X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Flov%2Flov_object.c;h=07ed284e66576a61108cf18ebcbfcd1ce80911c3;hp=4c15b9fbffa25753e8949ef344e8679b41f85335;hb=d821adb626e050206f1b4605f1159cf65ad22af5;hpb=65701b4a30efdb695776bcf690a2b3cabc928da1 diff --git a/lustre/lov/lov_object.c b/lustre/lov/lov_object.c index 4c15b9f..07ed284 100644 --- a/lustre/lov/lov_object.c +++ b/lustre/lov/lov_object.c @@ -1,6 +1,4 @@ -/* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*- - * vim:expandtab:shiftwidth=8:tabstop=8: - * +/* * GPL HEADER START * * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. @@ -29,7 +27,7 @@ * Copyright (c) 2008, 2010, Oracle and/or its affiliates. All rights reserved. * Use is subject to license terms. * - * Copyright (c) 2011, Whamcloud, Inc. + * Copyright (c) 2011, 2013, Intel Corporation. */ /* * This file is part of Lustre, http://www.lustre.org/ @@ -38,11 +36,13 @@ * Implementation of cl_object for LOV layer. * * Author: Nikita Danilov + * Author: Jinshan Xiong */ #define DEBUG_SUBSYSTEM S_LOV #include "lov_cl_internal.h" +#include /** \addtogroup lov * @{ @@ -59,7 +59,7 @@ struct lov_layout_operations { struct lov_object *lov, const struct cl_object_conf *conf, union lov_layout_state *state); - void (*llo_delete)(const struct lu_env *env, struct lov_object *lov, + int (*llo_delete)(const struct lu_env *env, struct lov_object *lov, union lov_layout_state *state); void (*llo_fini)(const struct lu_env *env, struct lov_object *lov, union lov_layout_state *state); @@ -67,10 +67,8 @@ struct lov_layout_operations { union lov_layout_state *state); int (*llo_print)(const struct lu_env *env, void *cookie, lu_printer_t p, const struct lu_object *o); - struct cl_page *(*llo_page_init)(const struct lu_env *env, - struct cl_object *obj, - struct cl_page *page, - cfs_page_t *vmpage); + int (*llo_page_init)(const struct lu_env *env, struct cl_object *obj, + struct cl_page *page, cfs_page_t *vmpage); int (*llo_lock_init)(const struct lu_env *env, struct cl_object *obj, struct cl_lock *lock, const struct cl_io *io); @@ -80,6 +78,8 @@ struct lov_layout_operations { struct cl_attr *attr); }; +static int lov_layout_wait(const struct lu_env *env, struct lov_object *lov); + /***************************************************************************** * * Lov object layout operations. @@ -107,7 +107,6 @@ static void lov_install_raid0(const struct lu_env *env, struct lov_object *lov, union lov_layout_state *state) { - lov->u = *state; } static struct cl_object *lov_sub_find(const struct lu_env *env, @@ -133,17 +132,27 @@ static int lov_init_sub(const struct lu_env *env, struct lov_object *lov, struct lov_oinfo *oinfo; int result; + if (OBD_FAIL_CHECK(OBD_FAIL_LOV_INIT)) { + /* For sanity:test_206. + * Do not leave the object in cache to avoid accessing + * freed memory. This is because osc_object is referring to + * lov_oinfo of lsm_stripe_data which will be freed due to + * this failure. */ + cl_object_kill(env, stripe); + cl_object_put(env, stripe); + return -EIO; + } + hdr = cl_object_header(lov2cl(lov)); subhdr = cl_object_header(stripe); parent = subhdr->coh_parent; - oinfo = r0->lo_lsm->lsm_oinfo[idx]; - CDEBUG(D_INODE, DFID"@%p[%d] -> "DFID"@%p: id: "LPU64" seq: "LPU64 - " idx: %d gen: %d\n", - PFID(&subhdr->coh_lu.loh_fid), subhdr, idx, - PFID(&hdr->coh_lu.loh_fid), hdr, - oinfo->loi_id, oinfo->loi_seq, - oinfo->loi_ost_idx, oinfo->loi_ost_gen); + oinfo = lov->lo_lsm->lsm_oinfo[idx]; + CDEBUG(D_INODE, DFID"@%p[%d] -> "DFID"@%p: ostid: "DOSTID + " idx: %d gen: %d\n", + PFID(&subhdr->coh_lu.loh_fid), subhdr, idx, + PFID(&hdr->coh_lu.loh_fid), hdr, POSTID(&oinfo->loi_oi), + oinfo->loi_ost_idx, oinfo->loi_ost_gen); if (parent == NULL) { subhdr->coh_parent = hdr; @@ -154,13 +163,28 @@ static int lov_init_sub(const struct lu_env *env, struct lov_object *lov, r0->lo_sub[idx]->lso_index = idx; result = 0; } else { - CERROR("Stripe is already owned by other file (%d).