Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / fs / btrfs / volumes.h
diff --git a/kernel/fs/btrfs/volumes.h b/kernel/fs/btrfs/volumes.h
new file mode 100644 (file)
index 0000000..ebc3133
--- /dev/null
@@ -0,0 +1,541 @@
+/*
+ * Copyright (C) 2007 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef __BTRFS_VOLUMES_
+#define __BTRFS_VOLUMES_
+
+#include <linux/bio.h>
+#include <linux/sort.h>
+#include <linux/btrfs.h>
+#include "async-thread.h"
+
+extern struct mutex uuid_mutex;
+
+#define BTRFS_STRIPE_LEN       (64 * 1024)
+
+struct buffer_head;
+struct btrfs_pending_bios {
+       struct bio *head;
+       struct bio *tail;
+};
+
+/*
+ * Use sequence counter to get consistent device stat data on
+ * 32-bit processors.
+ */
+#if BITS_PER_LONG==32 && defined(CONFIG_SMP)
+#include <linux/seqlock.h>
+#define __BTRFS_NEED_DEVICE_DATA_ORDERED
+#define btrfs_device_data_ordered_init(device) \
+       seqcount_init(&device->data_seqcount)
+#else
+#define btrfs_device_data_ordered_init(device) do { } while (0)
+#endif
+
+struct btrfs_device {
+       struct list_head dev_list;
+       struct list_head dev_alloc_list;
+       struct btrfs_fs_devices *fs_devices;
+
+       struct btrfs_root *dev_root;
+
+       struct rcu_string *name;
+
+       u64 generation;
+
+       spinlock_t io_lock ____cacheline_aligned;
+       int running_pending;
+       /* regular prio bios */
+       struct btrfs_pending_bios pending_bios;
+       /* WRITE_SYNC bios */
+       struct btrfs_pending_bios pending_sync_bios;
+
+       struct block_device *bdev;
+
+       /* the mode sent to blkdev_get */
+       fmode_t mode;
+
+       int writeable;
+       int in_fs_metadata;
+       int missing;
+       int can_discard;
+       int is_tgtdev_for_dev_replace;
+
+#ifdef __BTRFS_NEED_DEVICE_DATA_ORDERED
+       seqcount_t data_seqcount;
+#endif
+
+       /* the internal btrfs device id */
+       u64 devid;
+
+       /* size of the device in memory */
+       u64 total_bytes;
+
+       /* size of the device on disk */
+       u64 disk_total_bytes;
+
+       /* bytes used */
+       u64 bytes_used;
+
+       /* optimal io alignment for this device */
+       u32 io_align;
+
+       /* optimal io width for this device */
+       u32 io_width;
+       /* type and info about this device */
+       u64 type;
+
+       /* minimal io size for this device */
+       u32 sector_size;
+
+       /* physical drive uuid (or lvm uuid) */
+       u8 uuid[BTRFS_UUID_SIZE];
+
+       /*
+        * size of the device on the current transaction
+        *
+        * This variant is update when committing the transaction,
+        * and protected by device_list_mutex
+        */
+       u64 commit_total_bytes;
+
+       /* bytes used on the current transaction */
+       u64 commit_bytes_used;
+       /*
+        * used to manage the device which is resized
+        *
+        * It is protected by chunk_lock.
+        */
+       struct list_head resized_list;
+
+       /* for sending down flush barriers */
+       int nobarriers;
+       struct bio *flush_bio;
+       struct completion flush_wait;
+
+       /* per-device scrub information */
+       struct scrub_ctx *scrub_device;
+
+       struct btrfs_work work;
+       struct rcu_head rcu;
+       struct work_struct rcu_work;
+
+       /* readahead state */
+       spinlock_t reada_lock;
+       atomic_t reada_in_flight;
+       u64 reada_next;
+       struct reada_zone *reada_curr_zone;
+       struct radix_tree_root reada_zones;
+       struct radix_tree_root reada_extents;
+
+       /* disk I/O failure stats. For detailed description refer to
+        * enum btrfs_dev_stat_values in ioctl.h */
+       int dev_stats_valid;
+
+       /* Counter to record the change of device stats */
+       atomic_t dev_stats_ccnt;
+       atomic_t dev_stat_values[BTRFS_DEV_STAT_VALUES_MAX];
+};
+
+/*
+ * If we read those variants at the context of their own lock, we needn't
+ * use the following helpers, reading them directly is safe.
