X-Git-Url: http://ftp.safe.ca/?a=blobdiff_plain;f=fs%2Fxfs%2Fxfs_iget.c;h=e2fb6210d4c58e7e9609d9e5c584210d5a4058c3;hb=4f2d4ac6e5eb7d72e8df7f3fbf67a78dab8b91cf;hp=8e380a1fb79b691f03efa0e0aa126ce9c0e78727;hpb=75e17b3caf29b262000dc7348f1be9a7d5403463;p=safe%2Fjmp%2Flinux-2.6 diff --git a/fs/xfs/xfs_iget.c b/fs/xfs/xfs_iget.c index 8e380a1..e2fb621 100644 --- a/fs/xfs/xfs_iget.c +++ b/fs/xfs/xfs_iget.c @@ -24,14 +24,12 @@ #include "xfs_trans.h" #include "xfs_sb.h" #include "xfs_ag.h" -#include "xfs_dir.h" #include "xfs_dir2.h" #include "xfs_dmapi.h" #include "xfs_mount.h" #include "xfs_bmap_btree.h" #include "xfs_alloc_btree.h" #include "xfs_ialloc_btree.h" -#include "xfs_dir_sf.h" #include "xfs_dir2_sf.h" #include "xfs_attr_sf.h" #include "xfs_dinode.h" @@ -40,134 +38,268 @@ #include "xfs_ialloc.h" #include "xfs_quota.h" #include "xfs_utils.h" +#include "xfs_trans_priv.h" +#include "xfs_inode_item.h" +#include "xfs_bmap.h" +#include "xfs_btree_trace.h" +#include "xfs_dir2_trace.h" + /* - * Initialize the inode hash table for the newly mounted file system. - * Choose an initial table size based on user specified value, else - * use a simple algorithm using the maximum number of inodes as an - * indicator for table size, and clamp it between one and some large - * number of pages. + * Allocate and initialise an xfs_inode. */ -void -xfs_ihash_init(xfs_mount_t *mp) +STATIC struct xfs_inode * +xfs_inode_alloc( + struct xfs_mount *mp, + xfs_ino_t ino) { - __uint64_t icount; - uint i, flags = KM_SLEEP | KM_MAYFAIL; - - if (!mp->m_ihsize) { - icount = mp->m_maxicount ? mp->m_maxicount : - (mp->m_sb.sb_dblocks << mp->m_sb.sb_inopblog); - mp->m_ihsize = 1 << max_t(uint, 8, - (xfs_highbit64(icount) + 1) / 2); - mp->m_ihsize = min_t(uint, mp->m_ihsize, - (64 * NBPP) / sizeof(xfs_ihash_t)); - } + struct xfs_inode *ip; - while (!(mp->m_ihash = (xfs_ihash_t *)kmem_zalloc(mp->m_ihsize * - sizeof(xfs_ihash_t), flags))) { - if ((mp->m_ihsize >>= 1) <= NBPP) - flags = KM_SLEEP; - } - for (i = 0; i < mp->m_ihsize; i++) { - rwlock_init(&(mp->m_ihash[i].ih_lock)); + /* + * if this didn't occur in transactions, we could use + * KM_MAYFAIL and return NULL here on ENOMEM. Set the + * code up to do this anyway. + */ + ip = kmem_zone_alloc(xfs_inode_zone, KM_SLEEP); + if (!ip) + return NULL; + + ASSERT(atomic_read(&ip->i_iocount) == 0); + ASSERT(atomic_read(&ip->i_pincount) == 0); + ASSERT(!spin_is_locked(&ip->i_flags_lock)); + ASSERT(completion_done(&ip->i_flush)); + + /* + * initialise the VFS inode here to get failures + * out of the way early. + */ + if (!inode_init_always(mp->m_super, VFS_I(ip))) { + kmem_zone_free(xfs_inode_zone, ip); + return NULL; } -} -/* - * Free up structures allocated by xfs_ihash_init, at unmount time. - */ -void -xfs_ihash_free(xfs_mount_t *mp) -{ - kmem_free(mp->m_ihash, mp->m_ihsize*sizeof(xfs_ihash_t)); - mp->m_ihash = NULL; + /* initialise the xfs inode */ + ip->i_ino = ino; + ip->i_mount = mp; + memset(&ip->i_imap, 0, sizeof(struct xfs_imap)); + ip->i_afp = NULL; + memset(&ip->i_df, 0, sizeof(xfs_ifork_t)); + ip->i_flags = 0; + ip->i_update_core = 0; + ip->i_update_size = 0; + ip->i_delayed_blks = 0; + memset(&ip->i_d, 0, sizeof(xfs_icdinode_t)); + ip->i_size = 0; + ip->i_new_size = 0; + + /* + * Initialize inode's trace buffers. + */ +#ifdef XFS_INODE_TRACE + ip->i_trace = ktrace_alloc(INODE_TRACE_SIZE, KM_NOFS); +#endif +#ifdef XFS_BMAP_TRACE + ip->i_xtrace = ktrace_alloc(XFS_BMAP_KTRACE_SIZE, KM_NOFS); +#endif +#ifdef XFS_BTREE_TRACE + ip->i_btrace = ktrace_alloc(XFS_BMBT_KTRACE_SIZE, KM_NOFS); +#endif +#ifdef XFS_RW_TRACE + ip->i_rwtrace = ktrace_alloc(XFS_RW_KTRACE_SIZE, KM_NOFS); +#endif +#ifdef XFS_ILOCK_TRACE + ip->i_lock_trace = ktrace_alloc(XFS_ILOCK_KTRACE_SIZE, KM_NOFS); +#endif +#ifdef XFS_DIR2_TRACE + ip->i_dir_trace = ktrace_alloc(XFS_DIR2_KTRACE_SIZE, KM_NOFS); +#endif + + return ip; } /* - * Initialize the inode cluster hash table for the newly mounted file system. - * Its size is derived from the