Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / fs / ext4 / ext4_jbd2.h
diff --git a/kernel/fs/ext4/ext4_jbd2.h b/kernel/fs/ext4/ext4_jbd2.h
new file mode 100644 (file)
index 0000000..9c5b49f
--- /dev/null
@@ -0,0 +1,450 @@
+/*
+ * ext4_jbd2.h
+ *
+ * Written by Stephen C. Tweedie <sct@redhat.com>, 1999
+ *
+ * Copyright 1998--1999 Red Hat corp --- All Rights Reserved
+ *
+ * This file is part of the Linux kernel and is made available under
+ * the terms of the GNU General Public License, version 2, or at your
+ * option, any later version, incorporated herein by reference.
+ *
+ * Ext4-specific journaling extensions.
+ */
+
+#ifndef _EXT4_JBD2_H
+#define _EXT4_JBD2_H
+
+#include <linux/fs.h>
+#include <linux/jbd2.h>
+#include "ext4.h"
+
+#define EXT4_JOURNAL(inode)    (EXT4_SB((inode)->i_sb)->s_journal)
+
+/* Define the number of blocks we need to account to a transaction to
+ * modify one block of data.
+ *
+ * We may have to touch one inode, one bitmap buffer, up to three
+ * indirection blocks, the group and superblock summaries, and the data
+ * block to complete the transaction.
+ *
+ * For extents-enabled fs we may have to allocate and modify up to
+ * 5 levels of tree, data block (for each of these we need bitmap + group
+ * summaries), root which is stored in the inode, sb
+ */
+
+#define EXT4_SINGLEDATA_TRANS_BLOCKS(sb)                               \
+       (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)   \
+        ? 20U : 8U)
+
+/* Extended attribute operations touch at most two data buffers,
+ * two bitmap buffers, and two group summaries, in addition to the inode
+ * and the superblock, which are already accounted for. */
+
+#define EXT4_XATTR_TRANS_BLOCKS                6U
+
+/* Define the minimum size for a transaction which modifies data.  This
+ * needs to take into account the fact that we may end up modifying two
+ * quota files too (one for the group, one for the user quota).  The
+ * superblock only gets updated once, of course, so don't bother
+ * counting that again for the quota updates. */
+
+#define EXT4_DATA_TRANS_BLOCKS(sb)     (EXT4_SINGLEDATA_TRANS_BLOCKS(sb) + \
+                                        EXT4_XATTR_TRANS_BLOCKS - 2 + \
+                                        EXT4_MAXQUOTAS_TRANS_BLOCKS(sb))
+
+/*
+ * Define the number of metadata blocks we need to account to modify data.
+ *
+ * This include super block, inode block, quota blocks and xattr blocks
+ */
+#define EXT4_META_TRANS_BLOCKS(sb)     (EXT4_XATTR_TRANS_BLOCKS + \
+                                       EXT4_MAXQUOTAS_TRANS_BLOCKS(sb))
+
+/* Define an arbitrary limit for the amount of data we will anticipate
+ * writing to any given transaction.  For unbounded transactions such as
+ * write(2) and truncate(2) we can write more than this, but we always
+ * start off at the maximum transaction size and grow the transaction
+ * optimistically as we go. */
+
+#define EXT4_MAX_TRANS_DATA            64U
+
+/* We break up a large truncate or write transaction once the handle's
+ * buffer credits gets this low, we need either to extend the
+ * transaction or to start a new one.  Reserve enough space here for
+ * inode, bitmap, superblock, group and indirection updates for at least
+ * one block, plus two quota updates.  Quota allocations are not
+ * needed. */
+
+#define EXT4_RESERVE_TRANS_BLOCKS      12U
+
+#define EXT4_INDEX_EXTRA_TRANS_BLOCKS  8
+
+#ifdef CONFIG_QUOTA
+/* Amount of blocks needed for quota update - we know that the structure was
+ * allocated so we need to update only data block */
+#define EXT4_QUOTA_TRANS_BLOCKS(sb) ((test_opt(sb, QUOTA) ||\
+               EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) ?\
+               1 : 0)
+/* Amount of blocks needed for quota insert/delete - we do some block writes
+ * but inode, sb and group updates are done only once */
+#define EXT4_QUOTA_INIT_BLOCKS(sb) ((test_opt(sb, QUOTA) ||\
+               EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) ?\
+               (DQUOT_INIT_ALLOC*(EXT4_SINGLEDATA_TRANS_BLOCKS(sb)-3)\
+                +3+DQUOT_INIT_REWRITE) : 0)
+
+#define EXT4_QUOTA_DEL_BLOCKS(sb) ((test_opt(sb, QUOTA) ||\
+               EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) ?\
+               (DQUOT_DEL_ALLOC*(EXT4_SINGLEDATA_TRANS_BLOCKS(sb)-3)\
+                +3+DQUOT_DEL_REWRITE) : 0)
+#else
+#define EXT4_QUOTA_TRANS_BLOCKS(sb) 0
+#define EXT4_QUOTA_INIT_BLOCKS(sb) 0
+#define EXT4_QUOTA_DEL_BLOCKS(sb) 0
+#endif
+#define EXT4_MAXQUOTAS_TRANS_BLOCKS(sb) (EXT4_MAXQUOTAS*EXT4_QUOTA_TRANS_BLOCKS(sb))
