+// SPDX-License-Identifier: GPL-2.0
#ifndef NO_BCACHEFS_FS
#include "bcachefs.h"
#include "acl.h"
+#include "bkey_buf.h"
#include "btree_update.h"
#include "buckets.h"
#include "chardev.h"
#include "dirent.h"
#include "extents.h"
#include "fs.h"
+#include "fs-common.h"
#include "fs-io.h"
#include "fs-ioctl.h"
#include "fsck.h"
#include <linux/aio.h>
#include <linux/backing-dev.h>
#include <linux/exportfs.h>
+#include <linux/fiemap.h>
#include <linux/module.h>
#include <linux/posix_acl.h>
#include <linux/random.h>
struct bch_inode_info *,
struct bch_inode_unpacked *);
-/*
- * I_SIZE_DIRTY requires special handling:
- *
- * To the recovery code, the flag means that there is stale data past i_size
- * that needs to be deleted; it's used for implementing atomic appends and
- * truncates.
- *
- * On append, we set I_SIZE_DIRTY before doing the write, then after the write
- * we clear I_SIZE_DIRTY atomically with updating i_size to the new larger size
- * that exposes the data we just wrote.
- *
- * On truncate, it's the reverse: We set I_SIZE_DIRTY atomically with setting
- * i_size to the new smaller size, then we delete the data that we just made
- * invisible, and then we clear I_SIZE_DIRTY.
- *
- * Because there can be multiple appends in flight at a time, we need a refcount
- * (i_size_dirty_count) instead of manipulating the flag directly. Nonzero
- * refcount means I_SIZE_DIRTY is set, zero means it's cleared.
- *
- * Because write_inode() can be called at any time, i_size_dirty_count means
- * something different to the runtime code - it means to write_inode() "don't
- * update i_size yet".
- *
- * We don't clear I_SIZE_DIRTY directly, we let write_inode() clear it when
- * i_size_dirty_count is zero - but the reverse is not true, I_SIZE_DIRTY must
- * be set explicitly.
- */
-
-int __must_check __bch2_write_inode(struct bch_fs *c,
- struct bch_inode_info *inode,
- inode_set_fn set,
- void *p)
+static void journal_seq_copy(struct bch_fs *c,
+ struct bch_inode_info *dst,
+ u64 journal_seq)
{
- struct btree_iter iter;
- struct bch_inode_unpacked inode_u;
- struct bkey_inode_buf inode_p;
- u64 inum = inode->v.i_ino;
- unsigned i_nlink = READ_ONCE(inode->v.i_nlink);
- int ret;
-
/*
- * We can't write an inode with i_nlink == 0 because it's stored biased;
- * however, we don't need to because if i_nlink is 0 the inode is
- * getting deleted when it's evicted.
+ * atomic64_cmpxchg has a fallback for archs that don't support it,
+ * cmpxchg does not:
*/
- if (!i_nlink)
- return 0;
+ atomic64_t *dst_seq = (void *) &dst->ei_journal_seq;
+ u64 old, v = READ_ONCE(dst->ei_journal_seq);
- lockdep_assert_held(&inode->ei_update_lock);
+ do {
+ old = v;
- bch2_btree_iter_init(&iter, c, BTREE_ID_INODES, POS(inum, 0),
- BTREE_ITER_SLOTS|BTREE_ITER_INTENT);
+ if (old >= journal_seq)
+ break;
+ } 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)
+{
+ BUG_ON(atomic_long_read(&lock->v) == 0);
+
+ if (atomic_long_sub_return_release(i, &lock->v) == 0)
+ wake_up_all(&lock->wait);
+}
+
+static bool __pagecache_lock_tryget(struct pagecache_lock *lock, long i)
+{
+ long v = atomic_long_read(&lock->v), old;
do {
- struct bkey_s_c k = bch2_btree_iter_peek_slot(&iter);
+ old = v;
- if ((ret = btree_iter_err(k)))
- goto out;
+ if (i > 0 ? v < 0 : v > 0)
+ return false;
+ } while ((v = atomic_long_cmpxchg_acquire(&lock->v,
+ old, old + i)) != old);
+ return true;
+}
- if (WARN_ONCE(k.k->type != BCH_INODE_FS,
- "inode %llu not found when updating", inum)) {
- bch2_btree_iter_unlock(&iter);
- return -ENOENT;
- }
+static void __pagecache_lock_get(struct pagecache_lock *lock, long i)
+{
+ wait_event(lock->wait, __pagecache_lock_tryget(lock, i));
+}
- ret = bch2_inode_unpack(bkey_s_c_to_inode(k), &inode_u);
- if (WARN_ONCE(ret,
- "error %i unpacking inode %llu", ret, inum)) {
- ret = -ENOENT;
- break;
- }
+void bch2_pagecache_add_put(struct pagecache_lock *lock)
+{
+ __pagecache_lock_put(lock, 1);
+}
- if (set) {
- ret = set(inode, &inode_u, p);
- if (ret)
- goto out;
- }
+bool bch2_pagecache_add_tryget(struct pagecache_lock *lock)
+{
+ return __pagecache_lock_tryget(lock, 1);
+}
- BUG_ON(i_nlink < nlink_bias(inode->v.i_mode));
-
- inode_u.bi_mode = inode->v.i_mode;
- inode_u.bi_uid = i_uid_read(&inode->v);
- inode_u.bi_gid = i_gid_read(&inode->v);
- inode_u.bi_project = inode->ei_qid.q[QTYP_PRJ];
- inode_u.bi_nlink= i_nlink - nlink_bias(inode->v.i_mode);
- inode_u.bi_dev = inode->v.i_rdev;
- inode_u.bi_atime= timespec_to_bch2_time(c, inode->v.i_atime);
- inode_u.bi_mtime= timespec_to_bch2_time(c, inode->v.i_mtime);
- inode_u.bi_ctime= timespec_to_bch2_time(c, inode->v.i_ctime);
-
- bch2_inode_pack(&inode_p, &inode_u);
-
- ret = bch2_btree_insert_at(c, NULL, NULL,
- &inode->ei_journal_seq,
- BTREE_INSERT_ATOMIC|
- BTREE_INSERT_NOFAIL,
- BTREE_INSERT_ENTRY(&iter, &inode_p.inode.k_i));
- } while (ret == -EINTR);
-
- if (!ret) {
- inode->ei_inode = inode_u;
- inode->ei_qid = bch_qid(&inode_u);
- }
-out:
- bch2_btree_iter_unlock(&iter);
