static int xfs_iget_cache_miss( struct xfs_mount *mp, struct xfs_perag *pag, xfs_trans_t *tp, xfs_ino_t ino, struct xfs_inode **ipp, int flags, int lock_flags) { struct xfs_inode *ip; int error; xfs_agino_t agino = XFS_INO_TO_AGINO(mp, ino); int iflags; ip = xfs_inode_alloc(mp, ino); if (!ip) return ENOMEM; error = xfs_iread(mp, tp, ip, flags); if (error) goto out_destroy; trace_xfs_iget_miss(ip); if ((ip->i_d.di_mode == 0) && !(flags & XFS_IGET_CREATE)) { error = ENOENT; goto out_destroy; } /* * Preload the radix tree so we can insert safely under the * write spinlock. Note that we cannot sleep inside the preload * region. */ if (radix_tree_preload(GFP_KERNEL)) { error = EAGAIN; goto out_destroy; } /* * Because the inode hasn't been added to the radix-tree yet it can't * be found by another thread, so we can do the non-sleeping lock here. */ if (lock_flags) { if (!xfs_ilock_nowait(ip, lock_flags)) BUG(); } /* * These values must be set before inserting the inode into the radix * tree as the moment it is inserted a concurrent lookup (allowed by the * RCU locking mechanism) can find it and that lookup must see that this * is an inode currently under construction (i.e. that XFS_INEW is set). * The ip->i_flags_lock that protects the XFS_INEW flag forms the * memory barrier that ensures this detection works correctly at lookup * time. */ iflags = XFS_INEW; if (flags & XFS_IGET_DONTCACHE) iflags |= XFS_IDONTCACHE; ip->i_udquot = ip->i_gdquot = NULL; xfs_iflags_set(ip, iflags); /* insert the new inode */ spin_lock(&pag->pag_ici_lock); error = radix_tree_insert(&pag->pag_ici_root, agino, ip); if (unlikely(error)) { WARN_ON(error != -EEXIST); XFS_STATS_INC(xs_ig_dup); error = EAGAIN; goto out_preload_end; } spin_unlock(&pag->pag_ici_lock); radix_tree_preload_end(); *ipp = ip; return 0; out_preload_end: spin_unlock(&pag->pag_ici_lock); radix_tree_preload_end(); if (lock_flags) xfs_iunlock(ip, lock_flags); out_destroy: __destroy_inode(VFS_I(ip)); xfs_inode_free(ip); return error; }
/* * Look up an inode by number in the given file system. * The inode is looked up in the hash table for the file system * represented by the mount point parameter mp. Each bucket of * the hash table is guarded by an individual semaphore. * * If the inode is found in the hash table, its corresponding vnode * is obtained with a call to vn_get(). This call takes care of * coordination with the reclamation of the inode and vnode. Note * that the vmap structure is filled in while holding the hash lock. * This gives us the state of the inode/vnode when we found it and * is used for coordination in vn_get(). * * If it is not in core, read it in from the file system's device and * add the inode into the hash table. * * The inode is locked according to the value of the lock_flags parameter. * This flag parameter indicates how and if the inode's IO lock and inode lock * should be taken. * * mp -- the mount point structure for the current file system. It points * to the inode hash table. * tp -- a pointer to the current transaction if there is one. This is * simply passed through to the xfs_iread() call. * ino -- the number of the inode desired. This is the unique identifier * within the file system for the inode being requested. * lock_flags -- flags indicating how to lock the inode. See the comment * for xfs_ilock() for a list of valid values. * bno -- the block number starting the buffer containing the inode, * if known (as by bulkstat), else 0. */ int xfs_iget( xfs_mount_t *mp, xfs_trans_t *tp, xfs_ino_t ino, uint flags, uint lock_flags, xfs_inode_t **ipp, xfs_daddr_t bno) { xfs_ihash_t *ih; xfs_inode_t *ip; xfs_inode_t *iq; xfs_vnode_t *vp; ulong version; int error; /* REFERENCED */ int newnode; xfs_chash_t *ch; xfs_chashlist_t *chl, *chlnew; vmap_t vmap; SPLDECL(s); XFS_STATS_INC(xs_ig_attempts); ih = XFS_IHASH(mp, ino); again: read_lock(&ih->ih_lock); for (ip = ih->ih_next; ip != NULL; ip = ip->i_next) { if (ip->i_ino == ino) { vp = XFS_ITOV(ip); VMAP(vp, vmap); /* * Inode cache hit: if ip is not at the front of * its hash chain, move it there now. * Do this with the lock held for update, but * do statistics after releasing the lock. */ if (ip->i_prevp != &ih->ih_next && rwlock_trypromote(&ih->ih_lock)) { if ((iq = ip->i_next)) { iq->i_prevp = ip->i_prevp; } *ip->i_prevp = iq; iq = ih->ih_next; iq->i_prevp = &ip->i_next; ip->i_next = iq; ip->i_prevp = &ih->ih_next; ih->ih_next = ip; write_unlock(&ih->ih_lock); } else { read_unlock(&ih->ih_lock); } XFS_STATS_INC(xs_ig_found); /* * Get a reference to the vnode/inode. * vn_get() takes care of coordination with * the file system inode release and reclaim * functions. If it returns NULL, the inode * has been reclaimed so just start the search * over again. We probably won't find it, * but we could be racing with another cpu * looking for the same inode so we have to at * least look. */ if (!(vp = vn_get(vp, &vmap))) { XFS_STATS_INC(xs_ig_frecycle); goto again; } if (lock_flags != 0) { ip->i_flags &= ~XFS_IRECLAIM; xfs_ilock(ip, lock_flags); } newnode = (ip->i_d.di_mode == 0); if (newnode) { xfs_iocore_inode_reinit(ip); } ip->i_flags &= ~XFS_ISTALE; vn_trace_exit(vp, "xfs_iget.found", (inst_t *)__return_address); goto return_ip; } } /* * Inode cache miss: save the hash chain version stamp and unlock * the chain, so we don't deadlock in vn_alloc. */ XFS_STATS_INC(xs_ig_missed); version = ih->ih_version; read_unlock(&ih->ih_lock); /* * Read the disk inode attributes into a new inode structure and get * a new vnode for it. This should also initialize i_ino and i_mount. */ error = xfs_iread(mp, tp, ino, &ip, bno); if (error) { return error; } error = xfs_vn_allocate(mp, ip, &vp); if (error) { return error; } vn_trace_exit(vp, "xfs_iget.alloc", (inst_t *)__return_address); xfs_inode_lock_init(ip, vp); xfs_iocore_inode_init(ip); if (lock_flags != 0) { xfs_ilock(ip, lock_flags); } /* * Put ip on its hash chain, unless someone else hashed a duplicate * after we released the hash lock. */ write_lock(&ih->ih_lock); if (ih->ih_version != version) { for (iq = ih->ih_next; iq != NULL; iq = iq->i_next) { if (iq->i_ino == ino) { write_unlock(&ih->ih_lock); xfs_idestroy(ip); XFS_STATS_INC(xs_ig_dup); goto again; } } } /* * These values _must_ be set before releasing ihlock! */ ip->i_hash = ih; if ((iq = ih->ih_next)) { iq->i_prevp = &ip->i_next; } ip->i_next = iq; ip->i_prevp = &ih->ih_next; ih->ih_next = ip; ip->i_udquot = ip->i_gdquot = NULL; ih->ih_version++; write_unlock(&ih->ih_lock); /* * put ip on its cluster's hash chain */ ASSERT(ip->i_chash == NULL && ip->i_cprev == NULL && ip->i_cnext == NULL); chlnew = NULL; ch = XFS_CHASH(mp, ip->i_blkno); chlredo: s = mutex_spinlock(&ch->ch_lock); for (chl = ch->ch_list; chl != NULL; chl = chl->chl_next) { if (chl->chl_blkno == ip->i_blkno) { /* insert this inode into the doubly-linked list * where chl points */ if ((iq = chl->chl_ip)) { ip->i_cprev = iq->i_cprev; iq->i_cprev->i_cnext = ip; iq->i_cprev = ip; ip->i_cnext = iq; } else { ip->i_cnext = ip; ip->i_cprev = ip; } chl->chl_ip = ip; ip->i_chash = chl; break; } } /* no hash list found for this block; add a new hash list */ if (chl == NULL) { if (chlnew == NULL) { mutex_spinunlock(&ch->ch_lock, s); ASSERT(xfs_chashlist_zone != NULL); chlnew = (xfs_chashlist_t *) kmem_zone_alloc(xfs_chashlist_zone, KM_SLEEP); ASSERT(chlnew != NULL); goto chlredo; } else { ip->i_cnext = ip; ip->i_cprev = ip; ip->i_chash = chlnew; chlnew->chl_ip = ip; chlnew->chl_blkno = ip->i_blkno; chlnew->chl_next = ch->ch_list; ch->ch_list = chlnew; chlnew = NULL; } } else { if (chlnew != NULL) { kmem_zone_free(xfs_chashlist_zone, chlnew); } } mutex_spinunlock(&ch->ch_lock, s); /* * Link ip to its mount and thread it on the mount's inode list. */ XFS_MOUNT_ILOCK(mp); if ((iq = mp->m_inodes)) { ASSERT(iq->i_mprev->i_mnext == iq); ip->i_mprev = iq->i_mprev; iq->i_mprev->i_mnext = ip; iq->i_mprev = ip; ip->i_mnext = iq; } else { ip->i_mnext = ip; ip->i_mprev = ip; } mp->m_inodes = ip; XFS_MOUNT_IUNLOCK(mp); newnode = 1; return_ip: ASSERT(ip->i_df.if_ext_max == XFS_IFORK_DSIZE(ip) / sizeof(xfs_bmbt_rec_t)); ASSERT(((ip->i_d.di_flags & XFS_DIFLAG_REALTIME) != 0) == ((ip->i_iocore.io_flags & XFS_IOCORE_RT) != 0)); *ipp = ip; /* * If we have a real type for an on-disk inode, we can set ops(&unlock) * now. If it's a new inode being created, xfs_ialloc will handle it. */ XVFS_INIT_VNODE(XFS_MTOVFS(mp), vp, XFS_ITOBHV(ip), 1); return 0; }
static int xfs_iget_cache_miss( struct xfs_mount *mp, struct xfs_perag *pag, xfs_trans_t *tp, xfs_ino_t ino, struct xfs_inode **ipp, int flags, int lock_flags) { struct xfs_inode *ip; int error; xfs_agino_t agino = XFS_INO_TO_AGINO(mp, ino); ip = xfs_inode_alloc(mp, ino); if (!ip) return ENOMEM; error = xfs_iread(mp, tp, ip, flags); if (error) goto out_destroy; trace_xfs_iget_miss(ip); if ((ip->i_d.di_mode == 0) && !(flags & XFS_IGET_CREATE)) { error = ENOENT; goto out_destroy; } /* * Preload the radix tree so we can insert safely under the * write spinlock. Note that we cannot sleep inside the preload * region. */ if (radix_tree_preload(GFP_KERNEL)) { error = EAGAIN; goto out_destroy; } /* * Because the inode hasn't been added to the radix-tree yet it can't * be found by another thread, so we can do the non-sleeping lock here. */ if (lock_flags) { if (!xfs_ilock_nowait(ip, lock_flags)) BUG(); } spin_lock(&pag->pag_ici_lock); /* insert the new inode */ error = radix_tree_insert(&pag->pag_ici_root, agino, ip); if (unlikely(error)) { WARN_ON(error != -EEXIST); XFS_STATS_INC(xs_ig_dup); error = EAGAIN; goto out_preload_end; } /* These values _must_ be set before releasing the radix tree lock! */ ip->i_udquot = ip->i_gdquot = NULL; xfs_iflags_set(ip, XFS_INEW); spin_unlock(&pag->pag_ici_lock); radix_tree_preload_end(); *ipp = ip; return 0; out_preload_end: spin_unlock(&pag->pag_ici_lock); radix_tree_preload_end(); if (lock_flags) xfs_iunlock(ip, lock_flags); out_destroy: __destroy_inode(VFS_I(ip)); xfs_inode_free(ip); return error; }
static int xfs_iget_cache_miss( struct xfs_mount *mp, struct xfs_perag *pag, xfs_trans_t *tp, xfs_ino_t ino, struct xfs_inode **ipp, int flags, int lock_flags) { struct xfs_inode *ip; int error; xfs_agino_t agino = XFS_INO_TO_AGINO(mp, ino); int iflags; ip = xfs_inode_alloc(mp, ino); if (!ip) return ENOMEM; error = xfs_iread(mp, tp, ip, flags); if (error) goto out_destroy; trace_xfs_iget_miss(ip); if ((ip->i_d.di_mode == 0) && !(flags & XFS_IGET_CREATE)) { error = ENOENT; goto out_destroy; } if (radix_tree_preload(GFP_KERNEL)) { error = EAGAIN; goto out_destroy; } if (lock_flags) { if (!xfs_ilock_nowait(ip, lock_flags)) BUG(); } iflags = XFS_INEW; if (flags & XFS_IGET_DONTCACHE) iflags |= XFS_IDONTCACHE; ip->i_udquot = ip->i_gdquot = NULL; xfs_iflags_set(ip, iflags); spin_lock(&pag->pag_ici_lock); error = radix_tree_insert(&pag->pag_ici_root, agino, ip); if (unlikely(error)) { WARN_ON(error != -EEXIST); XFS_STATS_INC(xs_ig_dup); error = EAGAIN; goto out_preload_end; } spin_unlock(&pag->pag_ici_lock); radix_tree_preload_end(); *ipp = ip; return 0; out_preload_end: spin_unlock(&pag->pag_ici_lock); radix_tree_preload_end(); if (lock_flags) xfs_iunlock(ip, lock_flags); out_destroy: __destroy_inode(VFS_I(ip)); xfs_inode_free(ip); return error; }