));
- if (unlikely(old)) {
+ inode->v.i_state |= I_CREATING;
+ old = to_bch_ei(inode_insert5(&inode->v, inode->v.i_ino,
+ inum_test, NULL, &inode->v.i_ino));
+ BUG_ON(!old);
+
+ if (unlikely(old != inode)) {
/*
* We raced, another process pulled the new inode into cache
* before us:
*/
- journal_seq_copy(old, journal_seq);
+ journal_seq_copy(c, old, journal_seq);
make_bad_inode(&inode->v);
iput(&inode->v);
{
struct bch_fs *c = vdir->i_sb->s_fs_info;
struct bch_inode_info *dir = to_bch_ei(vdir);
+ struct bch_hash_info hash = bch2_hash_info_init(c, &dir->ei_inode);
struct inode *vinode = NULL;
u64 inum;
- inum = bch2_dirent_lookup(c, dir->v.i_ino,
- &dir->ei_str_hash,
+ inum = bch2_dirent_lookup(c, dir->v.i_ino, &hash,
&dentry->d_name);
if (inum)
return d_splice_alias(vinode, dentry);
}
-static int bch2_mknod(struct inode *vdir, struct dentry *dentry,
+static int bch2_mknod(struct user_namespace *mnt_userns,
+ struct inode *vdir, struct dentry *dentry,
umode_t mode, dev_t rdev)
{
struct bch_inode_info *inode =
- __bch2_create(to_bch_ei(vdir), dentry, mode, rdev, false);
+ __bch2_create(mnt_userns, to_bch_ei(vdir), dentry, mode, rdev, false);
if (IS_ERR(inode))
return PTR_ERR(inode);
return 0;
}
-static int bch2_create(struct inode *vdir, struct dentry *dentry,
+static int bch2_create(struct user_namespace *mnt_userns,
+ struct inode *vdir, struct dentry *dentry,
umode_t mode, bool excl)
{
- return bch2_mknod(vdir, dentry, mode|S_IFREG, 0);
+ return bch2_mknod(mnt_userns, vdir, dentry, mode|S_IFREG, 0);
}
static int __bch2_link(struct bch_fs *c,
mutex_lock(&inode->ei_update_lock);
bch2_trans_init(&trans, c, 4, 1024);
- do {
- bch2_trans_begin(&trans);
- ret = bch2_link_trans(&trans,
+ ret = __bch2_trans_do(&trans, NULL, &inode->ei_journal_seq, 0,
+ bch2_link_trans(&trans,
dir->v.i_ino,
inode->v.i_ino, &dir_u, &inode_u,
- &dentry->d_name) ?:
- bch2_trans_commit(&trans, NULL,
- &inode->ei_journal_seq,
- BTREE_INSERT_NOUNLOCK);
- } while (ret == -EINTR);
+ &dentry->d_name));
if (likely(!ret)) {
BUG_ON(inode_u.bi_inum != inode->v.i_ino);
- journal_seq_copy(inode, dir->ei_journal_seq);
+ journal_seq_copy(c, inode, dir->ei_journal_seq);
bch2_inode_update_after_write(c, dir, &dir_u,
ATTR_MTIME|ATTR_CTIME);
bch2_inode_update_after_write(c, inode, &inode_u, ATTR_CTIME);
bch2_lock_inodes(INODE_UPDATE_LOCK, dir, inode);
bch2_trans_init(&trans, c, 4, 1024);
- do {
- bch2_trans_begin(&trans);
-
- ret = bch2_unlink_trans(&trans,
+ ret = __bch2_trans_do(&trans, NULL, &dir->ei_journal_seq,
+ BTREE_INSERT_NOFAIL,
+ bch2_unlink_trans(&trans,
dir->v.i_ino, &dir_u,
- &inode_u, &dentry->d_name) ?:
- bch2_trans_commit(&trans, NULL,
- &dir->ei_journal_seq,
- BTREE_INSERT_NOUNLOCK|
- BTREE_INSERT_NOFAIL);
- } while (ret == -EINTR);
+ &inode_u, &dentry->d_name));
if (likely(!ret)) {
BUG_ON(inode_u.bi_inum != inode->v.i_ino);
- journal_seq_copy(inode, dir->ei_journal_seq);