+void bch2_pagecache_add_get(struct pagecache_lock *lock)
+{
+ __pagecache_lock_get(lock, 1);
+}
- return ret < 0 ? ret : 0;
+void bch2_pagecache_block_put(struct pagecache_lock *lock)
+{
+ __pagecache_lock_put(lock, -1);
}
-int __must_check bch2_write_inode(struct bch_fs *c,
- struct bch_inode_info *inode)
+void bch2_pagecache_block_get(struct pagecache_lock *lock)
{
- return __bch2_write_inode(c, inode, NULL, NULL);
+ __pagecache_lock_get(lock, -1);
}
-static int bch2_inc_nlink(struct bch_fs *c, struct bch_inode_info *inode)
+void bch2_inode_update_after_write(struct bch_fs *c,
+ struct bch_inode_info *inode,
+ struct bch_inode_unpacked *bi,
+ unsigned fields)
{
+ set_nlink(&inode->v, bch2_inode_nlink_get(bi));
+ i_uid_write(&inode->v, bi->bi_uid);
+ i_gid_write(&inode->v, bi->bi_gid);
+ inode->v.i_mode = bi->bi_mode;
+
+ if (fields & ATTR_ATIME)
+ inode->v.i_atime = bch2_time_to_timespec(c, bi->bi_atime);
+ if (fields & ATTR_MTIME)
+ inode->v.i_mtime = bch2_time_to_timespec(c, bi->bi_mtime);
+ if (fields & ATTR_CTIME)
+ inode->v.i_ctime = bch2_time_to_timespec(c, bi->bi_ctime);
+
+ inode->ei_inode = *bi;
+
+ bch2_inode_flags_to_vfs(inode);
+}
+
+int __must_check bch2_write_inode(struct bch_fs *c,
+ struct bch_inode_info *inode,
+ inode_set_fn set,
+ void *p, unsigned fields)
+{
+ struct btree_trans trans;
+ struct btree_iter *iter;
+ struct bch_inode_unpacked inode_u;
int ret;
- mutex_lock(&inode->ei_update_lock);
- inc_nlink(&inode->v);
- ret = bch2_write_inode(c, inode);
- mutex_unlock(&inode->ei_update_lock);
+ bch2_trans_init(&trans, c, 0, 0);
+retry:
+ bch2_trans_begin(&trans);
- return ret;
+ iter = bch2_inode_peek(&trans, &inode_u, inode->v.i_ino,
+ BTREE_ITER_INTENT);
+ ret = PTR_ERR_OR_ZERO(iter) ?:
+ (set ? set(inode, &inode_u, p) : 0) ?:
+ bch2_inode_write(&trans, iter, &inode_u) ?:
+ bch2_trans_commit(&trans, NULL,
+ &inode->ei_journal_seq,
+ BTREE_INSERT_NOUNLOCK|
+ BTREE_INSERT_NOFAIL);
+
+ /*
+ * the btree node lock protects inode->ei_inode, not ei_update_lock;
+ * this is important for inode updates via bchfs_write_index_update
+ */
+ 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;
}
-static int bch2_dec_nlink(struct bch_fs *c, struct bch_inode_info *inode)
+int bch2_fs_quota_transfer(struct bch_fs *c,
+ struct bch_inode_info *inode,
+ struct bch_qid new_qid,
+ unsigned qtypes,
+ enum quota_acct_mode mode)
{
- int ret = 0;
+ unsigned i;
+ int ret;
- mutex_lock(&inode->ei_update_lock);
- drop_nlink(&inode->v);
- ret = bch2_write_inode(c, inode);
- mutex_unlock(&inode->ei_update_lock);
+ qtypes &= enabled_qtypes(c);
+
+ for (i = 0; i < QTYP_NR; i++)
+ if (new_qid.q[i] == inode->ei_qid.q[i])
+ qtypes &= ~(1U << i);
+
+ if (!qtypes)
+ return 0;
+
+ mutex_lock(&inode->ei_quota_lock);
+
+ ret = bch2_quota_transfer(c, qtypes, new_qid,
+ inode->ei_qid,
+ inode->v.i_blocks +
+ inode->ei_quota_reserved,
+ mode);
+ if (!ret)
+ for (i = 0; i < QTYP_NR; i++)
+ if (qtypes & (1 << i))
+ inode->ei_qid.q[i] = new_qid.q[i];
+
+ mutex_unlock(&inode->ei_quota_lock);
return ret;
}
-static struct inode *bch2_vfs_inode_get(struct bch_fs *c, u64 inum)
+struct inode *bch2_vfs_inode_get(struct bch_fs *c, u64 inum)
{
struct bch_inode_unpacked inode_u;
struct bch_inode_info *inode;
return &inode->v;
}
-static struct bch_inode_info *bch2_vfs_inode_create(struct bch_fs *c,
- struct bch_inode_info *dir,
- umode_t mode, dev_t rdev)
+static int inum_test(struct inode *inode, void *p)
{
- struct posix_acl *default_acl = NULL, *acl = NULL;
- struct bch_inode_info *inode;
- struct bch_inode_unpacked inode_u;
- int ret;
+ unsigned long *ino = p;
- inode = to_bch_ei(new_inode(c->vfs_sb));
- if (unlikely(!inode))
- return ERR_PTR(-ENOMEM);
+ return *ino == inode->i_ino;
+}
- inode_init_owner(&inode->v, &dir->v, mode);
+static struct bch_inode_info *
+__bch2_create(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;
+ struct bch_inode_unpacked inode_u;
+ struct posix_acl *default_acl = NULL, *acl = NULL;
+ u64 journal_seq = 0;
+ int ret;
+ /*
+ * preallocate acls + vfs inode before btree transaction, so that
+ * nothing can fail after the transaction succeeds:
+ */
#ifdef CONFIG_BCACHEFS_POSIX_ACL
- ret = posix_acl_create(&dir->v, &inode->v.i_mode, &default_acl, &acl);
+ ret = posix_acl_create(&dir->v, &mode, &default_acl, &acl);
if (ret)
- goto err_make_bad;
+ return ERR_PTR(ret);
#endif
+ inode = to_bch_ei(new_inode(c->vfs_sb));
+ if (unlikely(!inode)) {
+ inode = ERR_PTR(-ENOMEM);
+ goto err;
+ }
- bch2_inode_init(c, &inode_u,
- i_uid_read(&inode->v),
- i_gid_read(&inode->v),
- inode->v.i_mode, rdev,
- &dir->ei_inode);
+ bch2_inode_init_early(c, &inode_u);
- inode_u.bi_project = dir->ei_qid.q[QTYP_PRJ];
+ if (!tmpfile)
+ mutex_lock(&dir->ei_update_lock);
- ret = bch2_quota_acct(c, bch_qid(&inode_u), Q_INO, 1, BCH_QUOTA_PREALLOC);
- if (ret)
- goto err_make_bad;
+ bch2_trans_init(&trans, c, 8,
+ 2048 + (!tmpfile ? dentry->d_name.len : 0));
+retry:
+ bch2_trans_begin(&trans);
- ret = bch2_inode_create(c, &inode_u,
- BLOCKDEV_INODE_MAX, 0,