+ */
+#if BITS_PER_LONG==32 && defined(CONFIG_SMP)
+#define BTRFS_DEVICE_GETSET_FUNCS(name)                                        \
+static inline u64                                                      \
+btrfs_device_get_##name(const struct btrfs_device *dev)                        \
+{                                                                      \
+       u64 size;                                                       \
+       unsigned int seq;                                               \
+                                                                       \
+       do {                                                            \
+               seq = read_seqcount_begin(&dev->data_seqcount);         \
+               size = dev->name;                                       \
+       } while (read_seqcount_retry(&dev->data_seqcount, seq));        \
+       return size;                                                    \
+}                                                                      \
+                                                                       \
+static inline void                                                     \
+btrfs_device_set_##name(struct btrfs_device *dev, u64 size)            \
+{                                                                      \
+       preempt_disable();                                              \
+       write_seqcount_begin(&dev->data_seqcount);                      \
+       dev->name = size;                                               \
+       write_seqcount_end(&dev->data_seqcount);                        \
+       preempt_enable();                                               \
+}
+#elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPT)
+#define BTRFS_DEVICE_GETSET_FUNCS(name)                                        \
+static inline u64                                                      \
+btrfs_device_get_##name(const struct btrfs_device *dev)                        \
+{                                                                      \
+       u64 size;                                                       \
+                                                                       \
+       preempt_disable();                                              \
+       size = dev->name;                                               \
+       preempt_enable();                                               \
+       return size;                                                    \
+}                                                                      \
+                                                                       \
+static inline void                                                     \
+btrfs_device_set_##name(struct btrfs_device *dev, u64 size)            \
+{                                                                      \
+       preempt_disable();                                              \
+       dev->name = size;                                               \
+       preempt_enable();                                               \
+}
+#else
+#define BTRFS_DEVICE_GETSET_FUNCS(name)                                        \
+static inline u64                                                      \
+btrfs_device_get_##name(const struct btrfs_device *dev)                        \
+{                                                                      \
+       return dev->name;                                               \
+}                                                                      \
+                                                                       \
+static inline void                                                     \
+btrfs_device_set_##name(struct btrfs_device *dev, u64 size)            \
+{                                                                      \
+       dev->name = size;                                               \
+}
+#endif
+
+BTRFS_DEVICE_GETSET_FUNCS(total_bytes);
+BTRFS_DEVICE_GETSET_FUNCS(disk_total_bytes);
+BTRFS_DEVICE_GETSET_FUNCS(bytes_used);
+
+struct btrfs_fs_devices {
+       u8 fsid[BTRFS_FSID_SIZE]; /* FS specific uuid */
+
+       u64 num_devices;
+       u64 open_devices;
+       u64 rw_devices;
+       u64 missing_devices;
+       u64 total_rw_bytes;
+       u64 total_devices;
+       struct block_device *latest_bdev;
+
+       /* all of the devices in the FS, protected by a mutex
+        * so we can safely walk it to write out the supers without
+        * worrying about add/remove by the multi-device code.
+        * Scrubbing super can kick off supers writing by holding
+        * this mutex lock.
+        */
+       struct mutex device_list_mutex;
+       struct list_head devices;
+
+       struct list_head resized_devices;
+       /* devices not currently being allocated */
+       struct list_head alloc_list;
+       struct list_head list;
+
+       struct btrfs_fs_devices *seed;
+       int seeding;
+
+       int opened;
+
+       /* set when we find or add a device that doesn't have the
+        * nonrot flag set
+        */
+       int rotating;
+};
+
+#define BTRFS_BIO_INLINE_CSUM_SIZE     64
+
+/*
+ * we need the mirror number and stripe index to be passed around
+ * the call chain while we are processing end_io (especially errors).
+ * Really, what we need is a btrfs_bio structure that has this info
+ * and is properly sized with its stripe array, but we're not there
+ * quite yet.  We have our own btrfs bioset, and all of the bios
+ * we allocate are actually btrfs_io_bios.  We'll cram as much of
+ * struct btrfs_bio as we can into this over time.