+ journal_seq_copy(c, inode, dir->ei_journal_seq);
bch2_inode_update_after_write(c, dir, &dir_u,
ATTR_MTIME|ATTR_CTIME);
bch2_inode_update_after_write(c, inode, &inode_u,
return ret;
}
-static int bch2_symlink(struct inode *vdir, struct dentry *dentry,
+static int bch2_symlink(struct user_namespace *mnt_userns,
+ struct inode *vdir, struct dentry *dentry,
const char *symname)
{
struct bch_fs *c = vdir->i_sb->s_fs_info;
struct bch_inode_info *dir = to_bch_ei(vdir), *inode;
int ret;
- inode = __bch2_create(dir, dentry, S_IFLNK|S_IRWXUGO, 0, true);
+ inode = __bch2_create(mnt_userns, dir, dentry, S_IFLNK|S_IRWXUGO, 0, true);
if (unlikely(IS_ERR(inode)))
return PTR_ERR(inode);
if (unlikely(ret))
goto err;
- journal_seq_copy(dir, inode->ei_journal_seq);
+ journal_seq_copy(c, dir, inode->ei_journal_seq);
ret = __bch2_link(c, inode, dir, dentry);
if (unlikely(ret))
return ret;
}
-static int bch2_mkdir(struct inode *vdir, struct dentry *dentry, umode_t mode)
+static int bch2_mkdir(struct user_namespace *mnt_userns,
+ struct inode *vdir, struct dentry *dentry, umode_t mode)
{
- return bch2_mknod(vdir, dentry, mode|S_IFDIR, 0);
+ return bch2_mknod(mnt_userns, vdir, dentry, mode|S_IFDIR, 0);
}
-static int bch2_rename2(struct inode *src_vdir, struct dentry *src_dentry,
+static int bch2_rename2(struct user_namespace *mnt_userns,
+ struct inode *src_vdir, struct dentry *src_dentry,
struct inode *dst_vdir, struct dentry *dst_dentry,
unsigned flags)
{
goto err;
}
-retry:
- bch2_trans_begin(&trans);
- ret = bch2_rename_trans(&trans,
- src_dir->v.i_ino, &src_dir_u,
- dst_dir->v.i_ino, &dst_dir_u,
- &src_inode_u,
- &dst_inode_u,
- &src_dentry->d_name,
- &dst_dentry->d_name,
- mode) ?:
- bch2_trans_commit(&trans, NULL,
- &journal_seq,
- BTREE_INSERT_NOUNLOCK);
- if (ret == -EINTR)
- goto retry;
+ ret = __bch2_trans_do(&trans, NULL, &journal_seq, 0,
+ bch2_rename_trans(&trans,
+ src_dir->v.i_ino, &src_dir_u,
+ dst_dir->v.i_ino, &dst_dir_u,
+ &src_inode_u,
+ &dst_inode_u,
+ &src_dentry->d_name,
+ &dst_dentry->d_name,
+ mode));
if (unlikely(ret))
goto err;
bch2_inode_update_after_write(c, src_dir, &src_dir_u,
ATTR_MTIME|ATTR_CTIME);
- journal_seq_copy(src_dir, journal_seq);
+ journal_seq_copy(c, src_dir, journal_seq);
if (src_dir != dst_dir) {
bch2_inode_update_after_write(c, dst_dir, &dst_dir_u,
ATTR_MTIME|ATTR_CTIME);
- journal_seq_copy(dst_dir, journal_seq);
+ journal_seq_copy(c, dst_dir, journal_seq);
}
bch2_inode_update_after_write(c, src_inode, &src_inode_u,
ATTR_CTIME);
- journal_seq_copy(src_inode, journal_seq);
+ journal_seq_copy(c, src_inode, journal_seq);
if (dst_inode) {
bch2_inode_update_after_write(c, dst_inode, &dst_inode_u,
ATTR_CTIME);
- journal_seq_copy(dst_inode, journal_seq);
+ journal_seq_copy(c, dst_inode, journal_seq);
}
err:
bch2_trans_exit(&trans);
return ret;
}
-void bch2_setattr_copy(struct bch_inode_info *inode,