- &c->unused_inode_hint);
+ 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()),
+ mode, rdev,
+ default_acl, acl) ?:
+ bch2_quota_acct(c, bch_qid(&inode_u), Q_INO, 1,
+ KEY_TYPE_QUOTA_PREALLOC);
if (unlikely(ret))
- goto err_acct_quota;
-
- bch2_vfs_inode_init(c, inode, &inode_u);
- atomic_long_inc(&c->nr_inodes);
+ goto err_before_quota;
- if (default_acl) {
- ret = bch2_set_acl(&inode->v, default_acl, ACL_TYPE_DEFAULT);
- if (unlikely(ret))
- goto err;
+ ret = bch2_trans_commit(&trans, NULL, &journal_seq,
+ BTREE_INSERT_NOUNLOCK);
+ if (unlikely(ret)) {
+ bch2_quota_acct(c, bch_qid(&inode_u), Q_INO, -1,
+ KEY_TYPE_QUOTA_WARN);
+err_before_quota:
+ if (ret == -EINTR)
+ goto retry;
+ goto err_trans;
}
- if (acl) {
- ret = bch2_set_acl(&inode->v, acl, ACL_TYPE_ACCESS);
- if (unlikely(ret))
- goto err;
+ if (!tmpfile) {
+ bch2_inode_update_after_write(c, dir, &dir_u,
+ ATTR_MTIME|ATTR_CTIME);
+ journal_seq_copy(c, dir, journal_seq);
+ mutex_unlock(&dir->ei_update_lock);
}
- insert_inode_hash(&inode->v);
-out:
- posix_acl_release(default_acl);
- posix_acl_release(acl);
- return inode;
-err_acct_quota:
- bch2_quota_acct(c, bch_qid(&inode_u), Q_INO, -1, BCH_QUOTA_WARN);
-err_make_bad:
- /*
- * indicate to bch_evict_inode that the inode was never actually
- * created:
- */
- make_bad_inode(&inode->v);
-err:
- clear_nlink(&inode->v);
- iput(&inode->v);
- inode = ERR_PTR(ret);
- goto out;
-}
-
-static int bch2_vfs_dirent_create(struct bch_fs *c,
- struct bch_inode_info *dir,
- u8 type, const struct qstr *name,
- u64 dst)
-{
- int ret;
-
- ret = bch2_dirent_create(c, dir->v.i_ino, &dir->ei_str_hash,
- type, name, dst,
- &dir->ei_journal_seq,
- BCH_HASH_SET_MUST_CREATE);
- if (unlikely(ret))
- return ret;
+ bch2_vfs_inode_init(c, inode, &inode_u);
+ journal_seq_copy(c, inode, journal_seq);
- dir->v.i_mtime = dir->v.i_ctime = current_time(&dir->v);
- mark_inode_dirty_sync(&dir->v);
- return 0;
-}
+ set_cached_acl(&inode->v, ACL_TYPE_ACCESS, acl);
+ set_cached_acl(&inode->v, ACL_TYPE_DEFAULT, default_acl);
-static int __bch2_create(struct bch_inode_info *dir, struct dentry *dentry,
- umode_t mode, dev_t rdev)
-{
- struct bch_fs *c = dir->v.i_sb->s_fs_info;
- struct bch_inode_info *inode;
- int ret;
+ /*
+ * we must insert the new inode into the inode cache before calling
+ * bch2_trans_exit() and dropping locks, else we could race with another
+ * thread pulling the inode in and modifying it:
+ */
- inode = bch2_vfs_inode_create(c, dir, mode, rdev);
- if (unlikely(IS_ERR(inode)))
- return PTR_ERR(inode);
+ 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);
- ret = bch2_vfs_dirent_create(c, dir, mode_to_type(mode),
- &dentry->d_name, inode->v.i_ino);
- if (unlikely(ret)) {
- clear_nlink(&inode->v);
+ if (unlikely(old != inode)) {
+ /*
+ * We raced, another process pulled the new inode into cache
+ * before us:
+ */
+ journal_seq_copy(c, old, journal_seq);
+ make_bad_inode(&inode->v);
iput(&inode->v);
- return ret;
+
+ inode = old;
+ } else {
+ /*
+ * we really don't want insert_inode_locked2() to be setting
+ * I_NEW...
+ */
+ unlock_new_inode(&inode->v);
}
- if (dir->ei_journal_seq > inode->ei_journal_seq)
- inode->ei_journal_seq = dir->ei_journal_seq;
+ bch2_trans_exit(&trans);
+err:
+ posix_acl_release(default_acl);
+ posix_acl_release(acl);
+ return inode;
+err_trans:
+ if (!tmpfile)
+ mutex_unlock(&dir->ei_update_lock);
- d_instantiate(dentry, &inode->v);
- return 0;
+ bch2_trans_exit(&trans);
+ make_bad_inode(&inode->v);
+ iput(&inode->v);
+ inode = ERR_PTR(ret);
+ goto err;
}
/* methods */
{
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,
+ umode_t mode, dev_t rdev)
+{
+ struct bch_inode_info *inode =
+ __bch2_create(to_bch_ei(vdir), dentry, mode, rdev, false);
+
+ if (IS_ERR(inode))
+ return PTR_ERR(inode);
+
+ d_instantiate(dentry, &inode->v);
+ return 0;
+}
+
static int bch2_create(struct inode *vdir, struct dentry *dentry,
umode_t mode, bool excl)
{
- return __bch2_create(to_bch_ei(vdir), dentry, mode|S_IFREG, 0);
+ return bch2_mknod(vdir, dentry, mode|S_IFREG, 0);
+}
+
+static int __bch2_link(struct bch_fs *c,
+ struct bch_inode_info *inode,
+ struct bch_inode_info *dir,
+ struct dentry *dentry)
+{
+ struct btree_trans trans;
+ struct bch_inode_unpacked dir_u, inode_u;
+ int ret;
+
+ mutex_lock(&inode->ei_update_lock);
+ bch2_trans_init(&trans, c, 4, 1024);
+
+ ret = __bch2_trans_do(&trans, NULL, &inode->ei_journal_seq,
+ BTREE_INSERT_NOUNLOCK,
+ bch2_link_trans(&trans,
+ dir->v.i_ino,
+ inode->v.i_ino, &dir_u, &inode_u,
+ &dentry->d_name));
+
+ if (likely(!ret)) {
+ BUG_ON(inode_u.bi_inum != inode->v.i_ino);
+
+ 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_trans_exit(&trans);
+ mutex_unlock(&inode->ei_update_lock);
+ return ret;
}
static int bch2_link(struct dentry *old_dentry, struct inode *vdir,
lockdep_assert_held(&inode->v.i_rwsem);
- inode->v.i_ctime = current_time(&dir->v);
-
- ret = bch2_inc_nlink(c, inode);
- if (ret)
+ ret = __bch2_link(c, inode, dir, dentry);
+ if (unlikely(ret))
return ret;