ihash table size. + * Check the validity of the inode we just found it the cache */ -void -xfs_chash_init(xfs_mount_t *mp) +static int +xfs_iget_cache_hit( + struct xfs_perag *pag, + struct xfs_inode *ip, + int flags, + int lock_flags) __releases(pag->pag_ici_lock) { - uint i; - - mp->m_chsize = max_t(uint, 1, mp->m_ihsize / - (XFS_INODE_CLUSTER_SIZE(mp) >> mp->m_sb.sb_inodelog)); - mp->m_chsize = min_t(uint, mp->m_chsize, mp->m_ihsize); - mp->m_chash = (xfs_chash_t *)kmem_zalloc(mp->m_chsize - * sizeof(xfs_chash_t), - KM_SLEEP); - for (i = 0; i < mp->m_chsize; i++) { - spinlock_init(&mp->m_chash[i].ch_lock,"xfshash"); + struct xfs_mount *mp = ip->i_mount; + int error = EAGAIN; + + /* + * If INEW is set this inode is being set up + * If IRECLAIM is set this inode is being torn down + * Pause and try again. + */ + if (xfs_iflags_test(ip, (XFS_INEW|XFS_IRECLAIM))) { + XFS_STATS_INC(xs_ig_frecycle); + goto out_error; } -} -/* - * Free up structures allocated by xfs_chash_init, at unmount time. - */ -void -xfs_chash_free(xfs_mount_t *mp) -{ - int i; + /* If IRECLAIMABLE is set, we've torn down the vfs inode part */ + if (xfs_iflags_test(ip, XFS_IRECLAIMABLE)) { + + /* + * If lookup is racing with unlink, then we should return an + * error immediately so we don't remove it from the reclaim + * list and potentially leak the inode. + */ + if ((ip->i_d.di_mode == 0) && !(flags & XFS_IGET_CREATE)) { + error = ENOENT; + goto out_error; + } - for (i = 0; i < mp->m_chsize; i++) { - spinlock_destroy(&mp->m_chash[i].ch_lock); + xfs_itrace_exit_tag(ip, "xfs_iget.alloc"); + + /* + * We need to re-initialise the VFS inode as it has been + * 'freed' by the VFS. Do this here so we can deal with + * errors cleanly, then tag it so it can be set up correctly + * later. + */ + if (!inode_init_always(mp->m_super, VFS_I(ip))) { + error = ENOMEM; + goto out_error; + } + + /* + * We must set the XFS_INEW flag before clearing the + * XFS_IRECLAIMABLE flag so that if a racing lookup does + * not find the XFS_IRECLAIMABLE above but has the igrab() + * below succeed we can safely check XFS_INEW to detect + * that this inode is still being initialised. + */ + xfs_iflags_set(ip, XFS_INEW); + xfs_iflags_clear(ip, XFS_IRECLAIMABLE); + + /* clear the radix tree reclaim flag as well. */ + __xfs_inode_clear_reclaim_tag(mp, pag, ip); + } else if (!igrab(VFS_I(ip))) { + /* If the VFS inode is being torn down, pause and try again. */ + XFS_STATS_INC(xs_ig_frecycle); + goto out_error; + } else if (xfs_iflags_test(ip, XFS_INEW)) { + /* + * We are racing with another cache hit that is + * currently recycling this inode out of the XFS_IRECLAIMABLE + * state. Wait for the initialisation to complete before + * continuing. + */ + wait_on_inode(VFS_I(ip)); + } + + if (ip->i_d.di_mode == 0 && !(flags & XFS_IGET_CREATE)) { + error = ENOENT; + iput(VFS_I(ip)); + goto out_error; } - kmem_free(mp->m_chash, mp->m_chsize*sizeof(xfs_chash_t)); - mp->m_chash = NULL; + /* We've got a live one. */ + read_unlock(&pag->pag_ici_lock); + + if (lock_flags != 0) + xfs_ilock(ip, lock_flags); + + xfs_iflags_clear(ip, XFS_ISTALE); + xfs_itrace_exit_tag(ip, "xfs_iget.found"); + XFS_STATS_INC(xs_ig_found); + return 0; + +out_error: + read_unlock(&pag->pag_ici_lock); + return error; } -/* - * Try to move an inode to the front of its hash list if possible - * (and if its not there already). Called right after obtaining - * the list version number and then dropping the read_lock on the - * hash list in question (which is done right after looking up the - * inode in question...). - */ -STATIC void -xfs_ihash_promote( - xfs_ihash_t *ih, - xfs_inode_t *ip, - ulong version) + +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, + xfs_daddr_t bno, + int flags, + int lock_flags) __releases(pag->pag_ici_lock) { - xfs_inode_t *iq; - - if ((ip->i_prevp != &ih->ih_next) && write_trylock(&ih->ih_lock)) { - if (likely(version == ih->ih_version)) { - /* remove from list */ - if ((iq = ip->i_next)) { - iq->i_prevp = ip->i_prevp; - } - *ip->i_prevp = iq; - - /* insert at list head */ - 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); + struct xfs_inode *ip; + int error; + unsigned long first_index, mask; + 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, bno, flags); + if (error) + goto out_destroy; + + xfs_itrace_exit_tag(ip, "xfs_iget.alloc"); + + if ((ip->i_d.di_mode == 0) && !