Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / include / linux / buffer_head.h
diff --git a/kernel/include/linux/buffer_head.h b/kernel/include/linux/buffer_head.h
new file mode 100644 (file)
index 0000000..ea72f52
--- /dev/null
@@ -0,0 +1,440 @@
+/*
+ * include/linux/buffer_head.h
+ *
+ * Everything to do with buffer_heads.
+ */
+
+#ifndef _LINUX_BUFFER_HEAD_H
+#define _LINUX_BUFFER_HEAD_H
+
+#include <linux/types.h>
+#include <linux/fs.h>
+#include <linux/linkage.h>
+#include <linux/pagemap.h>
+#include <linux/wait.h>
+#include <linux/atomic.h>
+
+#ifdef CONFIG_BLOCK
+
+enum bh_state_bits {
+       BH_Uptodate,    /* Contains valid data */
+       BH_Dirty,       /* Is dirty */
+       BH_Lock,        /* Is locked */
+       BH_Req,         /* Has been submitted for I/O */
+       BH_Uptodate_Lock,/* Used by the first bh in a page, to serialise
+                         * IO completion of other buffers in the page
+                         */
+
+       BH_Mapped,      /* Has a disk mapping */
+       BH_New,         /* Disk mapping was newly created by get_block */
+       BH_Async_Read,  /* Is under end_buffer_async_read I/O */
+       BH_Async_Write, /* Is under end_buffer_async_write I/O */
+       BH_Delay,       /* Buffer is not yet allocated on disk */
+       BH_Boundary,    /* Block is followed by a discontiguity */
+       BH_Write_EIO,   /* I/O error on write */
+       BH_Unwritten,   /* Buffer is allocated on disk but not written */
+       BH_Quiet,       /* Buffer Error Prinks to be quiet */
+       BH_Meta,        /* Buffer contains metadata */
+       BH_Prio,        /* Buffer should be submitted with REQ_PRIO */
+       BH_Defer_Completion, /* Defer AIO completion to workqueue */
+
+       BH_PrivateStart,/* not a state bit, but the first bit available
+                        * for private allocation by other entities
+                        */
+};
+
+#define MAX_BUF_PER_PAGE (PAGE_CACHE_SIZE / 512)
+
+struct page;
+struct buffer_head;
+struct address_space;
+typedef void (bh_end_io_t)(struct buffer_head *bh, int uptodate);
+
+/*
+ * Historically, a buffer_head was used to map a single block
+ * within a page, and of course as the unit of I/O through the
+ * filesystem and block layers.  Nowadays the basic I/O unit
+ * is the bio, and buffer_heads are used for extracting block
+ * mappings (via a get_block_t call), for tracking state within
+ * a page (via a page_mapping) and for wrapping bio submission
+ * for backward compatibility reasons (e.g. submit_bh).
+ */
+struct buffer_head {
+       unsigned long b_state;          /* buffer state bitmap (see above) */
+       struct buffer_head *b_this_page;/* circular list of page's buffers */
+       struct page *b_page;            /* the page this bh is mapped to */
+
+       sector_t b_blocknr;             /* start block number */
+       size_t b_size;                  /* size of mapping */
+       char *b_data;                   /* pointer to data within the page */
+
+       struct block_device *b_bdev;
+       bh_end_io_t *b_end_io;          /* I/O completion */
+       void *b_private;                /* reserved for b_end_io */
+       struct list_head b_assoc_buffers; /* associated with another mapping */
+       struct address_space *b_assoc_map;      /* mapping this buffer is
+                                                  associated with */
+       atomic_t b_count;               /* users using this buffer_head */
+#ifdef CONFIG_PREEMPT_RT_BASE
+       spinlock_t b_uptodate_lock;
+#if defined(CONFIG_JBD) || defined(CONFIG_JBD_MODULE) || \
+       defined(CONFIG_JBD2) || defined(CONFIG_JBD2_MODULE)
+       spinlock_t b_state_lock;
+       spinlock_t b_journal_head_lock;
+#endif
+#endif
+};
+
+static inline unsigned long bh_uptodate_lock_irqsave(struct buffer_head *bh)
+{
+       unsigned long flags;
+
+#ifndef CONFIG_PREEMPT_RT_BASE
+       local_irq_save(flags);
+       bit_spin_lock(BH_Uptodate_Lock, &bh->b_state);
+#else
+       spin_lock_irqsave(&bh->b_uptodate_lock, flags);
+#endif
+       return flags;
+}
+
+static inline void
+bh_uptodate_unlock_irqrestore(struct buffer_head *bh, unsigned long flags)
+{
+#ifndef CONFIG_PREEMPT_RT_BASE
+       bit_spin_unlock(BH_Uptodate_Lock, &bh->b_state);
+       local_irq_restore(flags);
+#else
+       spin_unlock_irqrestore(&bh->b_uptodate_lock, flags);
+#endif
+}
+
+static inline void buffer_head_init_locks(struct buffer_head *bh)
+{
+#ifdef CONFIG_PREEMPT_RT_BASE
+       spin_lock_init(&bh->b_uptodate_lock);
+#if defined(CONFIG_JBD) || defined(CONFIG_JBD_MODULE) || \
+       defined(CONFIG_JBD2) || defined(CONFIG_JBD2_MODULE)
+       spin_lock_init(&bh->b_state_lock);
+       spin_lock_init(&bh->b_journal_head_lock);
+#endif
+#endif
+}
+
+/*
+ * macro tricks to expand the set_buffer_foo(), clear_buffer_foo()
+ * and buffer_foo() functions.