ihold(&inode->v);
-
- ret = bch2_vfs_dirent_create(c, dir, mode_to_type(inode->v.i_mode),
- &dentry->d_name, inode->v.i_ino);
- if (unlikely(ret)) {
- bch2_dec_nlink(c, inode);
- iput(&inode->v);
- return ret;
- }
-
d_instantiate(dentry, &inode->v);
return 0;
}
struct bch_fs *c = vdir->i_sb->s_fs_info;
struct bch_inode_info *dir = to_bch_ei(vdir);
struct bch_inode_info *inode = to_bch_ei(dentry->d_inode);
+ struct bch_inode_unpacked dir_u, inode_u;
+ struct btree_trans trans;
int ret;
- lockdep_assert_held(&inode->v.i_rwsem);
-
- ret = bch2_dirent_delete(c, dir->v.i_ino, &dir->ei_str_hash,
- &dentry->d_name, &dir->ei_journal_seq);
- if (ret)
- return ret;
+ bch2_lock_inodes(INODE_UPDATE_LOCK, dir, inode);
+ bch2_trans_init(&trans, c, 4, 1024);
- if (dir->ei_journal_seq > inode->ei_journal_seq)
- inode->ei_journal_seq = dir->ei_journal_seq;
+ ret = __bch2_trans_do(&trans, NULL, &dir->ei_journal_seq,
+ BTREE_INSERT_NOUNLOCK|
+ BTREE_INSERT_NOFAIL,
+ bch2_unlink_trans(&trans,
+ dir->v.i_ino, &dir_u,
+ &inode_u, &dentry->d_name));
- inode->v.i_ctime = dir->v.i_ctime;
+ if (likely(!ret)) {
+ BUG_ON(inode_u.bi_inum != inode->v.i_ino);
- if (S_ISDIR(inode->v.i_mode)) {
- bch2_dec_nlink(c, dir);
- drop_nlink(&inode->v);
+ 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_MTIME);
}
- bch2_dec_nlink(c, inode);
+ bch2_trans_exit(&trans);
+ bch2_unlock_inodes(INODE_UPDATE_LOCK, dir, inode);
- return 0;
+ return ret;
}
static int bch2_symlink(struct inode *vdir, struct dentry *dentry,
struct bch_inode_info *dir = to_bch_ei(vdir), *inode;
int ret;
- inode = bch2_vfs_inode_create(c, dir, S_IFLNK|S_IRWXUGO, 0);
+ inode = __bch2_create(dir, dentry, S_IFLNK|S_IRWXUGO, 0, true);
if (unlikely(IS_ERR(inode)))
return PTR_ERR(inode);
if (unlikely(ret))
goto err;
- /* XXX: racy */
- if (dir->ei_journal_seq < inode->ei_journal_seq)
- dir->ei_journal_seq = inode->ei_journal_seq;
+ journal_seq_copy(c, dir, inode->ei_journal_seq);
- ret = bch2_vfs_dirent_create(c, dir, DT_LNK, &dentry->d_name,
- inode->v.i_ino);
+ ret = __bch2_link(c, inode, dir, dentry);
if (unlikely(ret))
goto err;
d_instantiate(dentry, &inode->v);
return 0;
err:
- clear_nlink(&inode->v);
iput(&inode->v);
return ret;
}
static int bch2_mkdir(struct inode *vdir, struct dentry *dentry, umode_t mode)
{
- struct bch_fs *c = vdir->i_sb->s_fs_info;
- struct bch_inode_info *dir = to_bch_ei(vdir);
- int ret;
-
- lockdep_assert_held(&dir->v.i_rwsem);
-
- ret = __bch2_create(dir, dentry, mode|S_IFDIR, 0);
- if (unlikely(ret))
- return ret;
-
- bch2_inc_nlink(c, dir);
-
- return 0;
-}
-
-static int bch2_rmdir(struct inode *vdir, struct dentry *dentry)
-{
- struct bch_fs *c = vdir->i_sb->s_fs_info;
-
- if (bch2_empty_dir(c, dentry->d_inode->i_ino))
- return -ENOTEMPTY;
-
- return bch2_unlink(vdir, dentry);
+ return bch2_mknod(vdir, dentry, mode|S_IFDIR, 0);
}
-static int bch2_mknod(struct inode *vdir, struct dentry *dentry,
- umode_t mode, dev_t rdev)
-{
- return __bch2_create(to_bch_ei(vdir), dentry, mode, rdev);
-}
-
-static int bch2_rename(struct bch_fs *c,
- struct bch_inode_info *old_dir,
- struct dentry *old_dentry,
- struct bch_inode_info *new_dir,
- struct dentry *new_dentry)
+static int bch2_rename2(struct inode *src_vdir, struct dentry *src_dentry,
+ struct inode *dst_vdir, struct dentry *dst_dentry,
+ unsigned flags)
{
- struct bch_inode_info *old_inode = to_bch_ei(old_dentry->d_inode);
- struct bch_inode_info *new_inode = to_bch_ei(new_dentry->d_inode);
- struct timespec now = current_time(&old_dir->v);
+ struct bch_fs *c = src_vdir->i_sb->s_fs_info;
+ struct bch_inode_info *src_dir = to_bch_ei(src_vdir);
+ struct bch_inode_info *dst_dir = to_bch_ei(dst_vdir);
+ struct bch_inode_info *src_inode = to_bch_ei(src_dentry->d_inode);
+ struct bch_inode_info *dst_inode = to_bch_ei(dst_dentry->d_inode);
+ struct bch_inode_unpacked dst_dir_u, src_dir_u;
+ struct bch_inode_unpacked src_inode_u, dst_inode_u;
+ struct btree_trans trans;
+ enum bch_rename_mode mode = flags & RENAME_EXCHANGE
+ ? BCH_RENAME_EXCHANGE
+ : dst_dentry->d_inode
+ ? BCH_RENAME_OVERWRITE : BCH_RENAME;
+ u64 journal_seq = 0;
int ret;
- lockdep_assert_held(&old_dir->v.i_rwsem);
- lockdep_assert_held(&new_dir->v.i_rwsem);
-
- if (new_inode)
- filemap_write_and_wait_range(old_inode->v.i_mapping,
- 0, LLONG_MAX);
-
- if (new_inode && S_ISDIR(old_inode->v.i_mode)) {
- lockdep_assert_held(&new_inode->v.i_rwsem);
-
- if (!S_ISDIR(new_inode->v.i_mode))
- return -ENOTDIR;
-
- if (bch2_empty_dir(c, new_inode->v.i_ino))
- return -ENOTEMPTY;
+ if (flags & ~(RENAME_NOREPLACE|RENAME_EXCHANGE))
+ return -EINVAL;
- ret = bch2_dirent_rename(c,
- old_dir, &old_dentry->d_name,
- new_dir, &new_dentry->d_name,
- &old_inode->ei_journal_seq, BCH_RENAME_OVERWRITE);
- if (unlikely(ret))
+ if (mode == BCH_RENAME_OVERWRITE) {
+ ret = filemap_write_and_wait_range(src_inode->v.i_mapping,
+ 0, LLONG_MAX);
+ if (ret)
return ret;
+ }
- clear_nlink(&new_inode->v);
- bch2_dec_nlink(c, old_dir);
- } else if (new_inode) {
- lockdep_assert_held(&new_inode->v.i_rwsem);