(flags & XFS_IGET_CREATE)) { + error = ENOENT; + goto out_destroy; + } + + if (lock_flags) + xfs_ilock(ip, lock_flags); + + /* + * 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_unlock; + } + + mask = ~(((XFS_INODE_CLUSTER_SIZE(mp) >> mp->m_sb.sb_inodelog)) - 1); + first_index = agino & mask; + write_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); + + write_unlock(&pag->pag_ici_lock); + radix_tree_preload_end(); + *ipp = ip; + return 0; + +out_preload_end: + write_unlock(&pag->pag_ici_lock); + radix_tree_preload_end(); +out_unlock: + if (lock_flags) + xfs_iunlock(ip, lock_flags); +out_destroy: + xfs_destroy_inode(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. + * The inode is looked up in the cache held in each AG. + * If the inode is found in the cache, initialise the vfs inode + * if necessary. * - * 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. + * If it is not in core, read it in from the file system's device, + * add it to the cache and initialise the vfs inode. * * 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 @@ -184,9 +316,8 @@ xfs_ihash_promote( * bno -- the block number starting the buffer containing the inode, * if known (as by bulkstat), else 0. */ -STATIC int -xfs_iget_core( - vnode_t *vp, +int +xfs_iget( xfs_mount_t *mp, xfs_trans_t *tp, xfs_ino_t ino, @@ -195,360 +326,63 @@ xfs_iget_core( xfs_inode_t **ipp, xfs_daddr_t bno) { - xfs_ihash_t *ih; xfs_inode_t *ip; - xfs_inode_t *iq; - vnode_t *inode_vp; - ulong version; int error; - /* REFERENCED */ - xfs_chash_t *ch; - xfs_chashlist_t *chl, *chlnew; - SPLDECL(s); + xfs_perag_t *pag; + xfs_agino_t agino; + /* the radix tree exists only in inode capable AGs */ + if (XFS_INO_TO_AGNO(mp, ino) >= mp->m_maxagi) + return EINVAL; - ih = XFS_IHASH(mp, ino); + /* get the perag structure and ensure that it's inode capable */ + pag = xfs_get_perag(mp, ino); + if (!pag->pagi_inodeok) + return EINVAL; + ASSERT(pag->pag_ici_init); + agino = XFS_INO_TO_AGINO(mp, ino); again: - read_lock(&ih->ih_lock); - - for (ip = ih->ih_next; ip != NULL; ip = ip->i_next) { - if (ip->i_ino == ino) { - /* - * If INEW is set this inode is being set up - * we need to pause and try again. - */ - if (ip->i_flags & XFS_INEW) { - read_unlock(&ih->ih_lock); - delay(1); - XFS_STATS_INC(xs_ig_frecycle); - - goto again; - } - - inode_vp = XFS_ITOV_NULL(ip); - if (inode_vp == NULL) { - /* - * If IRECLAIM is set this inode is - * on its way out of the system, - * we need to pause and try again. - */ - if (ip->i_flags & XFS_IRECLAIM) { - read_unlock(&ih->ih_lock); - delay(1); - XFS_STATS_INC(xs_ig_frecycle); - - goto again; - } - - vn_trace_exit(vp, "xfs_iget.alloc", - (inst_t *)__return_address); - - XFS_STATS_INC(xs_ig_found); - - ip->i_flags &= ~XFS_IRECLAIMABLE; - version = ih->ih_version; - read_unlock(&ih->ih_lock); - xfs_ihash_promote(ih, ip, version); - - XFS_MOUNT_ILOCK(mp); - list_del_init(&ip->i_reclaim); - XFS_MOUNT_IUNLOCK(mp); - - goto finish_inode; - - } else if (vp != inode_vp) { - struct inode *inode = LINVFS_GET_IP(inode_vp); - - /* The inode is being torn down, pause and - * try again. - */ - if (inode->i_state & (I_FREEING | I_CLEAR)) { - read_unlock(&ih->ih_lock); - delay(1); - XFS_STATS_INC(xs_ig_frecycle); - - goto again; - } -/* Chances are the other vnode (the one in the inode) is being torn - * down right now, and we landed on top of it. Question is, what do - * we do? Unhook the old inode and hook up the new one? - */ - cmn_err(CE_PANIC, - "xfs_iget_core: ambiguous vns: vp/0x%p, invp/0x%p", - inode_vp, vp); - } - - /* - * 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. - */ - version = ih->ih_version; - read_unlock(&ih->ih_lock); - xfs_ihash_promote(ih, ip, version); - XFS_STATS_INC(xs_ig_found); - -finish_inode: - if (ip->i_d.di_mode == 0) { - if (!