- struct bch_inode_unpacked *bi,
- struct iattr *attr)
+static void bch2_setattr_copy(struct user_namespace *mnt_userns,
+ struct bch_inode_info *inode,
+ struct bch_inode_unpacked *bi,
+ struct iattr *attr)
{
struct bch_fs *c = inode->v.i_sb->s_fs_info;
unsigned int ia_valid = attr->ia_valid;
if (ia_valid & ATTR_UID)
- bi->bi_uid = from_kuid(c->vfs_sb->s_user_ns, attr->ia_uid);
+ bi->bi_uid = from_kuid(mnt_userns, attr->ia_uid);
if (ia_valid & ATTR_GID)
- bi->bi_gid = from_kgid(c->vfs_sb->s_user_ns, attr->ia_gid);
+ bi->bi_gid = from_kgid(mnt_userns, attr->ia_gid);
+
+ if (ia_valid & ATTR_SIZE)
+ bi->bi_size = attr->ia_size;
if (ia_valid & ATTR_ATIME)
bi->bi_atime = timespec_to_bch2_time(c, attr->ia_atime);
: inode->v.i_gid;
if (!in_group_p(gid) &&
- !capable_wrt_inode_uidgid(&inode->v, CAP_FSETID))
+ !capable_wrt_inode_uidgid(mnt_userns, &inode->v, CAP_FSETID))
mode &= ~S_ISGID;
bi->bi_mode = mode;
}
}
-static int bch2_setattr_nonsize(struct bch_inode_info *inode,
- struct iattr *attr)
+int bch2_setattr_nonsize(struct user_namespace *mnt_userns,
+ struct bch_inode_info *inode,
+ struct iattr *attr)
{
struct bch_fs *c = inode->v.i_sb->s_fs_info;
struct bch_qid qid;
if (ret)
goto btree_err;
- bch2_setattr_copy(inode, &inode_u, attr);
+ bch2_setattr_copy(mnt_userns, inode, &inode_u, attr);
if (attr->ia_valid & ATTR_MODE) {
- ret = bch2_acl_chmod(&trans, inode, inode_u.bi_mode, &acl);
+ ret = bch2_acl_chmod(&trans, &inode_u, inode_u.bi_mode, &acl);
if (ret)
goto btree_err;
}
ret = bch2_inode_write(&trans, inode_iter, &inode_u) ?:
bch2_trans_commit(&trans, NULL,
&inode->ei_journal_seq,
- BTREE_INSERT_NOUNLOCK|
BTREE_INSERT_NOFAIL);
btree_err:
+ bch2_trans_iter_put(&trans, inode_iter);
+
if (ret == -EINTR)
goto retry;
if (unlikely(ret))
return ret;
}
-static int bch2_getattr(const struct path *path, struct kstat *stat,
+static int bch2_getattr(struct user_namespace *mnt_userns,
+ const struct path *path, struct kstat *stat,
u32 request_mask, unsigned query_flags)
{
struct bch_inode_info *inode = to_bch_ei(d_inode(path->dentry));
return 0;
}
-static int bch2_setattr(struct dentry *dentry, struct iattr *iattr)
+static int bch2_setattr(struct user_namespace *mnt_userns,
+ struct dentry *dentry, struct iattr *iattr)
{
struct bch_inode_info *inode = to_bch_ei(dentry->d_inode);
int ret;
lockdep_assert_held(&inode->v.i_rwsem);
- ret = setattr_prepare(dentry, iattr);
+ ret = setattr_prepare(mnt_userns, dentry, iattr);
if (ret)
return ret;
return iattr->ia_valid & ATTR_SIZE
- ? bch2_truncate(inode, iattr)
- : bch2_setattr_nonsize(inode, iattr);
+ ? bch2_truncate(mnt_userns, inode, iattr)
+ : bch2_setattr_nonsize(mnt_userns, inode, iattr);
}
-static int bch2_tmpfile(struct inode *vdir, struct dentry *dentry, umode_t mode)
+static int bch2_tmpfile(struct user_namespace *mnt_userns,
+ struct inode *vdir, struct dentry *dentry, umode_t mode)
{