-
- ret = bch2_dirent_rename(c,
- old_dir, &old_dentry->d_name,
- new_dir, &new_dentry->d_name,
- &old_inode->ei_journal_seq, BCH_RENAME_OVERWRITE);
- if (unlikely(ret))
- return ret;
+ bch2_trans_init(&trans, c, 8, 2048);
- new_inode->v.i_ctime = now;
- bch2_dec_nlink(c, new_inode);
- } else if (S_ISDIR(old_inode->v.i_mode)) {
- ret = bch2_dirent_rename(c,
- old_dir, &old_dentry->d_name,
- new_dir, &new_dentry->d_name,
- &old_inode->ei_journal_seq, BCH_RENAME);
- if (unlikely(ret))
- return ret;
+ bch2_lock_inodes(INODE_UPDATE_LOCK,
+ src_dir,
+ dst_dir,
+ src_inode,
+ dst_inode);
- bch2_inc_nlink(c, new_dir);
- bch2_dec_nlink(c, old_dir);
- } else {
- ret = bch2_dirent_rename(c,
- old_dir, &old_dentry->d_name,
- new_dir, &new_dentry->d_name,
- &old_inode->ei_journal_seq, BCH_RENAME);
- if (unlikely(ret))
- return ret;
+ if (inode_attr_changing(dst_dir, src_inode, Inode_opt_project)) {
+ ret = bch2_fs_quota_transfer(c, src_inode,
+ dst_dir->ei_qid,
+ 1 << QTYP_PRJ,
+ KEY_TYPE_QUOTA_PREALLOC);
+ if (ret)
+ goto err;
}
- old_dir->v.i_ctime = old_dir->v.i_mtime = now;
- new_dir->v.i_ctime = new_dir->v.i_mtime = now;
- mark_inode_dirty_sync(&old_dir->v);
- mark_inode_dirty_sync(&new_dir->v);
-
- old_inode->v.i_ctime = now;
- mark_inode_dirty_sync(&old_inode->v);
+ if (mode == BCH_RENAME_EXCHANGE &&
+ inode_attr_changing(src_dir, dst_inode, Inode_opt_project)) {
+ ret = bch2_fs_quota_transfer(c, dst_inode,
+ src_dir->ei_qid,
+ 1 << QTYP_PRJ,
+ KEY_TYPE_QUOTA_PREALLOC);
+ if (ret)
+ goto err;
+ }
- return 0;
-}
+ ret = __bch2_trans_do(&trans, NULL, &journal_seq,
+ BTREE_INSERT_NOUNLOCK,
+ 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;
-static int bch2_rename_exchange(struct bch_fs *c,
- struct bch_inode_info *old_dir,
- struct dentry *old_dentry,
- struct bch_inode_info *new_dir,
- struct dentry *new_dentry)
-{
- struct bch_inode_info *old_inode = to_bch_ei(old_dentry->d_inode);
- struct bch_inode_info *new_inode = to_bch_ei(new_dentry->d_inode);
- struct timespec now = current_time(&old_dir->v);
- int ret;
+ BUG_ON(src_inode->v.i_ino != src_inode_u.bi_inum);
+ BUG_ON(dst_inode &&
+ dst_inode->v.i_ino != dst_inode_u.bi_inum);
- ret = bch2_dirent_rename(c,
- old_dir, &old_dentry->d_name,
- new_dir, &new_dentry->d_name,
- &old_inode->ei_journal_seq, BCH_RENAME_EXCHANGE);
- if (unlikely(ret))
- return ret;
+ bch2_inode_update_after_write(c, src_dir, &src_dir_u,
+ ATTR_MTIME|ATTR_CTIME);
+ journal_seq_copy(c, src_dir, journal_seq);
- if (S_ISDIR(old_inode->v.i_mode) !=
- S_ISDIR(new_inode->v.i_mode)) {
- if (S_ISDIR(old_inode->v.i_mode)) {
- bch2_inc_nlink(c, new_dir);
- bch2_dec_nlink(c, old_dir);
- } else {
- bch2_dec_nlink(c, new_dir);
- bch2_inc_nlink(c, old_dir);
- }
+ if (src_dir != dst_dir) {
+ bch2_inode_update_after_write(c, dst_dir, &dst_dir_u,
+ ATTR_MTIME|ATTR_CTIME);
+ journal_seq_copy(c, dst_dir, journal_seq);
}
- old_dir->v.i_ctime = old_dir->v.i_mtime = now;
- new_dir->v.i_ctime = new_dir->v.i_mtime = now;
- mark_inode_dirty_sync(&old_dir->v);
- mark_inode_dirty_sync(&new_dir->v);
+ bch2_inode_update_after_write(c, src_inode, &src_inode_u,
+ ATTR_CTIME);
+ journal_seq_copy(c, src_inode, journal_seq);
- old_inode->v.i_ctime = now;
- new_inode->v.i_ctime = now;
- mark_inode_dirty_sync(&old_inode->v);
- mark_inode_dirty_sync(&new_inode->v);
+ if (dst_inode) {
+ bch2_inode_update_after_write(c, dst_inode, &dst_inode_u,
+ ATTR_CTIME);
+ journal_seq_copy(c, dst_inode, journal_seq);
+ }
+err:
+ bch2_trans_exit(&trans);
+
+ bch2_fs_quota_transfer(c, src_inode,
+ bch_qid(&src_inode->ei_inode),
+ 1 << QTYP_PRJ,
+ KEY_TYPE_QUOTA_NOCHECK);
+ if (dst_inode)
+ bch2_fs_quota_transfer(c, dst_inode,
+ bch_qid(&dst_inode->ei_inode),
+ 1 << QTYP_PRJ,
+ KEY_TYPE_QUOTA_NOCHECK);
+
+ bch2_unlock_inodes(INODE_UPDATE_LOCK,
+ src_dir,
+ dst_dir,
+ src_inode,
+ dst_inode);
- return 0;
+ return ret;
}
-static int bch2_rename2(struct inode *old_vdir, struct dentry *old_dentry,
- struct inode *new_vdir, struct dentry *new_dentry,
- unsigned flags)
+void bch2_setattr_copy(struct bch_inode_info *inode,
+ struct bch_inode_unpacked *bi,
+ struct iattr *attr)
{
- struct bch_fs *c = old_vdir->i_sb->s_fs_info;
- struct bch_inode_info *old_dir = to_bch_ei(old_vdir);
- struct bch_inode_info *new_dir = to_bch_ei(new_vdir);
-
- if (flags & ~(RENAME_NOREPLACE|RENAME_EXCHANGE))
- return -EINVAL;
-
- if (flags & RENAME_EXCHANGE)
- return bch2_rename_exchange(c, old_dir, old_dentry,
- new_dir, new_dentry);
-
- return bch2_rename(c, old_dir, old_dentry, new_dir, new_dentry);
+ 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);
+ if (ia_valid & ATTR_GID)
+ bi->bi_gid = from_kgid(c->vfs_sb->s_user_ns, attr->ia_gid);
+
+ if (ia_valid & ATTR_ATIME)
+ bi->bi_atime = timespec_to_bch2_time(c, attr->ia_atime);
+ if (ia_valid & ATTR_MTIME)
+ bi->bi_mtime = timespec_to_bch2_time(c, attr->ia_mtime);
+ if (ia_valid & ATTR_CTIME)
+ bi->bi_ctime = timespec_to_bch2_time(c, attr->ia_ctime);
+
+ if (ia_valid & ATTR_MODE) {
+ umode_t mode = attr->ia_mode;