+ */
+typedef void (btrfs_io_bio_end_io_t) (struct btrfs_io_bio *bio, int err);
+struct btrfs_io_bio {
+       unsigned int mirror_num;
+       unsigned int stripe_index;
+       u64 logical;
+       u8 *csum;
+       u8 csum_inline[BTRFS_BIO_INLINE_CSUM_SIZE];
+       u8 *csum_allocated;
+       btrfs_io_bio_end_io_t *end_io;
+       struct bio bio;
+};
+
+static inline struct btrfs_io_bio *btrfs_io_bio(struct bio *bio)
+{
+       return container_of(bio, struct btrfs_io_bio, bio);
+}
+
+struct btrfs_bio_stripe {
+       struct btrfs_device *dev;
+       u64 physical;
+       u64 length; /* only used for discard mappings */
+};
+
+struct btrfs_bio;
+typedef void (btrfs_bio_end_io_t) (struct btrfs_bio *bio, int err);
+
+#define BTRFS_BIO_ORIG_BIO_SUBMITTED   (1 << 0)
+
+struct btrfs_bio {
+       atomic_t refs;
+       atomic_t stripes_pending;
+       struct btrfs_fs_info *fs_info;
+       u64 map_type; /* get from map_lookup->type */
+       bio_end_io_t *end_io;
+       struct bio *orig_bio;
+       unsigned long flags;
+       void *private;
+       atomic_t error;
+       int max_errors;
+       int num_stripes;
+       int mirror_num;
+       int num_tgtdevs;
+       int *tgtdev_map;
+       /*
+        * logical block numbers for the start of each stripe
+        * The last one or two are p/q.  These are sorted,
+        * so raid_map[0] is the start of our full stripe
+        */
+       u64 *raid_map;
+       struct btrfs_bio_stripe stripes[];
+};
+
+struct btrfs_device_info {
+       struct btrfs_device *dev;
+       u64 dev_offset;
+       u64 max_avail;
+       u64 total_avail;
+};
+
+struct btrfs_raid_attr {
+       int sub_stripes;        /* sub_stripes info for map */
+       int dev_stripes;        /* stripes per dev */
+       int devs_max;           /* max devs to use */
+       int devs_min;           /* min devs needed */
+       int devs_increment;     /* ndevs has to be a multiple of this */
+       int ncopies;            /* how many copies to data has */
+};
+
+struct map_lookup {
+       u64 type;
+       int io_align;
+       int io_width;
+       int stripe_len;
+       int sector_size;
+       int num_stripes;
+       int sub_stripes;
+       struct btrfs_bio_stripe stripes[];
+};
+
+#define map_lookup_size(n) (sizeof(struct map_lookup) + \
+                           (sizeof(struct btrfs_bio_stripe) * (n)))
+
+/*
+ * Restriper's general type filter
+ */
+#define BTRFS_BALANCE_DATA             (1ULL << 0)
+#define BTRFS_BALANCE_SYSTEM           (1ULL << 1)
+#define BTRFS_BALANCE_METADATA         (1ULL << 2)
+
+#define BTRFS_BALANCE_TYPE_MASK                (BTRFS_BALANCE_DATA |       \
+                                        BTRFS_BALANCE_SYSTEM |     \
+                                        BTRFS_BALANCE_METADATA)
+
+#define BTRFS_BALANCE_FORCE            (1ULL << 3)
+#define BTRFS_BALANCE_RESUME           (1ULL << 4)
+
+/*
+ * Balance filters
+ */
+#define BTRFS_BALANCE_ARGS_PROFILES    (1ULL << 0)
+#define BTRFS_BALANCE_ARGS_USAGE       (1ULL << 1)
+#define BTRFS_BALANCE_ARGS_DEVID       (1ULL << 2)
+#define BTRFS_BALANCE_ARGS_DRANGE      (1ULL << 3)
+#define BTRFS_BALANCE_ARGS_VRANGE      (1ULL << 4)
+#define BTRFS_BALANCE_ARGS_LIMIT       (1ULL << 5)
+
+/*
+ * Profile changing flags.  When SOFT is set we won't relocate chunk if
+ * it already has the target profile (even though it may be
+ * half-filled).
