X-Git-Url: https://git.camperquake.de/gitweb.cgi?a=blobdiff_plain;f=module%2Fzfs%2Fdbuf.c;h=205abaada491ecdc1b9ef0e77d057573a19967ff;hb=5853fe790d1df58c5dd85ea52c5e165b6d43013c;hp=ccab0bce4009d2837d917117a6dc6fd0f14fffd2;hpb=fc5bb51f08a6c91ff9ad3559d0266eeeab0b1f61;p=zfs.git diff --git a/module/zfs/dbuf.c b/module/zfs/dbuf.c index ccab0bc..205abaa 100644 --- a/module/zfs/dbuf.c +++ b/module/zfs/dbuf.c @@ -20,9 +20,12 @@ */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. + * Copyright 2011 Nexenta Systems, Inc. All rights reserved. + * Copyright (c) 2012 by Delphix. All rights reserved. */ #include +#include #include #include #include @@ -258,7 +261,7 @@ dbuf_is_metadata(dmu_buf_impl_t *db) boolean_t is_metadata; DB_DNODE_ENTER(db); - is_metadata = dmu_ot[DB_DNODE(db)->dn_type].ot_metadata; + is_metadata = DMU_OT_IS_METADATA(DB_DNODE(db)->dn_type); DB_DNODE_EXIT(db); return (is_metadata); @@ -293,7 +296,13 @@ dbuf_init(void) retry: h->hash_table_mask = hsize - 1; +#if defined(_KERNEL) && defined(HAVE_SPL) + /* Large allocations which do not require contiguous pages + * should be using vmem_alloc() in the linux kernel */ + h->hash_table = vmem_zalloc(hsize * sizeof (void *), KM_PUSHPAGE); +#else h->hash_table = kmem_zalloc(hsize * sizeof (void *), KM_NOSLEEP); +#endif if (h->hash_table == NULL) { /* XXX - we should really return an error instead of assert */ ASSERT(hsize > (1ULL << 10)); @@ -317,7 +326,13 @@ dbuf_fini(void) for (i = 0; i < DBUF_MUTEXES; i++) mutex_destroy(&h->hash_mutexes[i]); +#if defined(_KERNEL) && defined(HAVE_SPL) + /* Large allocations which do not require contiguous pages + * should be using vmem_free() in the linux kernel */ + vmem_free(h->hash_table, (h->hash_table_mask + 1) * sizeof (void *)); +#else kmem_free(h->hash_table, (h->hash_table_mask + 1) * sizeof (void *)); +#endif kmem_cache_destroy(dbuf_cache); } @@ -1082,7 +1097,7 @@ dbuf_dirty(dmu_buf_impl_t *db, dmu_tx_t *tx) dn->dn_dirtyctx = (dmu_tx_is_syncing(tx) ? DN_DIRTY_SYNC : DN_DIRTY_OPEN); ASSERT(dn->dn_dirtyctx_firstset == NULL); - dn->dn_dirtyctx_firstset = kmem_alloc(1, KM_SLEEP); + dn->dn_dirtyctx_firstset = kmem_alloc(1, KM_PUSHPAGE); } mutex_exit(&dn->dn_mtx); @@ -1159,7 +1174,7 @@ dbuf_dirty(dmu_buf_impl_t *db, dmu_tx_t *tx) * to make a copy of it so that the changes we make in this * transaction group won't leak out when we sync the older txg. */ - dr = kmem_zalloc(sizeof (dbuf_dirty_record_t), KM_SLEEP); + dr = kmem_zalloc(sizeof (dbuf_dirty_record_t), KM_PUSHPAGE); list_link_init(&dr->dr_dirty_node); if (db->db_level == 0) { void *data_old = db->db_buf; @@ -1334,13 +1349,17 @@ dbuf_undirty(dmu_buf_impl_t *db, dmu_tx_t *tx) * it, since one of