+#define EXT4_MAXQUOTAS_INIT_BLOCKS(sb) (EXT4_MAXQUOTAS*EXT4_QUOTA_INIT_BLOCKS(sb))
+#define EXT4_MAXQUOTAS_DEL_BLOCKS(sb) (EXT4_MAXQUOTAS*EXT4_QUOTA_DEL_BLOCKS(sb))
+
+static inline int ext4_jbd2_credits_xattr(struct inode *inode)
+{
+       int credits = EXT4_DATA_TRANS_BLOCKS(inode->i_sb);
+
+       /*
+        * In case of inline data, we may push out the data to a block,
+        * so we need to reserve credits for this eventuality
+        */
+       if (ext4_has_inline_data(inode))
+               credits += ext4_writepage_trans_blocks(inode) + 1;
+       return credits;
+}
+
+
+/*
+ * Ext4 handle operation types -- for logging purposes
+ */
+#define EXT4_HT_MISC             0
+#define EXT4_HT_INODE            1
+#define EXT4_HT_WRITE_PAGE       2
+#define EXT4_HT_MAP_BLOCKS       3
+#define EXT4_HT_DIR              4
+#define EXT4_HT_TRUNCATE         5
+#define EXT4_HT_QUOTA            6
+#define EXT4_HT_RESIZE           7
+#define EXT4_HT_MIGRATE          8
+#define EXT4_HT_MOVE_EXTENTS     9
+#define EXT4_HT_XATTR           10
+#define EXT4_HT_EXT_CONVERT     11
+#define EXT4_HT_MAX             12
+
+/**
+ *   struct ext4_journal_cb_entry - Base structure for callback information.
+ *
+ *   This struct is a 'seed' structure for a using with your own callback
+ *   structs. If you are using callbacks you must allocate one of these
+ *   or another struct of your own definition which has this struct
+ *   as it's first element and pass it to ext4_journal_callback_add().
+ */
+struct ext4_journal_cb_entry {
+       /* list information for other callbacks attached to the same handle */
+       struct list_head jce_list;
+
+       /*  Function to call with this callback structure */
+       void (*jce_func)(struct super_block *sb,
+                        struct ext4_journal_cb_entry *jce, int error);
+
+       /* user data goes here */
+};
+
+/**
+ * ext4_journal_callback_add: add a function to call after transaction commit
+ * @handle: active journal transaction handle to register callback on
+ * @func: callback function to call after the transaction has committed:
+ *        @sb: superblock of current filesystem for transaction
+ *        @jce: returned journal callback data
+ *        @rc: journal state at commit (0 = transaction committed properly)
+ * @jce: journal callback data (internal and function private data struct)
+ *
+ * The registered function will be called in the context of the journal thread
+ * after the transaction for which the handle was created has completed.
+ *
+ * No locks are held when the callback function is called, so it is safe to
+ * call blocking functions from within the callback, but the callback should
+ * not block or run for too long, or the filesystem will be blocked waiting for
+ * the next transaction to commit. No journaling functions can be used, or
+ * there is a risk of deadlock.
+ *
+ * There is no guaranteed calling order of multiple registered callbacks on
+ * the same transaction.
+ */
+static inline void ext4_journal_callback_add(handle_t *handle,
+                       void (*func)(struct super_block *sb,
+                                    struct ext4_journal_cb_entry *jce,
+                                    int rc),
+                       struct ext4_journal_cb_entry *jce)
+{
+       struct ext4_sb_info *sbi =
+                       EXT4_SB(handle->h_transaction->t_journal->j_private);
+
+       /* Add the jce to transaction's private list */
+       jce->jce_func = func;
+       spin_lock(&sbi->s_md_lock);
+       list_add_tail(&jce->jce_list, &handle->h_transaction->t_private_list);
+       spin_unlock(&sbi->s_md_lock);
+}
+
+/**
+ * ext4_journal_callback_del: delete a registered callback
+ * @handle: active journal transaction handle on which callback was registered
+ * @jce: registered journal callback entry to unregister
+ * Return true if object was successfully removed
+ */
+static inline bool ext4_journal_callback_try_del(handle_t *handle,
+                                            struct ext4_journal_cb_entry *jce)
+{
+       bool deleted;
+       struct ext4_sb_info *sbi =
+                       EXT4_SB(handle->h_transaction->t_journal->j_private);
+
+       spin_lock(&sbi->s_md_lock);
+       deleted = !list_empty(&jce->jce_list);
+       list_del_init(&jce->jce_list);
+       spin_unlock(&sbi->s_md_lock);
+       return deleted;
+}
+
+int
+ext4_mark_iloc_dirty(handle_t *handle,
+                    struct inode *inode,
+                    struct ext4_iloc *iloc);
+
+/*
+ * On success, We end up with an outstanding reference count against
+ * iloc->bh.  This _must_ be cleaned up later.