(flags & IGET_CREATE)) - return ENOENT; - xfs_iocore_inode_reinit(ip); - } - - if (lock_flags != 0) - xfs_ilock(ip, lock_flags); - - 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; - } - - 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); - } - - if ((ip->i_d.di_mode == 0) && !(flags & IGET_CREATE)) { - xfs_idestroy(ip); - return ENOENT; - } - - /* - * 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++; - ip->i_flags |= XFS_INEW; - - 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; - } + error = 0; + read_lock(&pag->pag_ici_lock); + ip = radix_tree_lookup(&pag->pag_ici_root, agino); + + if (ip) { + error = xfs_iget_cache_hit(pag, ip, flags, lock_flags); + if (error) + goto out_error_or_again; } else { - if (chlnew != NULL) { - kmem_zone_free(xfs_chashlist_zone, chlnew); - } - } - - mutex_spinunlock(&ch->ch_lock, s); + read_unlock(&pag->pag_ici_lock); + XFS_STATS_INC(xs_ig_missed); - - /* - * 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; + error = xfs_iget_cache_miss(mp, pag, tp, ino, &ip, bno, + flags, lock_flags); + if (error) + goto out_error_or_again; } - mp->m_inodes = ip; + xfs_put_perag(mp, pag); - XFS_MOUNT_IUNLOCK(mp); + *ipp = ip; - 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. */ - VFS_INIT_VNODE(XFS_MTOVFS(mp), vp, XFS_ITOBHV(ip), 1); - + if (xfs_iflags_test(ip, XFS_INEW) && ip->i_d.di_mode != 0) + xfs_setup_inode(ip); return 0; -} - - -/* - * The 'normal' internal xfs_iget, if needed it will - * 'allocate', or 'get', the vnode. - */ -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) -{ - struct inode *inode; - vnode_t *vp = NULL; - int error; - - XFS_STATS_INC(xs_ig_attempts); - -retry: - if ((inode = iget_locked(XFS_MTOVFS(mp)->vfs_super, ino))) { - xfs_inode_t *ip; - - vp = LINVFS_GET_VP(inode); - if (inode->i_state & I_NEW) { - vn_initialize(inode); - error = xfs_iget_core(vp, mp, tp, ino, flags, - lock_flags, ipp, bno); - if (error) { - vn_mark_bad(vp); - if (inode->i_state & I_NEW) - unlock_new_inode(inode); - iput(inode); - } - } else { - /* - * If the inode is not fully constructed due to - * filehandle mistmatches wait for the inode to go - * away and try again. - * - * iget_locked will call __wait_on_freeing_inode - * to wait for the inode to go away. - */ - if (is_bad_inode(inode) || - ((ip = xfs_vtoi(vp)) == NULL)) { - iput(inode); - delay(1); - goto retry; - } - - if (lock_flags != 0) - xfs_ilock(ip, lock_flags); - XFS_STATS_INC(xs_ig_found); - *ipp = ip; - error = 0; - } - } else - error = ENOMEM; /* If we got no inode we are out of memory */ +out_error_or_again: + if (error == EAGAIN) { + delay(1); + goto again; + } + xfs_put_perag(mp, pag); return error; } -/* - * Do the setup for the various locks within the incore inode. - */ -void -xfs_inode_lock_init( - xfs_inode_t *ip, - vnode_t *vp) -{ - mrlock_init(&ip->i_lock, MRLOCK_ALLOW_EQUAL_PRI|MRLOCK_BARRIER, - "xfsino", (long)vp->v_number); - mrlock_init(&ip->i_iolock, MRLOCK_BARRIER, "xfsio", vp->v_number); - init_waitqueue_head(&ip->i_ipin_wait); - atomic_set(&ip->i_pincount, 0); - init_sema(&ip->i_flock, 1, "xfsfino", vp->v_number); -} /* * Look for the inode corresponding to the given ino in the hash table. @@ -560,32 +394,19 @@ xfs_inode_incore(xfs_mount_t *mp, xfs_ino_t ino, xfs_trans_t *tp) { - xfs_ihash_t *ih; xfs_inode_t *ip; - ulong version; - - ih = XFS_IHASH(mp, ino); - read_lock(&ih->ih_lock); - for (ip = ih->ih_next; ip != NULL; ip = ip->i_next) { - if (ip->i_ino == ino) { - /* - * If we find it and tp matches, return it. - * Also move it to the front of the hash list - * if we find it and it is not already there. - * Otherwise break from the loop and return - * NULL. - */ - if (ip->i_transp == tp) { - version = ih->ih_version; - read_unlock(&ih->ih_lock); - xfs_ihash_promote(ih, ip, version); - return (ip); - } - break; - } - } - read_unlock(&ih->ih_lock); - return (NULL); + xfs_perag_t *pag; + + pag = xfs_get_perag(mp, ino); + read_lock(&pag->pag_ici_lock); + ip = radix_tree_lookup(&pag->pag_ici_root, XFS_INO_TO_AGINO(mp, ino)); + read_unlock(&pag->pag_ici_lock); + xfs_put_perag(mp, pag); + + /* the returned inode must match the transaction */ + if (ip && (ip->i_transp != tp)) + return NULL; + return ip; } /* @@ -600,184 +421,137 @@ void xfs_iput(xfs_inode_t *ip, uint lock_flags) { - vnode_t *vp = XFS_ITOV(ip); - - vn_trace_entry(vp, "xfs_iput", (inst_t *)__return_address); - + xfs_itrace_entry(ip); xfs_iunlock(ip, lock_flags); - - VN_RELE(vp); + IRELE(ip); } /* * Special iput for brand-new inodes that are still locked */ void -xfs_iput_new(xfs_inode_t *ip, - uint lock_flags) +xfs_iput_new( + xfs_inode_t *ip, + uint lock_flags) { - vnode_t *vp = XFS_ITOV(ip); - struct inode *inode = LINVFS_GET_IP(vp); + struct inode *inode = VFS_I(ip); - vn_trace_entry(vp, "xfs_iput_new", (inst_t *)__return_address); + xfs_itrace_entry(ip); if ((ip->i_d.di_mode == 0)) { - ASSERT(!(ip->i_flags & XFS_IRECLAIMABLE)); - vn_mark_bad(vp); + ASSERT(!xfs_iflags_test(ip, XFS_IRECLAIMABLE)); + make_bad_inode(inode); } if (inode->i_state & I_NEW) unlock_new_inode(inode); if (lock_flags) xfs_iunlock(ip, lock_flags); - VN_RELE(vp); + IRELE(ip); } - /* - * This routine embodies the part of the reclaim code that pulls - * the inode from the inode hash table and the mount structure's - * inode list. - * This should only be called from xfs_reclaim(). + * This is called free all the memory associated with an inode. + * It must free the inode itself and any buffers allocated for + * if_extents/if_data and if_broot. It must also free the lock + * associated with the inode. + * + * Note: because we don't initialise everything on reallocation out + * of the zone, we must ensure we nullify everything correctly before + * freeing the structure. */ void -xfs_ireclaim(xfs_inode_t *ip) +xfs_ireclaim( + struct xfs_inode *ip) { - vnode_t *vp; + struct xfs_mount *mp = ip->i_mount; + struct xfs_perag *pag; - /* - * Remove from old hash list and mount list. - */ XFS_STATS_INC(xs_ig_reclaims); - xfs_iextract(ip); - - /* - * Here we do a spurious inode lock in order to coordinate with - * xfs_sync(). This is because xfs_sync() references the inodes - * in the mount list without taking references on the corresponding - * vnodes. We make that OK here by ensuring that we wait until - * the inode is unlocked in xfs_sync() before we go ahead and - * free it. We get both the regular lock and the io lock because - * the xfs_sync() code may need to drop the regular one but will - * still hold the io lock. - */ - xfs_ilock(ip, XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL); - - /* - * Release dquots (and their references) if any. An inode may escape - * xfs_inactive and get here via vn_alloc->vn_reclaim path. - */ - XFS_QM_DQDETACH(ip->i_mount, ip); - /* - * Pull our behavior descriptor from the vnode chain. + * Remove the inode from the per-AG radix tree. It doesn't matter + * if it was never added to it because radix_tree_delete can deal + * with that case just fine. */ - vp = XFS_ITOV_NULL(ip); - if (vp) { - vn_bhv_remove(VN_BHV_HEAD(vp), XFS_ITOBHV(ip)); - } + pag = xfs_get_perag(mp, ip->i_ino); + write_lock(&pag->pag_ici_lock); + radix_tree_delete(&pag->pag_ici_root, XFS_INO_TO_AGINO(mp, ip->i_ino)); + write_unlock(&pag->pag_ici_lock); + xfs_put_perag(mp, pag); /* - * Free all memory associated with the inode. + * Here we do an (almost) spurious inode lock in order to coordinate + * with inode cache radix tree lookups. This is because the lookup + * can reference the inodes in the cache without taking references. + * + * We make that OK here by ensuring that we wait until the inode is + * unlocked after the lookup before we go ahead and free it. We get + * both the ilock and the iolock because the code may need to drop the + * ilock one but will still hold the iolock. */ - xfs_idestroy(ip); -} - -/* - * This routine removes an about-to-be-destroyed inode from - * all of the lists in which it is located with the exception - * of the behavior chain. - */ -void -xfs_iextract( - xfs_inode_t *ip) -{ - xfs_ihash_t *ih; - xfs_inode_t *iq; - xfs_mount_t *mp; - xfs_chash_t *ch; - xfs_chashlist_t *chl, *chm; - SPLDECL(s); - - ih = ip->i_hash; - write_lock(&ih->ih_lock); - if ((iq = ip->i_next)) { - iq->i_prevp = ip->i_prevp; - } - *ip->i_prevp = iq; - ih->ih_version++; - write_unlock(&ih->ih_lock); - + xfs_ilock(ip, XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL); /* - * Remove from cluster hash list - * 1) delete the chashlist if this is the last inode on the chashlist - * 2) unchain from list of inodes - * 3) point chashlist->chl_ip to 'chl_next' if to this inode. + * Release dquots (and their references) if any. */ - mp = ip->i_mount; - ch = XFS_CHASH(mp, ip->i_blkno); - s = mutex_spinlock(&ch->ch_lock); - - if (ip->i_cnext == ip) { - /* Last inode on chashlist */ - ASSERT(ip->i_cnext == ip && ip->i_cprev == ip); - ASSERT(ip->i_chash != NULL); - chm=NULL; - for (chl = ch->ch_list; chl != NULL; chl = chl->chl_next) { - if (chl->chl_blkno == ip->i_blkno) { - if (chm == NULL) { - /* first item on the list */ - ch->ch_list = chl->chl_next; - } else { - chm->chl_next = chl->chl_next; - } - kmem_zone_free(xfs_chashlist_zone, chl); - break; - } else { - ASSERT(chl->chl_ip != ip); - chm = chl; - } - } - ASSERT_ALWAYS(chl != NULL); - } else { - /* delete one inode from a non-empty list */ - iq = ip->i_cnext; - iq->i_cprev = ip->i_cprev; - ip->i_cprev->i_cnext = iq; - if (ip->i_chash->chl_ip == ip) { - ip->i_chash->chl_ip = iq; - } - ip->i_chash = __return_address; - ip->i_cprev = __return_address; - ip->i_cnext = __return_address; + XFS_QM_DQDETACH(ip->i_mount, ip); + xfs_iunlock(ip, XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL); + + switch (ip->i_d.di_mode & S_IFMT) { + case S_IFREG: + case S_IFDIR: + case S_IFLNK: + xfs_idestroy_fork(ip, XFS_DATA_FORK); + break; } - mutex_spinunlock(&ch->ch_lock, s); - /* - * Remove from mount's inode list. - */ - XFS_MOUNT_ILOCK(mp); - ASSERT((ip->i_mnext != NULL) && (ip->i_mprev != NULL)); - iq = ip->i_mnext; - iq->i_mprev = ip->i_mprev; - ip->i_mprev->i_mnext = iq; - - /* - * Fix up the head pointer if it points to the inode being deleted. - */ - if (mp->m_inodes == ip) { - if (ip == iq) { - mp->m_inodes = NULL; - } else { - mp->m_inodes = iq; + if (ip->i_afp) + xfs_idestroy_fork(ip, XFS_ATTR_FORK); + +#ifdef XFS_INODE_TRACE + ktrace_free(ip->i_trace); +#endif +#ifdef XFS_BMAP_TRACE + ktrace_free(ip->i_xtrace); +#endif +#ifdef XFS_BTREE_TRACE + ktrace_free(ip->i_btrace); +#endif +#ifdef XFS_RW_TRACE + ktrace_free(ip->i_rwtrace); +#endif +#ifdef XFS_ILOCK_TRACE + ktrace_free(ip->i_lock_trace); +#endif +#ifdef XFS_DIR2_TRACE + ktrace_free(ip->i_dir_trace); +#endif + if (ip->i_itemp) { + /* + * Only if we are shutting down the fs will we see an + * inode still in the AIL. If it is there, we should remove + * it to prevent a use-after-free from occurring. + */ + xfs_log_item_t *lip = &ip->i_itemp->ili_item; + struct xfs_ail *ailp = lip->li_ailp; + + ASSERT(((lip->li_flags & XFS_LI_IN_AIL) == 0) || + XFS_FORCED_SHUTDOWN(ip->i_mount)); + if (lip->li_flags & XFS_LI_IN_AIL) { + spin_lock(&ailp->xa_lock); + if (lip->li_flags & XFS_LI_IN_AIL) + xfs_trans_ail_delete(ailp, lip); + else + spin_unlock(&ailp->xa_lock); } + xfs_inode_item_destroy(ip); + ip->i_itemp = NULL; } - - /* Deal with the deleted inodes list */ - list_del_init(&ip->i_reclaim); - - mp->m_ireclaims++; - XFS_MOUNT_IUNLOCK(mp); + /* asserts to verify all state is correct here */ + ASSERT(atomic_read(&ip->i_iocount) == 0); + ASSERT(atomic_read(&ip->i_pincount) == 0); + ASSERT(!spin_is_locked(&ip->i_flags_lock)); + ASSERT(completion_done(&ip->i_flush)); + kmem_zone_free(xfs_inode_zone, ip); } /* @@ -849,8 +623,9 @@ xfs_iunlock_map_shared( * XFS_IOLOCK_EXCL | XFS_ILOCK_EXCL */ void -xfs_ilock(xfs_inode_t *ip, - uint lock_flags) +xfs_ilock( + xfs_inode_t *ip, + uint lock_flags) { /* * You can't set both SHARED and EXCL for the same lock, @@ -861,18 +636,18 @@ xfs_ilock(xfs_inode_t *ip, (XFS_IOLOCK_SHARED | XFS_IOLOCK_EXCL)); ASSERT((lock_flags & (XFS_ILOCK_SHARED | XFS_ILOCK_EXCL)) != (XFS_ILOCK_SHARED | XFS_ILOCK_EXCL)); - ASSERT((lock_flags & ~XFS_LOCK_MASK) == 0); + ASSERT((lock_flags & ~(XFS_LOCK_MASK | XFS_LOCK_DEP_MASK)) == 0); + + if (lock_flags & XFS_IOLOCK_EXCL) + mrupdate_nested(&ip->i_iolock, XFS_IOLOCK_DEP(lock_flags)); + else if (lock_flags & XFS_IOLOCK_SHARED) + mraccess_nested(&ip->i_iolock, XFS_IOLOCK_DEP(lock_flags)); + + if (lock_flags & XFS_ILOCK_EXCL) + mrupdate_nested(&ip->i_lock, XFS_ILOCK_DEP(lock_flags)); + else if (lock_flags & XFS_ILOCK_SHARED) + mraccess_nested(&ip->i_lock, XFS_ILOCK_DEP(lock_flags)); - if (lock_flags & XFS_IOLOCK_EXCL) { - mrupdate(&ip->i_iolock); - } else if (lock_flags & XFS_IOLOCK_SHARED) { - mraccess(&ip->i_iolock); - } - if (lock_flags & XFS_ILOCK_EXCL) { - mrupdate(&ip->i_lock); - } else if (lock_flags & XFS_ILOCK_SHARED) { - mraccess(&ip->i_lock); - } xfs_ilock_trace(ip, 1, lock_flags, (inst_t *)__return_address); } @@ -887,15 +662,12 @@ xfs_ilock(xfs_inode_t *ip, * lock_flags -- this parameter indicates the inode's locks to be * to be locked. See the comment for xfs_ilock() for a list * of valid values. - * */ int -xfs_ilock_nowait(xfs_inode_t *ip, - uint lock_flags) +xfs_ilock_nowait( + xfs_inode_t *ip, + uint lock_flags) { - int iolocked; - int ilocked; - /* * You can't set both SHARED and EXCL for the same lock, * and only XFS_IOLOCK_SHARED, XFS_IOLOCK_EXCL, XFS_ILOCK_SHARED, @@ -905,39 +677,32 @@ xfs_ilock_nowait(xfs_inode_t *ip, (XFS_IOLOCK_SHARED | XFS_IOLOCK_EXCL)); ASSERT((lock_flags & (XFS_ILOCK_SHARED | XFS_ILOCK_EXCL)) != (XFS_ILOCK_SHARED | XFS_ILOCK_EXCL)); - ASSERT((lock_flags & ~XFS_LOCK_MASK) == 0); + ASSERT((lock_flags & ~(XFS_LOCK_MASK | XFS_LOCK_DEP_MASK)) == 0); - iolocked = 0; if (lock_flags & XFS_IOLOCK_EXCL) { - iolocked = mrtryupdate(&ip->i_iolock); - if (!iolocked) { - return 0; - } + if (!mrtryupdate(&ip->i_iolock)) + goto out; } else if (lock_flags & XFS_IOLOCK_SHARED) { - iolocked = mrtryaccess(&ip->i_iolock); - if (!iolocked) { - return 0; - } + if (!mrtryaccess(&ip->i_iolock)) + goto out; } if (lock_flags & XFS_ILOCK_EXCL) { - ilocked = mrtryupdate(&ip->i_lock); - if (!ilocked) { - if (iolocked) { - mrunlock(&ip->i_iolock); - } - return 0; - } + if (!mrtryupdate(&ip->i_lock)) + goto out_undo_iolock; } else if (lock_flags & XFS_ILOCK_SHARED) { - ilocked = mrtryaccess(&ip->i_lock); - if (!ilocked) { - if (iolocked) { - mrunlock(&ip->i_iolock); - } - return 0; - } + if (!mrtryaccess(&ip->i_lock)) + goto out_undo_iolock; } xfs_ilock_trace(ip, 2, lock_flags, (inst_t *)__return_address); return 1; + + out_undo_iolock: + if (lock_flags & XFS_IOLOCK_EXCL) + mrunlock_excl(&ip->i_iolock); + else if (lock_flags & XFS_IOLOCK_SHARED) + mrunlock_shared(&ip->i_iolock); + out: + return 0; } /* @@ -953,8 +718,9 @@ xfs_ilock_nowait(xfs_inode_t *ip, * */ void -xfs_iunlock(xfs_inode_t *ip, - uint lock_flags) +xfs_iunlock( + xfs_inode_t *ip, + uint lock_flags) { /* * You can't set both SHARED and EXCL for the same lock, @@ -965,34 +731,29 @@ xfs_iunlock(xfs_inode_t *ip, (XFS_IOLOCK_SHARED | XFS_IOLOCK_EXCL)); ASSERT((lock_flags & (XFS_ILOCK_SHARED | XFS_ILOCK_EXCL)) != (XFS_ILOCK_SHARED | XFS_ILOCK_EXCL)); - ASSERT((lock_flags & ~(XFS_LOCK_MASK | XFS_IUNLOCK_NONOTIFY)) == 0); + ASSERT((lock_flags & ~(XFS_LOCK_MASK | XFS_IUNLOCK_NONOTIFY | + XFS_LOCK_DEP_MASK)) == 0); ASSERT(lock_flags != 0); - if (lock_flags & (XFS_IOLOCK_SHARED | XFS_IOLOCK_EXCL)) { - ASSERT(!(lock_flags & XFS_IOLOCK_SHARED) || - (ismrlocked(&ip->i_iolock, MR_ACCESS))); - ASSERT(!