struct bch_inode_info *inode =
- __bch2_create(to_bch_ei(vdir), dentry, mode, 0, true);
+ __bch2_create(mnt_userns, to_bch_ei(vdir), dentry, mode, 0, true);
if (IS_ERR(inode))
return PTR_ERR(inode);
struct fiemap_extent_info *info,
struct bkey_s_c k, unsigned flags)
{
- if (bkey_extent_is_data(k.k)) {
+ if (bkey_extent_is_direct_data(k.k)) {
struct bkey_ptrs_c ptrs = bch2_bkey_ptrs_c(k);
const union bch_extent_entry *entry;
struct extent_ptr_decoded p;
}
return 0;
+ } else if (bkey_extent_is_inline_data(k.k)) {
+ return fiemap_fill_next_extent(info,
+ bkey_start_offset(k.k) << 9,
+ 0, k.k->size << 9,
+ flags|
+ FIEMAP_EXTENT_DATA_INLINE);
} else if (k.k->type == KEY_TYPE_reservation) {
return fiemap_fill_next_extent(info,
bkey_start_offset(k.k) << 9,
struct btree_trans trans;
struct btree_iter *iter;
struct bkey_s_c k;
- struct bkey_on_stack cur, prev;
+ struct bkey_buf cur, prev;
struct bpos end = POS(ei->v.i_ino, (start + len) >> 9);
unsigned offset_into_extent, sectors;
bool have_extent = false;
int ret = 0;
+ ret = fiemap_prep(&ei->v, info, start, &len, FIEMAP_FLAG_SYNC);
+ if (ret)
+ return ret;
+
if (start + len < start)
return -EINVAL;
- bkey_on_stack_init(&cur);
- bkey_on_stack_init(&prev);
+ bch2_bkey_buf_init(&cur);
+ bch2_bkey_buf_init(&prev);
bch2_trans_init(&trans, c, 0, 0);
- iter = bch2_trans_get_iter(&trans, BTREE_ID_EXTENTS,
+ iter = bch2_trans_get_iter(&trans, BTREE_ID_extents,
POS(ei->v.i_ino, start >> 9), 0);
retry:
+ bch2_trans_begin(&trans);
+
while ((k = bch2_btree_iter_peek(iter)).k &&
!(ret = bkey_err(k)) &&
bkey_cmp(iter->pos, end) < 0) {
+ enum btree_id data_btree = BTREE_ID_extents;
+
if (!bkey_extent_is_data(k.k) &&
k.k->type != KEY_TYPE_reservation) {
- bch2_btree_iter_next(iter);
+ bch2_btree_iter_advance(iter);
continue;
}
- bkey_on_stack_realloc(&cur, c, k.k->u64s);
- bkey_on_stack_realloc(&prev, c, k.k->u64s);
- bkey_reassemble(cur.k, k);
- k = bkey_i_to_s_c(cur.k);
-
offset_into_extent = iter->pos.offset -
bkey_start_offset(k.k);
sectors = k.k->size - offset_into_extent;
- ret = bch2_read_indirect_extent(&trans,
- &offset_into_extent, cur.k);
+ bch2_bkey_buf_reassemble(&cur, c, k);
+
+ ret = bch2_read_indirect_extent(&trans, &data_btree,
+ &offset_into_extent, &cur);
if (ret)
break;
+ k = bkey_i_to_s_c(cur.k);
+ bch2_bkey_buf_realloc(&prev, c, k.k->u64s);
+
sectors = min(sectors, k.k->size - offset_into_extent);
- if (offset_into_extent)
- bch2_cut_front(POS(k.k->p.inode,
- bkey_start_offset(k.k) +
- offset_into_extent),
- cur.k);
+ bch2_cut_front(POS(k.k->p.inode,
+ bkey_start_offset(k.k) +
+ offset_into_extent),
+ cur.k);
bch2_key_resize(&cur.k->k, sectors);
cur.k->k.p = iter->pos;
cur.k->k.p.offset += cur.k->k.size;
bkey_copy(prev.k, cur.k);
have_extent = true;
- if (k.k->type == KEY_TYPE_reflink_v)
- bch2_btree_iter_set_pos(iter, k.k->p);
- else
- bch2_btree_iter_next(iter);
+ bch2_btree_iter_set_pos(iter,