+ */
+#define BUFFER_FNS(bit, name)                                          \
+static inline void set_buffer_##name(struct buffer_head *bh)           \
+{                                                                      \
+       set_bit(BH_##bit, &(bh)->b_state);                              \
+}                                                                      \
+static inline void clear_buffer_##name(struct buffer_head *bh)         \
+{                                                                      \
+       clear_bit(BH_##bit, &(bh)->b_state);                            \
+}                                                                      \
+static inline int buffer_##name(const struct buffer_head *bh)          \
+{                                                                      \
+       return test_bit(BH_##bit, &(bh)->b_state);                      \
+}
+
+/*
+ * test_set_buffer_foo() and test_clear_buffer_foo()
+ */
+#define TAS_BUFFER_FNS(bit, name)                                      \
+static inline int test_set_buffer_##name(struct buffer_head *bh)       \
+{                                                                      \
+       return test_and_set_bit(BH_##bit, &(bh)->b_state);              \
+}                                                                      \
+static inline int test_clear_buffer_##name(struct buffer_head *bh)     \
+{                                                                      \
+       return test_and_clear_bit(BH_##bit, &(bh)->b_state);            \
+}                                                                      \
+
+/*
+ * Emit the buffer bitops functions.   Note that there are also functions
+ * of the form "mark_buffer_foo()".  These are higher-level functions which
+ * do something in addition to setting a b_state bit.
+ */
+BUFFER_FNS(Uptodate, uptodate)
+BUFFER_FNS(Dirty, dirty)
+TAS_BUFFER_FNS(Dirty, dirty)
+BUFFER_FNS(Lock, locked)
+BUFFER_FNS(Req, req)
+TAS_BUFFER_FNS(Req, req)
+BUFFER_FNS(Mapped, mapped)
+BUFFER_FNS(New, new)
+BUFFER_FNS(Async_Read, async_read)
+BUFFER_FNS(Async_Write, async_write)
+BUFFER_FNS(Delay, delay)
+BUFFER_FNS(Boundary, boundary)
+BUFFER_FNS(Write_EIO, write_io_error)
+BUFFER_FNS(Unwritten, unwritten)
+BUFFER_FNS(Meta, meta)
+BUFFER_FNS(Prio, prio)
+BUFFER_FNS(Defer_Completion, defer_completion)
+
+#define bh_offset(bh)          ((unsigned long)(bh)->b_data & ~PAGE_MASK)
+
+/* If we *know* page->private refers to buffer_heads */
+#define page_buffers(page)                                     \
+       ({                                                      \
+               BUG_ON(!PagePrivate(page));                     \
+               ((struct buffer_head *)page_private(page));     \
+       })
+#define page_has_buffers(page) PagePrivate(page)
+
+void buffer_check_dirty_writeback(struct page *page,
+                                    bool *dirty, bool *writeback);
+
+/*
+ * Declarations
+ */
+
+void mark_buffer_dirty(struct buffer_head *bh);
+void init_buffer(struct buffer_head *, bh_end_io_t *, void *);
+void touch_buffer(struct buffer_head *bh);
+void set_bh_page(struct buffer_head *bh,
+               struct page *page, unsigned long offset);
+int try_to_free_buffers(struct page *);
+struct buffer_head *alloc_page_buffers(struct page *page, unsigned long size,
+               int retry);
+void create_empty_buffers(struct page *, unsigned long,
+                       unsigned long b_state);
+void end_buffer_read_sync(struct buffer_head *bh, int uptodate);
+void end_buffer_write_sync(struct buffer_head *bh, int uptodate);
+void end_buffer_async_write(struct buffer_head *bh, int uptodate);
+
+/* Things to do with buffers at mapping->private_list */
+void mark_buffer_dirty_inode(struct buffer_head *bh, struct inode *inode);
+int inode_has_buffers(struct inode *);
+void invalidate_inode_buffers(struct inode *);
+int remove_inode_buffers(struct inode *inode);
+int sync_mapping_buffers(struct address_space *mapping);
+void unmap_underlying_metadata(struct block_device *bdev, sector_t block);
+
+void mark_buffer_async_write(struct buffer_head *bh);
