X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=libbcachefs%2Ffs.h;h=c3af7225ff693ec9c5af06502e22f3fbc8354fd5;hb=3a0cc86e767b95366b9cbdba7a1453454f1b5a41;hp=3df85ffb450ccbdb87cbc809d6c5e0efdb2279f1;hpb=db931a4571817d7d61be6bce306f1d42f7cd3398;p=bcachefs-tools-debian diff --git a/libbcachefs/fs.h b/libbcachefs/fs.h index 3df85ff..c3af722 100644 --- a/libbcachefs/fs.h +++ b/libbcachefs/fs.h @@ -6,57 +6,68 @@ #include "opts.h" #include "str_hash.h" #include "quota_types.h" +#include "two_state_shared_lock.h" #include #include -/* - * Two-state lock - can be taken for add or block - both states are shared, - * like read side of rwsem, but conflict with other state: - */ -struct pagecache_lock { - atomic_long_t v; - wait_queue_head_t wait; -}; - -static inline void pagecache_lock_init(struct pagecache_lock *lock) -{ - atomic_long_set(&lock->v, 0); - init_waitqueue_head(&lock->wait); -} - -void bch2_pagecache_add_put(struct pagecache_lock *); -bool bch2_pagecache_add_tryget(struct pagecache_lock *); -void bch2_pagecache_add_get(struct pagecache_lock *); -void bch2_pagecache_block_put(struct pagecache_lock *); -void bch2_pagecache_block_get(struct pagecache_lock *); - struct bch_inode_info { struct inode v; + struct list_head ei_vfs_inode_list; unsigned long ei_flags; struct mutex ei_update_lock; - u64 ei_journal_seq; u64 ei_quota_reserved; unsigned long ei_last_dirtied; - - struct pagecache_lock ei_pagecache_lock; + two_state_lock_t ei_pagecache_lock; struct mutex ei_quota_lock; struct bch_qid ei_qid; - struct bch_hash_info ei_str_hash; + u32 ei_subvol; + + /* + * When we've been doing nocow writes we'll need to issue flushes to the + * underlying block devices + * + * XXX: a device may have had a flush issued by some other codepath. It + * would be better to keep for each device a sequence number that's + * incremented when we isusue a cache flush, and track here the sequence + * number that needs flushing. + */ + struct bch_devs_mask ei_devs_need_flush; /* copy of inode in btree: */ struct bch_inode_unpacked ei_inode; }; +#define bch2_pagecache_add_put(i) bch2_two_state_unlock(&i->ei_pagecache_lock, 0) +#define bch2_pagecache_add_tryget(i) bch2_two_state_trylock(&i->ei_pagecache_lock, 0) +#define bch2_pagecache_add_get(i) bch2_two_state_lock(&i->ei_pagecache_lock, 0) + +#define bch2_pagecache_block_put(i) bch2_two_state_unlock(&i->ei_pagecache_lock, 1) +#define bch2_pagecache_block_get(i) bch2_two_state_lock(&i->ei_pagecache_lock, 1) + +static inline subvol_inum inode_inum(struct bch_inode_info *inode) +{ + return (subvol_inum) { + .subvol = inode->ei_subvol, + .inum = inode->ei_inode.bi_inum, + }; +} + /* * Set if we've gotten a btree error for this inode, and thus the vfs inode and * btree inode may be inconsistent: */ #define EI_INODE_ERROR 0 +/* + * Set in the inode is in a snapshot subvolume - we don't do quota accounting in + * those: + */ +#define EI_INODE_SNAPSHOT 1 + #define to_bch_ei(_inode) \ container_of_or_null(_inode, struct bch_inode_info, v) @@ -66,9 +77,8 @@ static inline int ptrcmp(void *l, void *r) } enum bch_inode_lock_op { - INODE_LOCK = (1U << 0), - INODE_PAGECACHE_BLOCK = (1U << 1), - INODE_UPDATE_LOCK = (1U << 2), + INODE_PAGECACHE_BLOCK = (1U << 0), + INODE_UPDATE_LOCK = (1U << 1), }; #define bch2_lock_inodes(_locks, ...) \ @@ -80,10 +90,8 @@ do { \ \ for (i = 1; i < ARRAY_SIZE(a); i++) \ if (a[i] != a[i - 1]) { \ - if ((_locks) & INODE_LOCK) \ - down_write_nested(&a[i]->v.i_rwsem, i); \ if ((_locks) & INODE_PAGECACHE_BLOCK) \ - bch2_pagecache_block_get(&a[i]->ei_pagecache_lock);\ + bch2_pagecache_block_get(a[i]);\ if ((_locks) & INODE_UPDATE_LOCK) \ mutex_lock_nested(&a[i]->ei_update_lock, i);\ } \ @@ -98,10 +106,8 @@ do { \ \ for (i = 1; i < ARRAY_SIZE(a); i++) \ if (a[i] != a[i - 1]) { \ - if ((_locks) & INODE_LOCK) \ - up_write(&a[i]->v.i_rwsem); \ if ((_locks) & INODE_PAGECACHE_BLOCK) \ - bch2_pagecache_block_put(&a[i]->ei_pagecache_lock);\ + bch2_pagecache_block_put(a[i]);\ if ((_locks) & INODE_UPDATE_LOCK) \ mutex_unlock(&a[i]->ei_update_lock); \ } \ @@ -137,6 +143,10 @@ struct bch_inode_unpacked; #ifndef NO_BCACHEFS_FS +struct bch_inode_info * +__bch2_create(struct mnt_idmap *, struct bch_inode_info *, + struct dentry *, umode_t, dev_t, subvol_inum, unsigned); + int bch2_fs_quota_transfer(struct bch_fs *, struct bch_inode_info *, struct bch_qid, @@ -156,24 +166,36 @@ static inline int bch2_set_projid(struct bch_fs *c, KEY_TYPE_QUOTA_PREALLOC); } -struct inode *bch2_vfs_inode_get(struct bch_fs *, u64); +struct inode *bch2_vfs_inode_get(struct bch_fs *, subvol_inum); /* returns 0 if we want to do the update, or error is passed up */ -typedef int (*inode_set_fn)(struct bch_inode_info *, +typedef int (*inode_set_fn)(struct btree_trans *, + struct bch_inode_info *, struct bch_inode_unpacked *, void *); -void bch2_inode_update_after_write(struct bch_fs *, +void bch2_inode_update_after_write(struct btree_trans *, struct bch_inode_info *, struct bch_inode_unpacked *, unsigned); int __must_check bch2_write_inode(struct bch_fs *, struct bch_inode_info *, inode_set_fn, void *, unsigned); +int bch2_setattr_nonsize(struct mnt_idmap *, + struct bch_inode_info *, + struct iattr *); +int __bch2_unlink(struct inode *, struct dentry *, bool); + +void bch2_evict_subvolume_inodes(struct bch_fs *, snapshot_id_list *); + void bch2_vfs_exit(void); int bch2_vfs_init(void); #else +#define bch2_inode_update_after_write(_trans, _inode, _inode_u, _fields) ({ do {} while (0); }) + +static inline void bch2_evict_subvolume_inodes(struct bch_fs *c, + snapshot_id_list *s) {} static inline void bch2_vfs_exit(void) {} static inline int bch2_vfs_init(void) { return 0; }