+ */
+#define BTRFS_BALANCE_ARGS_CONVERT     (1ULL << 8)
+#define BTRFS_BALANCE_ARGS_SOFT                (1ULL << 9)
+
+struct btrfs_balance_args;
+struct btrfs_balance_progress;
+struct btrfs_balance_control {
+       struct btrfs_fs_info *fs_info;
+
+       struct btrfs_balance_args data;
+       struct btrfs_balance_args meta;
+       struct btrfs_balance_args sys;
+
+       u64 flags;
+
+       struct btrfs_balance_progress stat;
+};
+
+int btrfs_account_dev_extents_size(struct btrfs_device *device, u64 start,
+                                  u64 end, u64 *length);
+void btrfs_get_bbio(struct btrfs_bio *bbio);
+void btrfs_put_bbio(struct btrfs_bio *bbio);
+int btrfs_map_block(struct btrfs_fs_info *fs_info, int rw,
+                   u64 logical, u64 *length,
+                   struct btrfs_bio **bbio_ret, int mirror_num);
+int btrfs_map_sblock(struct btrfs_fs_info *fs_info, int rw,
+                    u64 logical, u64 *length,
+                    struct btrfs_bio **bbio_ret, int mirror_num,
+                    int need_raid_map);
+int btrfs_rmap_block(struct btrfs_mapping_tree *map_tree,
+                    u64 chunk_start, u64 physical, u64 devid,
+                    u64 **logical, int *naddrs, int *stripe_len);
+int btrfs_read_sys_array(struct btrfs_root *root);
+int btrfs_read_chunk_tree(struct btrfs_root *root);
+int btrfs_alloc_chunk(struct btrfs_trans_handle *trans,
+                     struct btrfs_root *extent_root, u64 type);
+void btrfs_mapping_init(struct btrfs_mapping_tree *tree);
+void btrfs_mapping_tree_free(struct btrfs_mapping_tree *tree);
+int btrfs_map_bio(struct btrfs_root *root, int rw, struct bio *bio,
+                 int mirror_num, int async_submit);
+int btrfs_open_devices(struct btrfs_fs_devices *fs_devices,
+                      fmode_t flags, void *holder);
+int btrfs_scan_one_device(const char *path, fmode_t flags, void *holder,
+                         struct btrfs_fs_devices **fs_devices_ret);
+int btrfs_close_devices(struct btrfs_fs_devices *fs_devices);
+void btrfs_close_extra_devices(struct btrfs_fs_devices *fs_devices, int step);
+int btrfs_find_device_missing_or_by_path(struct btrfs_root *root,
+                                        char *device_path,
+                                        struct btrfs_device **device);
+struct btrfs_device *btrfs_alloc_device(struct btrfs_fs_info *fs_info,
+                                       const u64 *devid,
+                                       const u8 *uuid);
+int btrfs_rm_device(struct btrfs_root *root, char *device_path);
+void btrfs_cleanup_fs_uuids(void);
+int btrfs_num_copies(struct btrfs_fs_info *fs_info, u64 logical, u64 len);
+int btrfs_grow_device(struct btrfs_trans_handle *trans,
+                     struct btrfs_device *device, u64 new_size);
+struct btrfs_device *btrfs_find_device(struct btrfs_fs_info *fs_info, u64 devid,
+                                      u8 *uuid, u8 *fsid);
+int btrfs_shrink_device(struct btrfs_device *device, u64 new_size);
+int btrfs_init_new_device(struct btrfs_root *root, char *path);
+int btrfs_init_dev_replace_tgtdev(struct btrfs_root *root, char *device_path,
+                                 struct btrfs_device *srcdev,
+                                 struct btrfs_device **device_out);
+int btrfs_balance(struct btrfs_balance_control *bctl,
+                 struct btrfs_ioctl_balance_args *bargs);
+int btrfs_resume_balance_async(struct btrfs_fs_info *fs_info);
+int btrfs_recover_balance(struct btrfs_fs_info *fs_info);
+int btrfs_pause_balance(struct btrfs_fs_info *fs_info);
+int btrfs_cancel_balance(struct btrfs_fs_info *fs_info);
+int btrfs_create_uuid_tree(struct btrfs_fs_info *fs_info);
+int btrfs_check_uuid_tree(struct btrfs_fs_info *fs_info);