the current holders may be in the * middle of an update. Note that users of dbuf_undirty() * should not place a hold on the dbuf before the call. + * Also note: we can get here with a spill block, so + * test for that similar to how dbuf_dirty does. */ if (refcount_count(&db->db_holds) > db->db_dirtycnt) { mutex_exit(&db->db_mtx); /* Make sure we don't toss this buffer at sync phase */ - mutex_enter(&dn->dn_mtx); - dnode_clear_range(dn, db->db_blkid, 1, tx); - mutex_exit(&dn->dn_mtx); + if (db->db_blkid != DMU_SPILL_BLKID) { + mutex_enter(&dn->dn_mtx); + dnode_clear_range(dn, db->db_blkid, 1, tx); + mutex_exit(&dn->dn_mtx); + } DB_DNODE_EXIT(db); return (0); } @@ -1353,11 +1372,18 @@ dbuf_undirty(dmu_buf_impl_t *db, dmu_tx_t *tx) *drp = dr->dr_next; + /* + * Note that there are three places in dbuf_dirty() + * where this dirty record may be put on a list. + * Make sure to do a list_remove corresponding to + * every one of those list_insert calls. + */ if (dr->dr_parent) { mutex_enter(&dr->dr_parent->dt.di.dr_mtx); list_remove(&dr->dr_parent->dt.di.dr_children, dr); mutex_exit(&dr->dr_parent->dt.di.dr_mtx); - } else if (db->db_level+1 == dn->dn_nlevels) { + } else if (db->db_blkid == DMU_SPILL_BLKID || + db->db_level+1 == dn->dn_nlevels) { ASSERT(db->db_blkptr == NULL || db->db_parent == dn->dn_dbuf); mutex_enter(&dn->dn_mtx); list_remove(&dn->dn_dirty_records[txg & TXG_MASK], dr); @@ -1607,7 +1633,8 @@ dbuf_clear(dmu_buf_impl_t *db) dbuf_rele(parent, db); } -static int +__attribute__((always_inline)) +static inline int dbuf_findbp(dnode_t *dn, int level, uint64_t blkid, int fail_sparse, dmu_buf_impl_t **parentp, blkptr_t **bpp, struct dbuf_hold_impl_data *dh) { @@ -1693,7 +1720,7 @@ dbuf_create(dnode_t *dn, uint8_t level, uint64_t blkid, ASSERT(RW_LOCK_HELD(&dn->dn_struct_rwlock)); ASSERT(dn->dn_type != DMU_OT_NONE); - db = kmem_cache_alloc(dbuf_cache, KM_SLEEP); + db = kmem_cache_alloc(dbuf_cache, KM_PUSHPAGE); db->db_objset = os; db->db.db_object = dn->dn_object; @@ -1993,7 +2020,7 @@ dbuf_hold_impl(dnode_t *dn, uint8_t level, uint64_t blkid, int fail_sparse, int error; dh = kmem_zalloc(sizeof(struct dbuf_hold_impl_data) * - DBUF_HOLD_IMPL_MAX_DEPTH, KM_SLEEP); + DBUF_HOLD_IMPL_MAX_DEPTH, KM_PUSHPAGE); __dbuf_hold_impl_init(dh, dn, level, blkid, fail_sparse, tag, dbp, 0); error = __dbuf_hold_impl(dh); @@ -2162,7 +2189,24 @@ dbuf_rele_and_unlock(dmu_buf_impl_t *db, void *tag) dbuf_evict(db); } else { VERIFY(arc_buf_remove_ref(db->db_buf, db) == 0); - if (!DBUF_IS_CACHEABLE(db)) + + /* + * A dbuf will be eligible for eviction if either the + * 'primarycache' property is set or a duplicate + * copy of this buffer is already cached in the