(lock_flags & XFS_IOLOCK_EXCL) || - (ismrlocked(&ip->i_iolock, MR_UPDATE))); - mrunlock(&ip->i_iolock); - } + if (lock_flags & XFS_IOLOCK_EXCL) + mrunlock_excl(&ip->i_iolock); + else if (lock_flags & XFS_IOLOCK_SHARED) + mrunlock_shared(&ip->i_iolock); - if (lock_flags & (XFS_ILOCK_SHARED | XFS_ILOCK_EXCL)) { - ASSERT(!(lock_flags & XFS_ILOCK_SHARED) || - (ismrlocked(&ip->i_lock, MR_ACCESS))); - ASSERT(!(lock_flags & XFS_ILOCK_EXCL) || - (ismrlocked(&ip->i_lock, MR_UPDATE))); - mrunlock(&ip->i_lock); + if (lock_flags & XFS_ILOCK_EXCL) + mrunlock_excl(&ip->i_lock); + else if (lock_flags & XFS_ILOCK_SHARED) + mrunlock_shared(&ip->i_lock); + if ((lock_flags & (XFS_ILOCK_SHARED | XFS_ILOCK_EXCL)) && + !(lock_flags & XFS_IUNLOCK_NONOTIFY) && ip->i_itemp) { /* * Let the AIL know that this item has been unlocked in case * it is in the AIL and anyone is waiting on it. Don't do * this if the caller has asked us not to. */ - if (!(lock_flags & XFS_IUNLOCK_NONOTIFY) && - ip->i_itemp != NULL) { - xfs_trans_unlocked_item(ip->i_mount, - (xfs_log_item_t*)(ip->i_itemp)); - } + xfs_trans_unlocked_item(ip->i_itemp->ili_item.li_ailp, + (xfs_log_item_t*)(ip->i_itemp)); } xfs_ilock_trace(ip, 3, lock_flags, (inst_t *)__return_address); } @@ -1002,42 +763,93 @@ xfs_iunlock(xfs_inode_t *ip, * if it is being demoted. */ void -xfs_ilock_demote(xfs_inode_t *ip, - uint lock_flags) +xfs_ilock_demote( + xfs_inode_t *ip, + uint lock_flags) { ASSERT(lock_flags & (XFS_IOLOCK_EXCL|XFS_ILOCK_EXCL)); ASSERT((lock_flags & ~(XFS_IOLOCK_EXCL|XFS_ILOCK_EXCL)) == 0); - if (lock_flags & XFS_ILOCK_EXCL) { - ASSERT(ismrlocked(&ip->i_lock, MR_UPDATE)); + if (lock_flags & XFS_ILOCK_EXCL) mrdemote(&ip->i_lock); - } - if (lock_flags & XFS_IOLOCK_EXCL) { - ASSERT(ismrlocked(&ip->i_iolock, MR_UPDATE)); + if (lock_flags & XFS_IOLOCK_EXCL) mrdemote(&ip->i_iolock); +} + +#ifdef DEBUG +/* + * Debug-only routine, without additional rw_semaphore APIs, we can + * now only answer requests regarding whether we hold the lock for write + * (reader state is outside our visibility, we only track writer state). + * + * Note: this means !xfs_isilocked would give false positives, so don't do that. + */ +int +xfs_isilocked( + xfs_inode_t *ip, + uint lock_flags) +{ + if ((lock_flags & (XFS_ILOCK_EXCL|XFS_ILOCK_SHARED)) == + XFS_ILOCK_EXCL) { + if (!ip->i_lock.mr_writer) + return 0; } + + if ((lock_flags & (XFS_IOLOCK_EXCL|XFS_IOLOCK_SHARED)) == + XFS_IOLOCK_EXCL) { + if (!ip->i_iolock.mr_writer) + return 0; + } + + return 1; } +#endif + +#ifdef XFS_INODE_TRACE + +#define KTRACE_ENTER(ip, vk, s, line, ra) \ + ktrace_enter((ip)->i_trace, \ +/* 0 */ (void *)(__psint_t)(vk), \ +/* 1 */ (void *)(s), \ +/* 2 */ (void *)(__psint_t) line, \ +/* 3 */ (void *)(__psint_t)atomic_read(&VFS_I(ip)->i_count), \ +/* 4 */ (void *)(ra), \ +/* 5 */ NULL, \ +/* 6 */ (void *)(__psint_t)current_cpu(), \ +/* 7 */ (void *)(__psint_t)current_pid(), \ +/* 8 */ (void *)__return_address, \ +/* 9 */ NULL, NULL, NULL, NULL, NULL, NULL, NULL) /* - * The following three routines simply manage the i_flock - * semaphore embedded in the inode. This semaphore synchronizes - * processes attempting to flush the in-core inode back to disk. + * Vnode tracing code. */ void -xfs_iflock(xfs_inode_t *ip) +_xfs_itrace_entry(xfs_inode_t *ip, const char *func, inst_t *ra) { - psema(&(ip->i_flock), PINOD|PLTWAIT); + KTRACE_ENTER(ip, INODE_KTRACE_ENTRY, func, 0, ra); } -int -xfs_iflock_nowait(xfs_inode_t *ip) +void +_xfs_itrace_exit(xfs_inode_t *ip, const char *func, inst_t *ra) +{ + KTRACE_ENTER(ip, INODE_KTRACE_EXIT, func, 0, ra); +} + +void +xfs_itrace_hold(xfs_inode_t *ip, char *file, int line, inst_t *ra) +{ + KTRACE_ENTER(ip, INODE_KTRACE_HOLD, file, line, ra); +} + +void +_xfs_itrace_ref(xfs_inode_t *ip, char *file, int line, inst_t *ra) { - return (cpsema(&(ip->i_flock))); + KTRACE_ENTER(ip, INODE_KTRACE_REF, file, line, ra); } void -xfs_ifunlock(xfs_inode_t *ip) +xfs_itrace_rele(xfs_inode_t *ip, char *file, int line, inst_t *ra) { - ASSERT(valusema(&(ip->i_flock)) <= 0); - vsema(&(ip->i_flock)); + KTRACE_ENTER(ip, INODE_KTRACE_RELE, file, line, ra); } +#endif /* XFS_INODE_TRACE */