+ */
+
+int ext4_reserve_inode_write(handle_t *handle, struct inode *inode,
+                       struct ext4_iloc *iloc);
+
+int ext4_mark_inode_dirty(handle_t *handle, struct inode *inode);
+
+/*
+ * Wrapper functions with which ext4 calls into JBD.
+ */
+int __ext4_journal_get_write_access(const char *where, unsigned int line,
+                                   handle_t *handle, struct buffer_head *bh);
+
+int __ext4_forget(const char *where, unsigned int line, handle_t *handle,
+                 int is_metadata, struct inode *inode,
+                 struct buffer_head *bh, ext4_fsblk_t blocknr);
+
+int __ext4_journal_get_create_access(const char *where, unsigned int line,
+                               handle_t *handle, struct buffer_head *bh);
+
+int __ext4_handle_dirty_metadata(const char *where, unsigned int line,
+                                handle_t *handle, struct inode *inode,
+                                struct buffer_head *bh);
+
+int __ext4_handle_dirty_super(const char *where, unsigned int line,
+                             handle_t *handle, struct super_block *sb);
+
+#define ext4_journal_get_write_access(handle, bh) \
+       __ext4_journal_get_write_access(__func__, __LINE__, (handle), (bh))
+#define ext4_forget(handle, is_metadata, inode, bh, block_nr) \
+       __ext4_forget(__func__, __LINE__, (handle), (is_metadata), (inode), \
+                     (bh), (block_nr))
+#define ext4_journal_get_create_access(handle, bh) \
+       __ext4_journal_get_create_access(__func__, __LINE__, (handle), (bh))
+#define ext4_handle_dirty_metadata(handle, inode, bh) \
+       __ext4_handle_dirty_metadata(__func__, __LINE__, (handle), (inode), \
+                                    (bh))
+#define ext4_handle_dirty_super(handle, sb) \
+       __ext4_handle_dirty_super(__func__, __LINE__, (handle), (sb))
+
+handle_t *__ext4_journal_start_sb(struct super_block *sb, unsigned int line,
+                                 int type, int blocks, int rsv_blocks);
+int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle);
+
+#define EXT4_NOJOURNAL_MAX_REF_COUNT ((unsigned long) 4096)
+
+/* Note:  Do not use this for NULL handles.  This is only to determine if
+ * a properly allocated handle is using a journal or not. */
+static inline int ext4_handle_valid(handle_t *handle)
+{
+       if ((unsigned long)handle < EXT4_NOJOURNAL_MAX_REF_COUNT)
+               return 0;
+       return 1;
+}
+
+static inline void ext4_handle_sync(handle_t *handle)
+{
+       if (ext4_handle_valid(handle))
+               handle->h_sync = 1;
+}
+
+static inline int ext4_handle_is_aborted(handle_t *handle)
+{
+       if (ext4_handle_valid(handle))
+               return is_handle_aborted(handle);
+       return 0;
+}
+
+static inline int ext4_handle_has_enough_credits(handle_t *handle, int needed)
+{
+       if (ext4_handle_valid(handle) && handle->h_buffer_credits < needed)
+               return 0;
+       return 1;
+}
+
+#define ext4_journal_start_sb(sb, type, nblocks)                       \
+       __ext4_journal_start_sb((sb), __LINE__, (type), (nblocks), 0)
+
+#define ext4_journal_start(inode, type, nblocks)                       \
+       __ext4_journal_start((inode), __LINE__, (type), (nblocks), 0)
+
+#define ext4_journal_start_with_reserve(inode, type, blocks, rsv_blocks) \
+       __ext4_journal_start((inode), __LINE__, (type), (blocks), (rsv_blocks))
+
+static inline handle_t *__ext4_journal_start(struct inode *inode,
+                                            unsigned int line, int type,
+                                            int blocks, int rsv_blocks)
+{
+       return __ext4_journal_start_sb(inode->i_sb, line, type, blocks,
+                                      rsv_blocks);
+}
+
+#define ext4_journal_stop(handle) \
+       __ext4_journal_stop(__func__, __LINE__, (handle))
+
+#define ext4_journal_start_reserved(handle, type) \
+       __ext4_journal_start_reserved((handle), __LINE__, (type))
+
+handle_t *__ext4_journal_start_reserved(handle_t *handle, unsigned int line,
+                                       int type);
+
+static inline void ext4_journal_free_reserved(handle_t *handle)
+{
+       if (ext4_handle_valid(handle))