+ kgid_t gid = ia_valid & ATTR_GID
+ ? attr->ia_gid
+ : inode->v.i_gid;
+
+ if (!in_group_p(gid) &&
+ !capable_wrt_inode_uidgid(&inode->v, CAP_FSETID))
+ mode &= ~S_ISGID;
+ bi->bi_mode = mode;
+ }
}
-static int bch2_setattr_nonsize(struct bch_inode_info *inode, struct iattr *iattr)
+static int bch2_setattr_nonsize(struct bch_inode_info *inode,
+ struct iattr *attr)
{
struct bch_fs *c = inode->v.i_sb->s_fs_info;
- struct bch_qid qid = inode->ei_qid;
- unsigned qtypes = 0;
+ struct bch_qid qid;
+ struct btree_trans trans;
+ struct btree_iter *inode_iter;
+ struct bch_inode_unpacked inode_u;
+ struct posix_acl *acl = NULL;
int ret;
mutex_lock(&inode->ei_update_lock);
- if (c->opts.usrquota &&
- (iattr->ia_valid & ATTR_UID) &&
- !uid_eq(iattr->ia_uid, inode->v.i_uid)) {
- qid.q[QTYP_USR] = from_kuid(&init_user_ns, iattr->ia_uid),
- qtypes |= 1 << QTYP_USR;
- }
+ qid = inode->ei_qid;
- if (c->opts.grpquota &&
- (iattr->ia_valid & ATTR_GID) &&
- !gid_eq(iattr->ia_gid, inode->v.i_gid)) {
- qid.q[QTYP_GRP] = from_kgid(&init_user_ns, iattr->ia_gid);
- qtypes |= 1 << QTYP_GRP;
- }
+ if (attr->ia_valid & ATTR_UID)
+ qid.q[QTYP_USR] = from_kuid(&init_user_ns, attr->ia_uid);
+
+ if (attr->ia_valid & ATTR_GID)
+ qid.q[QTYP_GRP] = from_kgid(&init_user_ns, attr->ia_gid);
- if (qtypes) {
- ret = bch2_quota_transfer(c, qtypes, qid, inode->ei_qid,
- inode->v.i_blocks +
- inode->ei_quota_reserved);
+ ret = bch2_fs_quota_transfer(c, inode, qid, ~0,
+ KEY_TYPE_QUOTA_PREALLOC);
+ if (ret)
+ goto err;
+
+ bch2_trans_init(&trans, c, 0, 0);
+retry:
+ bch2_trans_begin(&trans);
+ kfree(acl);
+ acl = NULL;
+
+ inode_iter = bch2_inode_peek(&trans, &inode_u, inode->v.i_ino,
+ BTREE_ITER_INTENT);
+ ret = PTR_ERR_OR_ZERO(inode_iter);
+ if (ret)
+ goto btree_err;
+
+ bch2_setattr_copy(inode, &inode_u, attr);
+
+ if (attr->ia_valid & ATTR_MODE) {
+ ret = bch2_acl_chmod(&trans, &inode_u, inode_u.bi_mode, &acl);
if (ret)
- goto out_unlock;
+ goto btree_err;
}
- setattr_copy(&inode->v, iattr);
+ 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);
- ret = bch2_write_inode(c, inode);
-out_unlock:
- mutex_unlock(&inode->ei_update_lock);
+ if (ret == -EINTR)
+ goto retry;
+ if (unlikely(ret))
+ goto err_trans;
- if (!ret &&
- iattr->ia_valid & ATTR_MODE)
- ret = posix_acl_chmod(&inode->v, inode->v.i_mode);
+ bch2_inode_update_after_write(c, inode, &inode_u, attr->ia_valid);
+
+ if (acl)
+ set_cached_acl(&inode->v, ACL_TYPE_ACCESS, acl);
+err_trans:
+ bch2_trans_exit(&trans);
+err:
+ mutex_unlock(&inode->ei_update_lock);
return ret;
}
if (inode->ei_inode.bi_flags & BCH_INODE_IMMUTABLE)
stat->attributes |= STATX_ATTR_IMMUTABLE;
+ stat->attributes_mask |= STATX_ATTR_IMMUTABLE;
+
if (inode->ei_inode.bi_flags & BCH_INODE_APPEND)
stat->attributes |= STATX_ATTR_APPEND;
+ stat->attributes_mask |= STATX_ATTR_APPEND;
+
if (inode->ei_inode.bi_flags & BCH_INODE_NODUMP)
stat->attributes |= STATX_ATTR_NODUMP;
+ stat->attributes_mask |= STATX_ATTR_NODUMP;
return 0;
}
static int bch2_tmpfile(struct inode *vdir, struct dentry *dentry, umode_t mode)
{
- struct bch_fs *c = vdir->i_sb->s_fs_info;
- struct bch_inode_info *dir = to_bch_ei(vdir);
- struct bch_inode_info *inode;
+ struct bch_inode_info *inode =
+ __bch2_create(to_bch_ei(vdir), dentry, mode, 0, true);
- /* XXX: i_nlink should be 0? */
- inode = bch2_vfs_inode_create(c, dir, mode, 0);
- if (unlikely(IS_ERR(inode)))
+ if (IS_ERR(inode))
return PTR_ERR(inode);
- d_tmpfile(dentry, &inode->v);
+ d_mark_tmpfile(dentry, &inode->v);
+ d_instantiate(dentry, &inode->v);
return 0;
}
-static int bch2_fill_extent(struct fiemap_extent_info *info,
- const struct bkey_i *k, unsigned flags)
+static int bch2_fill_extent(struct bch_fs *c,
+ struct fiemap_extent_info *info,
+ struct bkey_s_c k, unsigned flags)
{
- if (bkey_extent_is_data(&k->k)) {
- struct bkey_s_c_extent e = bkey_i_to_s_c_extent(k);
- const struct bch_extent_ptr *ptr;
- struct bch_extent_crc_unpacked crc;
+ 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;
int ret;
- extent_for_each_ptr_crc(e, ptr, crc) {
+ if (k.k->type == KEY_TYPE_reflink_v)
+ flags |= FIEMAP_EXTENT_SHARED;
+
+ bkey_for_each_ptr_decode(k.k, ptrs, p, entry) {
int flags2 = 0;
- u64 offset = ptr->offset;
+ u64 offset = p.ptr.offset;
- if (crc.compression_type)
+ if (p.crc.compression_type)
flags2 |= FIEMAP_EXTENT_ENCODED;
else
- offset += crc.offset;
+ offset += p.crc.offset;
- if ((offset & (PAGE_SECTORS - 1)) ||
- (e.k->size & (PAGE_SECTORS - 1)))
+ if ((offset & (c->opts.block_size - 1)) ||
+ (k.k->size & (c->opts.block_size - 1)))
flags2 |= FIEMAP_EXTENT_NOT_ALIGNED;
ret = fiemap_fill_next_extent(info,
- bkey_start_offset(e.k) << 9,
- offset << 9,
- e.k->size << 9, flags|flags2);
+ bkey_start_offset(k.k) << 9,
+ offset << 9,
+ k.k->size << 9, flags|flags2);
if (ret)
return ret;
}
return 0;
- } else if (k->k.type == BCH_RESERVATION) {
+ } 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,
- 0, k->k.size << 9,
+ bkey_start_offset(k.k) << 9,
+ 0, k.k->size << 9,
flags|