+ POS(iter->pos.inode, iter->pos.offset + sectors));
}
if (ret == -EINTR)
ret = bch2_fill_extent(c, info, bkey_i_to_s_c(prev.k),
FIEMAP_EXTENT_LAST);
+ bch2_trans_iter_put(&trans, iter);
ret = bch2_trans_exit(&trans) ?: ret;
- bkey_on_stack_exit(&cur, c);
- bkey_on_stack_exit(&prev, c);
+ bch2_bkey_buf_exit(&cur, c);
+ bch2_bkey_buf_exit(&prev, c);
return ret < 0 ? ret : 0;
}
.open = generic_file_open,
.fsync = bch2_fsync,
.splice_read = generic_file_splice_read,
- /*
- * Broken, on v5.3:
.splice_write = iter_file_splice_write,
- */
.fallocate = bch2_fallocate_dispatch,
.unlocked_ioctl = bch2_fs_file_ioctl,
#ifdef CONFIG_COMPAT
.writepage = bch2_writepage,
.readpage = bch2_readpage,
.writepages = bch2_writepages,
- .readpages = bch2_readpages,
+ .readahead = bch2_readahead,
.set_page_dirty = __set_page_dirty_nobuffers,
.write_begin = bch2_write_begin,
.write_end = bch2_write_end,
inode->v.i_generation = bi->bi_generation;
inode->v.i_size = bi->bi_size;
+ inode->ei_flags = 0;
inode->ei_journal_seq = 0;
inode->ei_quota_reserved = 0;
- inode->ei_str_hash = bch2_hash_info_init(c, bi);
inode->ei_qid = bch_qid(bi);
inode->v.i_mapping->a_ops = &bch_address_space_operations;
KEY_TYPE_QUOTA_WARN);
bch2_quota_acct(c, inode->ei_qid, Q_INO, -1,
KEY_TYPE_QUOTA_WARN);
- bch2_inode_rm(c, inode->v.i_ino);
+ bch2_inode_rm(c, inode->v.i_ino, true);
}
}
struct bch_fs *c = sb->s_fs_info;
struct bch_fs_usage_short usage = bch2_fs_usage_read_short(c);
unsigned shift = sb->s_blocksize_bits - 9;
+ /*
+ * this assumes inodes take up 64 bytes, which is a decent average
+ * number:
+ */
+ u64 avail_inodes = ((usage.capacity - usage.used) << 3);
u64 fsid;
buf->f_type = BCACHEFS_STATFS_MAGIC;
buf->f_bsize = sb->s_blocksize;
buf->f_blocks = usage.capacity >> shift;
- buf->f_bfree = (usage.capacity - usage.used) >> shift;
- buf->f_bavail = buf->f_bfree;
- buf->f_files = usage.nr_inodes;
- buf->f_ffree = U64_MAX;
+ buf->f_bfree = usage.free >> shift;
+ buf->f_bavail = avail_factor(usage.free) >> shift;
+
+ buf->f_files = usage.nr_inodes + avail_inodes;
+ buf->f_ffree = avail_inodes;
fsid = le64_to_cpup((void *) c->sb.user_uuid.b) ^
le64_to_cpup((void *) c->sb.user_uuid.b + sizeof(u64));
return bch2_journal_flush(&c->journal);
}
-static struct bch_fs *bch2_path_to_fs(const char *dev)
+static struct bch_fs *bch2_path_to_fs(const char *path)
{
struct bch_fs *c;
- struct block_device *bdev = lookup_bdev(dev);
-
- if (IS_ERR(bdev))
- return ERR_CAST(bdev);
-
- c = bch2_bdev_to_fs(bdev);
- bdput(bdev);
- return c ?: ERR_PTR(-ENOENT);
-}
-
-static struct bch_fs *__bch2_open_as_blockdevs(const char *dev_name, char * const *devs,
- unsigned nr_devs, struct bch_opts opts)
-{
- struct bch_fs *c, *c1, *c2;
- size_t i;
-
- if (!nr_devs)
- return ERR_PTR(-EINVAL);
-
- c = bch2_fs_open(devs, nr_devs, opts);
-
- if (IS_ERR(c) && PTR_ERR(c) == -EBUSY) {
- /*
- * Already open?