+               jbd2_journal_free_reserved(handle);
+}
+
+static inline handle_t *ext4_journal_current_handle(void)
+{
+       return journal_current_handle();
+}
+
+static inline int ext4_journal_extend(handle_t *handle, int nblocks)
+{
+       if (ext4_handle_valid(handle))
+               return jbd2_journal_extend(handle, nblocks);
+       return 0;
+}
+
+static inline int ext4_journal_restart(handle_t *handle, int nblocks)
+{
+       if (ext4_handle_valid(handle))
+               return jbd2_journal_restart(handle, nblocks);
+       return 0;
+}
+
+static inline int ext4_journal_blocks_per_page(struct inode *inode)
+{
+       if (EXT4_JOURNAL(inode) != NULL)
+               return jbd2_journal_blocks_per_page(inode);
+       return 0;
+}
+
+static inline int ext4_journal_force_commit(journal_t *journal)
+{
+       if (journal)
+               return jbd2_journal_force_commit(journal);
+       return 0;
+}
+
+static inline int ext4_jbd2_file_inode(handle_t *handle, struct inode *inode)
+{
+       if (ext4_handle_valid(handle))
+               return jbd2_journal_file_inode(handle, EXT4_I(inode)->jinode);
+       return 0;
+}
+
+static inline void ext4_update_inode_fsync_trans(handle_t *handle,
+                                                struct inode *inode,
+                                                int datasync)
+{
+       struct ext4_inode_info *ei = EXT4_I(inode);
+
+       if (ext4_handle_valid(handle)) {
+               ei->i_sync_tid = handle->h_transaction->t_tid;
+               if (datasync)
+                       ei->i_datasync_tid = handle->h_transaction->t_tid;
+       }
+}
+
+/* super.c */
+int ext4_force_commit(struct super_block *sb);
+
+/*
+ * Ext4 inode journal modes
+ */
+#define EXT4_INODE_JOURNAL_DATA_MODE   0x01 /* journal data mode */
+#define EXT4_INODE_ORDERED_DATA_MODE   0x02 /* ordered data mode */
+#define EXT4_INODE_WRITEBACK_DATA_MODE 0x04 /* writeback data mode */
+
+static inline int ext4_inode_journal_mode(struct inode *inode)
+{
+       if (EXT4_JOURNAL(inode) == NULL)
+               return EXT4_INODE_WRITEBACK_DATA_MODE;  /* writeback */
+       /* We do not support data journalling with delayed allocation */
+       if (!S_ISREG(inode->i_mode) ||
+           test_opt(inode->i_sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
+               return EXT4_INODE_JOURNAL_DATA_MODE;    /* journal data */
+       if (ext4_test_inode_flag(inode, EXT4_INODE_JOURNAL_DATA) &&
+           !test_opt(inode->i_sb, DELALLOC))
+               return EXT4_INODE_JOURNAL_DATA_MODE;    /* journal data */
+       if (test_opt(inode->i_sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
+               return EXT4_INODE_ORDERED_DATA_MODE;    /* ordered */
+       if (test_opt(inode->i_sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
+               return EXT4_INODE_WRITEBACK_DATA_MODE;  /* writeback */
+       else
+               BUG();
+}
+
+static inline int ext4_should_journal_data(struct inode *inode)
+{
+       return ext4_inode_journal_mode(inode) & EXT4_INODE_JOURNAL_DATA_MODE;
+}
+
+static inline int ext4_should_order_data(struct inode *inode)
+{
+       return ext4_inode_journal_mode(inode) & EXT4_INODE_ORDERED_DATA_MODE;
+}
+
+static inline int ext4_should_writeback_data(struct inode *inode)
+{
+       return ext4_inode_journal_mode(inode) & EXT4_INODE_WRITEBACK_DATA_MODE;
+}
+
+/*
+ * This function controls whether or not we should try to go down the
+ * dioread_nolock code paths, which makes it safe to avoid taking
+ * i_mutex for direct I/O reads.  This only works for extent-based
+ * files, and it doesn't work if data journaling is enabled, since the
+ * dioread_nolock code uses b_private to pass information back to the
+ * I/O completion handler, and this conflicts with the jbd's use of
+ * b_private.
+ */
+static inline int ext4_should_dioread_nolock(struct inode *inode)
+{
+       if (!test_opt(inode->i_sb, DIOREAD_NOLOCK))
+               return 0;
+       if (!S_ISREG(inode->i_mode))
+               return 0;
+       if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)))
+               return 0;
+       if (ext4_should_journal_data(inode))
+               return 0;
+       return 1;
+}
+
+#endif /* _EXT4_JBD2_H */