FIEMAP_EXTENT_DELALLOC|
FIEMAP_EXTENT_UNWRITTEN);
{
struct bch_fs *c = vinode->i_sb->s_fs_info;
struct bch_inode_info *ei = to_bch_ei(vinode);
- struct btree_iter iter;
+ struct btree_trans trans;
+ struct btree_iter *iter;
struct bkey_s_c k;
- BKEY_PADDED(k) tmp;
+ 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;
- for_each_btree_key(&iter, c, BTREE_ID_EXTENTS,
- POS(ei->v.i_ino, start >> 9), 0, k)
- if (bkey_extent_is_data(k.k) ||
- k.k->type == BCH_RESERVATION) {
- if (bkey_cmp(bkey_start_pos(k.k),
- POS(ei->v.i_ino, (start + len) >> 9)) >= 0)
- break;
+ 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,
+ POS(ei->v.i_ino, start >> 9), 0);
+retry:
+ 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);
+ continue;
+ }
- if (have_extent) {
- ret = bch2_fill_extent(info, &tmp.k, 0);
- if (ret)
- goto out;
- }
+ offset_into_extent = iter->pos.offset -
+ bkey_start_offset(k.k);
+ sectors = k.k->size - offset_into_extent;
+
+ 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);
- bkey_reassemble(&tmp.k, k);
- have_extent = true;
+ sectors = min(sectors, k.k->size - offset_into_extent);
+
+ 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;
+
+ if (have_extent) {
+ ret = bch2_fill_extent(c, info,
+ bkey_i_to_s_c(prev.k), 0);
+ if (ret)
+ break;
}
- if (have_extent)
- ret = bch2_fill_extent(info, &tmp.k, FIEMAP_EXTENT_LAST);
-out:
- bch2_btree_iter_unlock(&iter);
+ bkey_copy(prev.k, cur.k);
+ have_extent = true;
+
+ bch2_btree_iter_set_pos(iter,
+ POS(iter->pos.inode, iter->pos.offset + sectors));
+ }
+
+ if (ret == -EINTR)
+ goto retry;
+
+ if (!ret && have_extent)
+ 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;
+ bch2_bkey_buf_exit(&cur, c);
+ bch2_bkey_buf_exit(&prev, c);
return ret < 0 ? ret : 0;
}
static const struct vm_operations_struct bch_vm_ops = {
- .fault = filemap_fault,
+ .fault = bch2_page_fault,
.map_pages = filemap_map_pages,
.page_mkwrite = bch2_page_mkwrite,
};
static int bch2_vfs_readdir(struct file *file, struct dir_context *ctx)
{
- struct bch_fs *c = file_inode(file)->i_sb->s_fs_info;
+ struct bch_inode_info *inode = file_bch_inode(file);
+ struct bch_fs *c = inode->v.i_sb->s_fs_info;
+
+ if (!dir_emit_dots(file, ctx))
+ return 0;
- return bch2_readdir(c, file, ctx);
+ return bch2_readdir(c, inode->v.i_ino, ctx);
}
static const struct file_operations bch_file_operations = {
.llseek = bch2_llseek,
- .read_iter = generic_file_read_iter,
+ .read_iter = bch2_read_iter,
.write_iter = bch2_write_iter,
.mmap = bch2_mmap,
.open = generic_file_open,
.fsync = bch2_fsync,
.splice_read = generic_file_splice_read,
+#if 0
+ /* Busted: */
.splice_write = iter_file_splice_write,
+#endif
.fallocate = bch2_fallocate_dispatch,
.unlocked_ioctl = bch2_fs_file_ioctl,
#ifdef CONFIG_COMPAT
.compat_ioctl = bch2_compat_fs_ioctl,
#endif
+ .remap_file_range = bch2_remap_file_range,
};
static const struct inode_operations bch_file_inode_operations = {
.unlink = bch2_unlink,
.symlink = bch2_symlink,
.mkdir = bch2_mkdir,
- .rmdir = bch2_rmdir,
+ .rmdir = bch2_unlink,
.mknod = bch2_mknod,
.rename = bch2_rename2,
.getattr = bch2_getattr,
static const struct file_operations bch_dir_file_operations = {
.llseek = bch2_dir_llseek,
.read = generic_read_dir,
- .iterate = bch2_vfs_readdir,
+ .iterate_shared = bch2_vfs_readdir,
.fsync = bch2_fsync,
.unlocked_ioctl = bch2_fs_file_ioctl,
#ifdef CONFIG_COMPAT
.writepage = bch2_writepage,
.readpage = bch2_readpage,
.writepages = bch2_writepages,
- .readpages = bch2_readpages,
- .set_page_dirty = bch2_set_page_dirty,
+ .readahead = bch2_readahead,
+ .set_page_dirty = __set_page_dirty_nobuffers,
.write_begin = bch2_write_begin,
.write_end = bch2_write_end,
.invalidatepage = bch2_invalidatepage,
.releasepage = bch2_releasepage,
- .direct_IO = bch2_direct_IO,
+ .direct_IO = noop_direct_IO,
#ifdef CONFIG_MIGRATION
.migratepage = bch2_migrate_page,
#endif
struct bch_inode_info *inode,
struct bch_inode_unpacked *bi)
{
- inode->v.i_mode = bi->bi_mode;
- i_uid_write(&inode->v, bi->bi_uid);
- i_gid_write(&inode->v, bi->bi_gid);
+ bch2_inode_update_after_write(c, inode, bi, ~0);
+
inode->v.i_blocks = bi->bi_sectors;
inode->v.i_ino = bi->bi_inum;
- set_nlink(&inode->v, bi->bi_nlink + nlink_bias(inode->v.i_mode));
inode->v.i_rdev = bi->bi_dev;
inode->v.i_generation = bi->bi_generation;
inode->v.i_size = bi->bi_size;
- inode->v.i_atime = bch2_time_to_timespec(c, bi->bi_atime);
- inode->v.i_mtime = bch2_time_to_timespec(c, bi->bi_mtime);
- inode->v.i_ctime = bch2_time_to_timespec(c, bi->bi_ctime);
+ inode->ei_flags = 0;
inode->ei_journal_seq = 0;
inode->ei_quota_reserved = 0;
inode->ei_qid = bch_qid(bi);
- inode->ei_str_hash = bch2_hash_info_init(c, bi);
- inode->ei_inode = *bi;
-
- bch2_inode_flags_to_vfs(inode);
inode->v.i_mapping->a_ops = &bch_address_space_operations;
inode_init_once(&inode->v);
mutex_init(&inode->ei_update_lock);
+ pagecache_lock_init(&inode->ei_pagecache_lock);
+ mutex_init(&inode->ei_quota_lock);
inode->ei_journal_seq = 0;
return &inode->v;
call_rcu(&vinode->i_rcu, bch2_i_callback);