+int btrfs_chunk_readonly(struct btrfs_root *root, u64 chunk_offset);
+int find_free_dev_extent(struct btrfs_trans_handle *trans,
+                        struct btrfs_device *device, u64 num_bytes,
+                        u64 *start, u64 *max_avail);
+void btrfs_dev_stat_inc_and_print(struct btrfs_device *dev, int index);
+int btrfs_get_dev_stats(struct btrfs_root *root,
+                       struct btrfs_ioctl_get_dev_stats *stats);
+void btrfs_init_devices_late(struct btrfs_fs_info *fs_info);
+int btrfs_init_dev_stats(struct btrfs_fs_info *fs_info);
+int btrfs_run_dev_stats(struct btrfs_trans_handle *trans,
+                       struct btrfs_fs_info *fs_info);
+void btrfs_rm_dev_replace_remove_srcdev(struct btrfs_fs_info *fs_info,
+                                       struct btrfs_device *srcdev);
+void btrfs_rm_dev_replace_free_srcdev(struct btrfs_fs_info *fs_info,
+                                     struct btrfs_device *srcdev);
+void btrfs_destroy_dev_replace_tgtdev(struct btrfs_fs_info *fs_info,
+                                     struct btrfs_device *tgtdev);
+void btrfs_init_dev_replace_tgtdev_for_resume(struct btrfs_fs_info *fs_info,
+                                             struct btrfs_device *tgtdev);
+int btrfs_scratch_superblock(struct btrfs_device *device);
+int btrfs_is_parity_mirror(struct btrfs_mapping_tree *map_tree,
+                          u64 logical, u64 len, int mirror_num);
+unsigned long btrfs_full_stripe_len(struct btrfs_root *root,
+                                   struct btrfs_mapping_tree *map_tree,
+                                   u64 logical);
+int btrfs_finish_chunk_alloc(struct btrfs_trans_handle *trans,
+                               struct btrfs_root *extent_root,
+                               u64 chunk_offset, u64 chunk_size);
+int btrfs_remove_chunk(struct btrfs_trans_handle *trans,
+                      struct btrfs_root *root, u64 chunk_offset);
+
+static inline int btrfs_dev_stats_dirty(struct btrfs_device *dev)
+{
+       return atomic_read(&dev->dev_stats_ccnt);
+}
+
+static inline void btrfs_dev_stat_inc(struct btrfs_device *dev,
+                                     int index)
+{
+       atomic_inc(dev->dev_stat_values + index);
+       smp_mb__before_atomic();
+       atomic_inc(&dev->dev_stats_ccnt);
+}
+
+static inline int btrfs_dev_stat_read(struct btrfs_device *dev,
+                                     int index)
+{
+       return atomic_read(dev->dev_stat_values + index);
+}
+
+static inline int btrfs_dev_stat_read_and_reset(struct btrfs_device *dev,
+                                               int index)
+{
+       int ret;
+
+       ret = atomic_xchg(dev->dev_stat_values + index, 0);
+       smp_mb__before_atomic();
+       atomic_inc(&dev->dev_stats_ccnt);
+       return ret;
+}
+
+static inline void btrfs_dev_stat_set(struct btrfs_device *dev,
+                                     int index, unsigned long val)
+{
+       atomic_set(dev->dev_stat_values + index, val);
+       smp_mb__before_atomic();
+       atomic_inc(&dev->dev_stats_ccnt);
+}
+
+static inline void btrfs_dev_stat_reset(struct btrfs_device *dev,
+                                       int index)
+{
+       btrfs_dev_stat_set(dev, index, 0);
+}
+
+void btrfs_update_commit_device_size(struct btrfs_fs_info *fs_info);
+void btrfs_update_commit_device_bytes_used(struct btrfs_root *root,
+                                       struct btrfs_transaction *transaction);
+
+static inline void lock_chunks(struct btrfs_root *root)
+{
+       mutex_lock(&root->fs_info->chunk_mutex);
+}
+
+static inline void unlock_chunks(struct btrfs_root *root)
+{
+       mutex_unlock(&root->fs_info->chunk_mutex);
+}
+
+
+#endif