- * Look up each block device, make sure they all belong to a
- * filesystem and they all belong to the _same_ filesystem
- */
-
- c1 = bch2_path_to_fs(devs[0]);
- if (IS_ERR(c1))
- return c;
-
- for (i = 1; i < nr_devs; i++) {
- c2 = bch2_path_to_fs(devs[i]);
- if (!IS_ERR(c2))
- closure_put(&c2->cl);
-
- if (c1 != c2) {
- closure_put(&c1->cl);
- return c;
- }
- }
-
- c = c1;
- }
-
- if (IS_ERR(c))
- return c;
+ dev_t dev;
+ int ret;
- mutex_lock(&c->state_lock);
+ ret = lookup_bdev(path, &dev);
+ if (ret)
+ return ERR_PTR(ret);
- if (!test_bit(BCH_FS_STARTED, &c->flags)) {
- mutex_unlock(&c->state_lock);
+ c = bch2_dev_to_fs(dev);
+ if (c)
closure_put(&c->cl);
- pr_err("err mounting %s: incomplete filesystem", dev_name);
- return ERR_PTR(-EINVAL);
- }
-
- mutex_unlock(&c->state_lock);
-
- set_bit(BCH_FS_BDEV_MOUNTED, &c->flags);
- return c;
+ return c ?: ERR_PTR(-ENOENT);
}
-static struct bch_fs *bch2_open_as_blockdevs(const char *_dev_name,
- struct bch_opts opts)
+static char **split_devs(const char *_dev_name, unsigned *nr)
{
char *dev_name = NULL, **devs = NULL, *s;
- struct bch_fs *c = ERR_PTR(-ENOMEM);
size_t i, nr_devs = 0;
dev_name = kstrdup(_dev_name, GFP_KERNEL);
if (!dev_name)
- goto err;
+ return NULL;
for (s = dev_name; s; s = strchr(s + 1, ':'))
nr_devs++;
- devs = kcalloc(nr_devs, sizeof(const char *), GFP_KERNEL);
- if (!devs)
- goto err;
+ devs = kcalloc(nr_devs + 1, sizeof(const char *), GFP_KERNEL);
+ if (!devs) {
+ kfree(dev_name);
+ return NULL;
+ }
for (i = 0, s = dev_name;
s;
(s = strchr(s, ':')) && (*s++ = '\0'))
devs[i++] = s;
- c = __bch2_open_as_blockdevs(_dev_name, devs, nr_devs, opts);
-err:
- kfree(devs);
- kfree(dev_name);
- return c;
+ *nr = nr_devs;
+ return devs;
}
static int bch2_remount(struct super_block *sb, int *flags, char *data)
opt_set(opts, read_only, (*flags & SB_RDONLY) != 0);
- ret = bch2_parse_mount_opts(&opts, data);
+ ret = bch2_parse_mount_opts(c, &opts, data);
if (ret)
return ret;
if (opts.read_only != c->opts.read_only) {
- mutex_lock(&c->state_lock);
+ down_write(&c->state_lock);
if (opts.read_only) {
bch2_fs_read_only(c);
ret = bch2_fs_read_write(c);
if (ret) {
bch_err(c, "error going rw: %i", ret);
- mutex_unlock(&c->state_lock);
+ up_write(&c->state_lock);
return -EINVAL;
}
c->opts.read_only = opts.read_only;
- mutex_unlock(&c->state_lock);
+ up_write(&c->state_lock);
}
if (opts.errors >= 0)
return ret;
}
+static int bch2_show_devname(struct seq_file *seq, struct dentry *root)
+{
+ struct bch_fs *c = root->d_sb->s_fs_info;
+ struct bch_dev *ca;
+ unsigned i;
+ bool first = true;
+
+ for_each_online_member(ca, c, i) {
+ if (!first)
+ seq_putc(seq, ':');
+ first = false;
+ seq_puts(seq, "/dev/");
+ seq_puts(seq, ca->name);
+ }
+
+ return 0;
+}
+
static int bch2_show_options(struct seq_file *seq, struct dentry *root)
{
struct bch_fs *c = root->d_sb->s_fs_info;
}
return 0;
+}
+static void bch2_put_super(struct super_block *sb)
+{
+ struct bch_fs *c = sb->s_fs_info;
+
+ __bch2_fs_stop(c);
}
static const struct super_operations bch_super_operations = {
.evict_inode = bch2_evict_inode,
.sync_fs = bch2_sync_fs,
.statfs = bch2_statfs,
+ .show_devname = bch2_show_devname,
.show_options = bch2_show_options,
.remount_fs = bch2_remount,
-#if 0
.put_super = bch2_put_super,
+#if 0
.freeze_fs = bch2_freeze,
.unfreeze_fs = bch2_unfreeze,
#endif
};