+void __wait_on_buffer(struct buffer_head *);
+wait_queue_head_t *bh_waitq_head(struct buffer_head *bh);
+struct buffer_head *__find_get_block(struct block_device *bdev, sector_t block,
+                       unsigned size);
+struct buffer_head *__getblk_gfp(struct block_device *bdev, sector_t block,
+                                 unsigned size, gfp_t gfp);
+void __brelse(struct buffer_head *);
+void __bforget(struct buffer_head *);
+void __breadahead(struct block_device *, sector_t block, unsigned int size);
+struct buffer_head *__bread_gfp(struct block_device *,
+                               sector_t block, unsigned size, gfp_t gfp);
+void invalidate_bh_lrus(void);
+struct buffer_head *alloc_buffer_head(gfp_t gfp_flags);
+void free_buffer_head(struct buffer_head * bh);
+void unlock_buffer(struct buffer_head *bh);
+void __lock_buffer(struct buffer_head *bh);
+void ll_rw_block(int, int, struct buffer_head * bh[]);
+int sync_dirty_buffer(struct buffer_head *bh);
+int __sync_dirty_buffer(struct buffer_head *bh, int rw);
+void write_dirty_buffer(struct buffer_head *bh, int rw);
+int _submit_bh(int rw, struct buffer_head *bh, unsigned long bio_flags);
+int submit_bh(int, struct buffer_head *);
+void write_boundary_block(struct block_device *bdev,
+                       sector_t bblock, unsigned blocksize);
+int bh_uptodate_or_lock(struct buffer_head *bh);
+int bh_submit_read(struct buffer_head *bh);
+
+extern int buffer_heads_over_limit;
+
+/*
+ * Generic address_space_operations implementations for buffer_head-backed
+ * address_spaces.
+ */
+void block_invalidatepage(struct page *page, unsigned int offset,
+                         unsigned int length);
+int block_write_full_page(struct page *page, get_block_t *get_block,
+                               struct writeback_control *wbc);
+int block_read_full_page(struct page*, get_block_t*);
+int block_is_partially_uptodate(struct page *page, unsigned long from,
+                               unsigned long count);
+int block_write_begin(struct address_space *mapping, loff_t pos, unsigned len,
+               unsigned flags, struct page **pagep, get_block_t *get_block);
+int __block_write_begin(struct page *page, loff_t pos, unsigned len,
+               get_block_t *get_block);
+int block_write_end(struct file *, struct address_space *,
+                               loff_t, unsigned, unsigned,
+                               struct page *, void *);
+int generic_write_end(struct file *, struct address_space *,
+                               loff_t, unsigned, unsigned,
+                               struct page *, void *);
+void page_zero_new_buffers(struct page *page, unsigned from, unsigned to);
+int cont_write_begin(struct file *, struct address_space *, loff_t,
+                       unsigned, unsigned, struct page **, void **,
+                       get_block_t *, loff_t *);
+int generic_cont_expand_simple(struct inode *inode, loff_t size);
+int block_commit_write(struct page *page, unsigned from, unsigned to);
+int __block_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf,
+                               get_block_t get_block);
+int block_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf,
+                               get_block_t get_block);
+/* Convert errno to return value from ->page_mkwrite() call */
+static inline int block_page_mkwrite_return(int err)
+{
+       if (err == 0)
+               return VM_FAULT_LOCKED;
+       if (err == -EFAULT)
+               return VM_FAULT_NOPAGE;
+       if (err == -ENOMEM)
+               return VM_FAULT_OOM;
+       if (err == -EAGAIN)
+               return VM_FAULT_RETRY;
+       /* -ENOSPC, -EDQUOT, -EIO ... */
+       return VM_FAULT_SIGBUS;
+}
+sector_t generic_block_bmap(struct address_space *, sector_t, get_block_t *);
+int block_truncate_page(struct address_space *, loff_t, get_block_t *);
+int nobh_write_begin(struct address_space *, loff_t, unsigned, unsigned,
+                               struct page **, void **, get_block_t*);
+int nobh_write_end(struct file *, struct address_space *,
+                               loff_t, unsigned, unsigned,