arc. + * + * In the case of the 'primarycache' a buffer + * is considered for eviction if it matches the + * criteria set in the property. + * + * To decide if our buffer is considered a + * duplicate, we must call into the arc to determine + * if multiple buffers are referencing the same + * block on-disk. If so, then we simply evict + * ourselves. + */ + if (!DBUF_IS_CACHEABLE(db) || + arc_buf_eviction_needed(db->db_buf)) dbuf_clear(db); else mutex_exit(&db->db_mtx); @@ -2290,7 +2334,11 @@ dbuf_check_blkptr(dnode_t *dn, dmu_buf_impl_t *db) } } -static void +/* dbuf_sync_indirect() is called recursively from dbuf_sync_list() so it + * is critical the we not allow the compiler to inline this function in to + * dbuf_sync_list() thereby drastically bloating the stack usage. + */ +noinline static void dbuf_sync_indirect(dbuf_dirty_record_t *dr, dmu_tx_t *tx) { dmu_buf_impl_t *db = dr->dr_dbuf; @@ -2333,7 +2381,11 @@ dbuf_sync_indirect(dbuf_dirty_record_t *dr, dmu_tx_t *tx) zio_nowait(zio); } -static void +/* dbuf_sync_leaf() is called recursively from dbuf_sync_list() so it is + * critical the we not allow the compiler to inline this function in to + * dbuf_sync_list() thereby drastically bloating the stack usage. + */ +noinline static void dbuf_sync_leaf(dbuf_dirty_record_t *dr, dmu_tx_t *tx) { arc_buf_t **datap = &dr->dt.dl.dr_data; @@ -2795,3 +2847,41 @@ dbuf_write(dbuf_dirty_record_t *dr, arc_buf_t *data, dmu_tx_t *tx) ZIO_PRIORITY_ASYNC_WRITE, ZIO_FLAG_MUSTSUCCEED, &zb); } } + +#if defined(_KERNEL) && defined(HAVE_SPL) +EXPORT_SYMBOL(dbuf_find); +EXPORT_SYMBOL(dbuf_is_metadata); +EXPORT_SYMBOL(dbuf_evict); +EXPORT_SYMBOL(dbuf_loan_arcbuf); +EXPORT_SYMBOL(dbuf_whichblock); +EXPORT_SYMBOL(dbuf_read); +EXPORT_SYMBOL(dbuf_unoverride); +EXPORT_SYMBOL(dbuf_free_range); +EXPORT_SYMBOL(dbuf_new_size); +EXPORT_SYMBOL(dbuf_release_bp); +EXPORT_SYMBOL(dbuf_dirty); +EXPORT_SYMBOL(dmu_buf_will_dirty); +EXPORT_SYMBOL(dmu_buf_will_not_fill); +EXPORT_SYMBOL(dmu_buf_will_fill); +EXPORT_SYMBOL(dmu_buf_fill_done); +EXPORT_SYMBOL(dmu_buf_rele); +EXPORT_SYMBOL(dbuf_assign_arcbuf); +EXPORT_SYMBOL(dbuf_clear); +EXPORT_SYMBOL(dbuf_prefetch); +EXPORT_SYMBOL(dbuf_hold_impl); +EXPORT_SYMBOL(dbuf_hold); +EXPORT_SYMBOL(dbuf_hold_level); +EXPORT_SYMBOL(dbuf_create_bonus); +EXPORT_SYMBOL(dbuf_spill_set_blksz); +EXPORT_SYMBOL(dbuf_rm_spill); +EXPORT_SYMBOL(dbuf_add_ref); +EXPORT_SYMBOL(dbuf_rele); +EXPORT_SYMBOL(dbuf_rele_and_unlock); +EXPORT_SYMBOL(dbuf_refcount); +EXPORT_SYMBOL(dbuf_sync_list); +EXPORT_SYMBOL(dmu_buf_set_user); +EXPORT_SYMBOL(dmu_buf_set_user_ie); +EXPORT_SYMBOL(dmu_buf_update_user); +EXPORT_SYMBOL(dmu_buf_get_user); +EXPORT_SYMBOL(dmu_buf_freeable); +#endif