}
+static int inode_update_times_fn(struct bch_inode_info *inode,
+ struct bch_inode_unpacked *bi,
+ void *p)
+{
+ struct bch_fs *c = inode->v.i_sb->s_fs_info;
+
+ bi->bi_atime = timespec_to_bch2_time(c, inode->v.i_atime);
+ bi->bi_mtime = timespec_to_bch2_time(c, inode->v.i_mtime);
+ bi->bi_ctime = timespec_to_bch2_time(c, inode->v.i_ctime);
+
+ return 0;
+}
+
static int bch2_vfs_write_inode(struct inode *vinode,
struct writeback_control *wbc)
{
int ret;
mutex_lock(&inode->ei_update_lock);
- ret = bch2_write_inode(c, inode);
+ ret = bch2_write_inode(c, inode, inode_update_times_fn, NULL,
+ ATTR_ATIME|ATTR_MTIME|ATTR_CTIME);
mutex_unlock(&inode->ei_update_lock);
- if (c->opts.journal_flush_disabled)
- return ret;
-
- if (!ret && wbc->sync_mode == WB_SYNC_ALL)
- ret = bch2_journal_flush_seq(&c->journal, inode->ei_journal_seq);
-
return ret;
}
if (!inode->v.i_nlink && !is_bad_inode(&inode->v)) {
bch2_quota_acct(c, inode->ei_qid, Q_SPC, -((s64) inode->v.i_blocks),
- BCH_QUOTA_WARN);
+ KEY_TYPE_QUOTA_WARN);
bch2_quota_acct(c, inode->ei_qid, Q_INO, -1,
- BCH_QUOTA_WARN);
- bch2_inode_rm(c, inode->v.i_ino);
- atomic_long_dec(&c->nr_inodes);
+ KEY_TYPE_QUOTA_WARN);
+ bch2_inode_rm(c, inode->v.i_ino, true);
}
}
{
struct super_block *sb = dentry->d_sb;
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 = c->capacity >> PAGE_SECTOR_SHIFT;
- buf->f_bfree = bch2_fs_sectors_free(c, bch2_fs_usage_read(c)) >>
- PAGE_SECTOR_SHIFT;
+ buf->f_blocks = usage.capacity >> shift;
+ buf->f_bfree = (usage.capacity - usage.used) >> shift;
buf->f_bavail = buf->f_bfree;
- buf->f_files = atomic_long_read(&c->nr_inodes);
- buf->f_ffree = U64_MAX;
+
+ 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));
buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
- buf->f_namelen = NAME_MAX;
+ buf->f_namelen = BCH_NAME_MAX;
return 0;
}
{
struct bch_fs *c = sb->s_fs_info;
+ if (c->opts.journal_flush_disabled)
+ return 0;
+
if (!wait) {
bch2_journal_flush_async(&c->journal, NULL);
return 0;
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 (!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;
-
- mutex_lock(&c->state_lock);
-
- if (!bch2_fs_running(c)) {
- mutex_unlock(&c->state_lock);
+ 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)
struct bch_opts opts = bch2_opts_empty();
int ret;
- opt_set(opts, read_only, (*flags & MS_RDONLY) != 0);
+ 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) {
- const char *err = NULL;
-
- mutex_lock(&c->state_lock);
+ down_write(&c->state_lock);
if (opts.read_only) {
bch2_fs_read_only(c);
- sb->s_flags |= MS_RDONLY;
+ sb->s_flags |= SB_RDONLY;
} else {
- err = bch2_fs_read_write(c);
- if (err) {
- bch_err(c, "error going rw: %s", err);
+ ret = bch2_fs_read_write(c);
+ if (ret) {
+ bch_err(c, "error going rw: %i", ret);
+ up_write(&c->state_lock);
return -EINVAL;
}
- sb->s_flags &= ~MS_RDONLY;
+ sb->s_flags &= ~SB_RDONLY;
}
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;
const struct bch_option *opt = &bch2_opt_table[i];
u64 v = bch2_opt_get_by_id(&c->opts, i);
- if (opt->mode < OPT_MOUNT)
+ if (!(opt->mode & OPT_MOUNT))
continue;
if (v == bch2_opt_get_by_id(&bch2_opts_default, i))
continue;
- bch2_opt_to_text(c, buf, sizeof(buf), opt, v,
+ bch2_opt_to_text(&PBUF(buf), c, opt, v,
OPT_SHOW_MOUNT_STYLE);
seq_putc(seq, ',');
seq_puts(seq, buf);
}
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 & MS_RDONLY) != 0);
+ 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);
+ devs = split_devs(dev_name, &nr_devs);
+ if (!devs)
+ return ERR_PTR(-ENOMEM);
- sb = sget(fs_type, bch2_test_super, bch2_set_super, flags|MS_NOSEC, c);
- if (IS_ERR(sb)) {
- closure_put(&c->cl);
- return ERR_CAST(sb);
+ 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 ((flags ^ sb->s_flags) & MS_RDONLY) {
+ if (sb->s_root) {
+ if ((flags ^ sb->s_flags) & SB_RDONLY) {
ret = -EBUSY;
goto err_put_super;
}
goto out;
}
- /* XXX: blocksize */
- sb->s_blocksize = PAGE_SIZE;
- sb->s_blocksize_bits = PAGE_SHIFT;
+ sb->s_blocksize = block_bytes(c);
+ sb->s_blocksize_bits = ilog2(block_bytes(c));
sb->s_maxbytes = MAX_LFS_FILESIZE;
sb->s_op = &bch_super_operations;
sb->s_export_op = &bch_export_ops;
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_MAX_READAHEAD * 1024 / PAGE_SIZE;
+ sb->s_bdi->ra_pages = VM_READAHEAD_PAGES;
for_each_online_member(ca, c, i) {
struct block_device *bdev = ca->disk_sb.bdev;
#ifdef CONFIG_BCACHEFS_POSIX_ACL
if (c->opts.acl)
- sb->s_flags |= MS_POSIXACL;
+ sb->s_flags |= SB_POSIXACL;
#endif
vinode = bch2_vfs_inode_get(c, BCACHEFS_ROOT_INO);
if (IS_ERR(vinode)) {
+ bch_err(c, "error mounting: error getting root inode %i",
+ (int) PTR_ERR(vinode));
ret = PTR_ERR(vinode);
goto err_put_super;
}
sb->s_root = d_make_root(vinode);
if (!sb->s_root) {
+ bch_err(c, "error mounting: error allocating root dentry");
ret = -ENOMEM;
goto err_put_super;
}
- sb->s_flags |= MS_ACTIVE;
+ sb->s_flags |= SB_ACTIVE;
out:
return dget(sb->s_root);
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 = {