+                               struct page *, void *);
+int nobh_truncate_page(struct address_space *, loff_t, get_block_t *);
+int nobh_writepage(struct page *page, get_block_t *get_block,
+                        struct writeback_control *wbc);
+
+void buffer_init(void);
+
+/*
+ * inline definitions
+ */
+
+static inline void attach_page_buffers(struct page *page,
+               struct buffer_head *head)
+{
+       page_cache_get(page);
+       SetPagePrivate(page);
+       set_page_private(page, (unsigned long)head);
+}
+
+static inline void get_bh(struct buffer_head *bh)
+{
+        atomic_inc(&bh->b_count);
+}
+
+static inline void put_bh(struct buffer_head *bh)
+{
+        smp_mb__before_atomic();
+        atomic_dec(&bh->b_count);
+}
+
+static inline void brelse(struct buffer_head *bh)
+{
+       if (bh)
+               __brelse(bh);
+}
+
+static inline void bforget(struct buffer_head *bh)
+{
+       if (bh)
+               __bforget(bh);
+}
+
+static inline struct buffer_head *
+sb_bread(struct super_block *sb, sector_t block)
+{
+       return __bread_gfp(sb->s_bdev, block, sb->s_blocksize, __GFP_MOVABLE);
+}
+
+static inline struct buffer_head *
+sb_bread_unmovable(struct super_block *sb, sector_t block)
+{
+       return __bread_gfp(sb->s_bdev, block, sb->s_blocksize, 0);
+}
+
+static inline void
+sb_breadahead(struct super_block *sb, sector_t block)
+{
+       __breadahead(sb->s_bdev, block, sb->s_blocksize);
+}
+
+static inline struct buffer_head *
+sb_getblk(struct super_block *sb, sector_t block)
+{
+       return __getblk_gfp(sb->s_bdev, block, sb->s_blocksize, __GFP_MOVABLE);
+}
+
+static inline struct buffer_head *
+sb_find_get_block(struct super_block *sb, sector_t block)
+{
+       return __find_get_block(sb->s_bdev, block, sb->s_blocksize);
+}
+
+static inline void
+map_bh(struct buffer_head *bh, struct super_block *sb, sector_t block)
+{
+       set_buffer_mapped(bh);
+       bh->b_bdev = sb->s_bdev;
+       bh->b_blocknr = block;
+       bh->b_size = sb->s_blocksize;
+}
+
+static inline void wait_on_buffer(struct buffer_head *bh)
+{
+       might_sleep();
+       if (buffer_locked(bh))
+               __wait_on_buffer(bh);
+}
+
+static inline int trylock_buffer(struct buffer_head *bh)
+{
+       return likely(!test_and_set_bit_lock(BH_Lock, &bh->b_state));
+}
+
+static inline void lock_buffer(struct buffer_head *bh)
+{
+       might_sleep();
+       if (!trylock_buffer(bh))
+               __lock_buffer(bh);
+}
+
+static inline struct buffer_head *getblk_unmovable(struct block_device *bdev,
+                                                  sector_t block,
+                                                  unsigned size)
+{
+       return __getblk_gfp(bdev, block, size, 0);
+}
+
+static inline struct buffer_head *__getblk(struct block_device *bdev,
+                                          sector_t block,
+                                          unsigned size)
+{
+       return __getblk_gfp(bdev, block, size, __GFP_MOVABLE);
+}
+
+/**
+ *  __bread() - reads a specified block and returns the bh
+ *  @bdev: the block_device to read from
+ *  @block: number of block
+ *  @size: size (in bytes) to read
+ *
+ *  Reads a specified block, and returns buffer head that contains it.
+ *  The page cache is allocated from movable area so that it can be migrated.
+ *  It returns NULL if the block was unreadable.
+ */
+static inline struct buffer_head *
+__bread(struct block_device *bdev, sector_t block, unsigned size)
+{
+       return __bread_gfp(bdev, block, size, __GFP_MOVABLE);
+}
+
+extern int __set_page_dirty_buffers(struct page *page);
+
+#else /* CONFIG_BLOCK */
+
+static inline void buffer_init(void) {}
+static inline int try_to_free_buffers(struct page *page) { return 1; }
+static inline int inode_has_buffers(struct inode *inode) { return 0; }
+static inline void invalidate_inode_buffers(struct inode *inode) {}
+static inline int remove_inode_buffers(struct inode *inode) { return 1; }
+static inline int sync_mapping_buffers(struct address_space *mapping) { return 0; }
+
+#endif /* CONFIG_BLOCK */
+#endif /* _LINUX_BUFFER_HEAD_H */