-static int bch2_test_super(struct super_block *s, void *data)
-{
- return s->s_fs_info == data;
-}
-
static int bch2_set_super(struct super_block *s, void *data)
{
s->s_fs_info = data;
return 0;
}
+static int bch2_noset_super(struct super_block *s, void *data)
+{
+ return -EBUSY;
+}
+
+static int bch2_test_super(struct super_block *s, void *data)
+{
+ struct bch_fs *c = s->s_fs_info;
+ struct bch_fs **devs = data;
+ unsigned i;
+
+ if (!c)
+ return false;
+
+ for (i = 0; devs[i]; i++)
+ if (c != devs[i])
+ return false;
+ return true;
+}
+
static struct dentry *bch2_mount(struct file_system_type *fs_type,
int flags, const char *dev_name, void *data)
{
struct super_block *sb;
struct inode *vinode;
struct bch_opts opts = bch2_opts_empty();
- unsigned i;
+ char **devs;
+ struct bch_fs **devs_to_fs = NULL;
+ unsigned i, nr_devs;
int ret;
opt_set(opts, read_only, (flags & SB_RDONLY) != 0);
- ret = bch2_parse_mount_opts(&opts, data);
+ ret = bch2_parse_mount_opts(NULL, &opts, data);
if (ret)
return ERR_PTR(ret);
- c = bch2_open_as_blockdevs(dev_name, opts);
- if (IS_ERR(c))
- return ERR_CAST(c);
+ if (!dev_name || strlen(dev_name) == 0)
+ return ERR_PTR(-EINVAL);
- sb = sget(fs_type, bch2_test_super, bch2_set_super, flags|SB_NOSEC, c);
- if (IS_ERR(sb)) {
- closure_put(&c->cl);
- return ERR_CAST(sb);
+ devs = split_devs(dev_name, &nr_devs);
+ if (!devs)
+ return ERR_PTR(-ENOMEM);
+
+ devs_to_fs = kcalloc(nr_devs + 1, sizeof(void *), GFP_KERNEL);
+ if (!devs_to_fs) {
+ sb = ERR_PTR(-ENOMEM);
+ goto got_sb;
}
- BUG_ON(sb->s_fs_info != c);
+ for (i = 0; i < nr_devs; i++)
+ devs_to_fs[i] = bch2_path_to_fs(devs[i]);
- if (sb->s_root) {
- closure_put(&c->cl);
+ sb = sget(fs_type, bch2_test_super, bch2_noset_super,
+ flags|SB_NOSEC, devs_to_fs);
+ if (!IS_ERR(sb))
+ goto got_sb;
+
+ c = bch2_fs_open(devs, nr_devs, opts);
+ if (IS_ERR(c)) {
+ sb = ERR_CAST(c);
+ goto got_sb;
+ }
+
+ /* Some options can't be parsed until after the fs is started: */
+ ret = bch2_parse_mount_opts(c, &opts, data);
+ if (ret) {
+ bch2_fs_stop(c);
+ sb = ERR_PTR(ret);
+ goto got_sb;
+ }
+
+ bch2_opts_apply(&c->opts, opts);
+
+ sb = sget(fs_type, NULL, bch2_set_super, flags|SB_NOSEC, c);
+ if (IS_ERR(sb))
+ bch2_fs_stop(c);
+got_sb:
+ kfree(devs_to_fs);
+ kfree(devs[0]);
+ kfree(devs);
+
+ if (IS_ERR(sb))
+ return ERR_CAST(sb);
+ c = sb->s_fs_info;
+
+ if (sb->s_root) {
if ((flags ^ sb->s_flags) & SB_RDONLY) {
ret = -EBUSY;
goto err_put_super;
#endif
sb->s_xattr = bch2_xattr_handlers;
sb->s_magic = BCACHEFS_STATFS_MAGIC;
- sb->s_time_gran = c->sb.time_precision;
+ sb->s_time_gran = c->sb.nsec_per_time_unit;
+ sb->s_time_min = div_s64(S64_MIN, c->sb.time_units_per_sec) + 1;
+ sb->s_time_max = div_s64(S64_MAX, c->sb.time_units_per_sec);
c->vfs_sb = sb;
strlcpy(sb->s_id, c->name, sizeof(sb->s_id));
if (ret)
goto err_put_super;
- sb->s_bdi->congested_fn = bch2_congested;
- sb->s_bdi->congested_data = c;
sb->s_bdi->ra_pages = VM_READAHEAD_PAGES;
for_each_online_member(ca, c, i) {
break;
}
+ c->dev = sb->s_dev;
+
#ifdef CONFIG_BCACHEFS_POSIX_ACL
if (c->opts.acl)
sb->s_flags |= SB_POSIXACL;
struct bch_fs *c = sb->s_fs_info;
generic_shutdown_super(sb);
-
- if (test_bit(BCH_FS_BDEV_MOUNTED, &c->flags))
- bch2_fs_stop(c);
- else
- closure_put(&c->cl);
+ bch2_fs_free(c);
}
static struct file_system_type bcache_fs_type = {
projects / bcachefs-tools-debian / blobdiff