\n", idx); - LU_OBJECT_DEBUG(D_ERROR, env, &stripe->co_lu, "\n"); - LU_OBJECT_DEBUG(D_ERROR, env, lu_object_top(&parent->coh_lu), - "old\n"); - LU_OBJECT_HEADER(D_ERROR, env, lov2lu(lov), "new\n"); - cl_object_put(env, stripe); - result = -EIO; + struct lu_object *old_obj; + struct lov_object *old_lov; + unsigned int mask = D_INODE; + + old_obj = lu_object_locate(&parent->coh_lu, &lov_device_type); + LASSERT(old_obj != NULL); + old_lov = cl2lov(lu2cl(old_obj)); + if (old_lov->lo_layout_invalid) { + /* the object's layout has already changed but isn't + * refreshed */ + lu_object_unhash(env, &stripe->co_lu); + result = -EAGAIN; + } else { + mask = D_ERROR; + result = -EIO; + } + + LU_OBJECT_DEBUG(mask, env, &stripe->co_lu, + "stripe %d is already owned.\n", idx); + LU_OBJECT_DEBUG(mask, env, old_obj, "owned.\n"); + LU_OBJECT_HEADER(mask, env, lov2lu(lov), "try to own.\n"); + cl_object_put(env, stripe); } return result; } @@ -181,15 +205,23 @@ static int lov_init_raid0(const struct lu_env *env, struct lov_layout_raid0 *r0 = &state->raid0; ENTRY; - r0->lo_nr = conf->u.coc_md->lsm->lsm_stripe_count; - r0->lo_lsm = conf->u.coc_md->lsm; + + if (lsm->lsm_magic != LOV_MAGIC_V1 && lsm->lsm_magic != LOV_MAGIC_V3) { + dump_lsm(D_ERROR, lsm); + LASSERTF(0, "magic mismatch, expected %d/%d, actual %d.\n", + LOV_MAGIC_V1, LOV_MAGIC_V3, lsm->lsm_magic); + } + + LASSERT(lov->lo_lsm == NULL); + lov->lo_lsm = lsm_addref(lsm); + r0->lo_nr = lsm->lsm_stripe_count; LASSERT(r0->lo_nr <= lov_targets_nr(dev)); OBD_ALLOC_LARGE(r0->lo_sub, r0->lo_nr * sizeof r0->lo_sub[0]); if (r0->lo_sub != NULL) { result = 0; subconf->coc_inode = conf->coc_inode; - cfs_spin_lock_init(&r0->lo_sub_lock); + spin_lock_init(&r0->lo_sub_lock); /* * Create stripe cl_objects. */ @@ -198,26 +230,43 @@ static int lov_init_raid0(const struct lu_env *env, struct lov_oinfo *oinfo = lsm->lsm_oinfo[i]; int ost_idx = oinfo->loi_ost_idx; - fid_ostid_unpack(ofid, &oinfo->loi_oi, - oinfo->loi_ost_idx); + result = ostid_to_fid(ofid, &oinfo->loi_oi, + oinfo->loi_ost_idx); + if (result != 0) + GOTO(out, result); + subdev = lovsub2cl_dev(dev->ld_target[ost_idx]); subconf->u.coc_oinfo = oinfo; LASSERTF(subdev != NULL, "not init ost %d\n", ost_idx); + /* In the function below, .hs_keycmp resolves to + * lu_obj_hop_keycmp() */ + /* coverity[overrun-buffer-val] */ stripe = lov_sub_find(env, subdev, ofid, subconf); - if (!IS_ERR(stripe)) + if (!IS_ERR(stripe)) { result = lov_init_sub(env, lov, stripe, r0, i); - else + if (result == -EAGAIN) { /* try again */ + --i; + result = 0; + } + } else { result = PTR_ERR(stripe); + } } } else result = -ENOMEM; +out: RETURN(result); } -static void lov_delete_empty(const struct lu_env *env, struct lov_object *lov, - union lov_layout_state *state) +static int lov_delete_empty(const struct lu_env *env, struct lov_object *lov, + union lov_layout_state *state) { - LASSERT(lov->lo_type == LLT_EMPTY); + LASSERT(lov->lo_type == LLT_EMPTY); + + lov_layout_wait(env, lov); + + cl_object_prune(env, &lov->lo_cl); + return 0; } static void lov_subobject_kill(const struct lu_env *env, struct lov_object *lov, @@ -252,12 +301,12 @@ static void lov_subobject_kill(const struct lu_env *env, struct lov_object *lov, /* this wait-queue is signaled at the end of * lu_object_free(). */ cfs_set_current_state(CFS_TASK_UNINT); - cfs_spin_lock(&r0->lo_sub_lock); - if (r0->lo_sub[idx] == los) { - cfs_spin_unlock(&r0->lo_sub_lock); - cfs_waitq_wait(waiter, CFS_TASK_UNINT); - } else { - cfs_spin_unlock(&r0->lo_sub_lock); + spin_lock(&r0->lo_sub_lock); + if (r0->lo_sub[idx] == los) { + spin_unlock(&r0->lo_sub_lock); + cfs_waitq_wait(waiter, CFS_TASK_UNINT); + } else { + spin_unlock(&r0->lo_sub_lock); cfs_set_current_state(CFS_TASK_RUNNING); break; } @@ -267,26 +316,34 @@ static void lov_subobject_kill(const struct lu_env *env, struct lov_object *lov, LASSERT(r0->lo_sub[idx] == NULL); } -static void lov_delete_raid0(const struct lu_env *env, struct lov_object *lov, - union lov_layout_state *state) +static int lov_delete_raid0(const struct lu_env *env, struct lov_object *lov, + union lov_layout_state *state) { - struct lov_layout_raid0 *r0 = &state->raid0; - int i; + struct lov_layout_raid0 *r0 = &state->raid0; + struct lov_stripe_md *lsm = lov->lo_lsm; + int i; - ENTRY; + ENTRY; + + dump_lsm(D_INODE, lsm); + + lov_layout_wait(env, lov); if (r0->lo_sub != NULL) { for (i = 0; i < r0->lo_nr; ++i) { struct lovsub_object *los = r0->lo_sub[i]; - if (los != NULL) + if (los != NULL) { + cl_locks_prune(env, &los->lso_cl, 1); /* * If top-level object is to be evicted from * the cache, so are its sub-objects. */ lov_subobject_kill(env, lov, los, i); + } } } - EXIT; + cl_object_prune(env, &lov->lo_cl); + RETURN(0); } static void lov_fini_empty(const struct lu_env *env, struct lov_object *lov, @@ -298,20 +355,24 @@ static void lov_fini_empty(const struct lu_env *env, struct lov_object *lov, static void lov_fini_raid0(const struct lu_env *env, struct lov_object *lov, union lov_layout_state *state) { - struct lov_layout_raid0 *r0 = &state->raid0; + struct lov_layout_raid0 *r0 = &state->raid0; + ENTRY; - ENTRY; - if (r0->lo_sub != NULL) { - OBD_FREE_LARGE(r0->lo_sub, r0->lo_nr * sizeof r0->lo_sub[0]); - r0->lo_sub = NULL; - } - EXIT; + if (r0->lo_sub != NULL) { + OBD_FREE_LARGE(r0->lo_sub, r0->lo_nr * sizeof r0->lo_sub[0]); + r0->lo_sub = NULL; + } + + dump_lsm(D_INODE, lov->lo_lsm); + lov_free_memmd(&lov->lo_lsm); + + EXIT; } static int lov_print_empty(const struct lu_env *env, void *cookie, lu_printer_t p, const struct lu_object *o) { - (*p)(env, cookie, "empty\n"); + (*p)(env, cookie, "empty %d\n", lu2lov(o)->lo_layout_invalid); return 0; } @@ -320,9 +381,13 @@ static int lov_print_raid0(const struct lu_env *env, void *cookie, { struct lov_object *lov = lu2lov(o); struct lov_layout_raid0 *r0 = lov_r0(lov); + struct lov_stripe_md *lsm = lov->lo_lsm; int i; - (*p)(env, cookie, "stripes: %d:\n", r0->lo_nr); + (*p)(env, cookie, "stripes: %d, %svalid, lsm{%p 0x%08X %d %u %u}: \n", + r0->lo_nr, lov->lo_layout_invalid ? "in" : "", lsm, + lsm->lsm_magic, cfs_atomic_read(&lsm->lsm_refc), + lsm->lsm_stripe_count, lsm->lsm_layout_gen); for (i = 0; i < r0->lo_nr; ++i) { struct lu_object *sub; @@ -352,46 +417,68 @@ static int lov_attr_get_empty(const struct lu_env *env, struct cl_object *obj, static int lov_attr_get_raid0(const struct lu_env *env, struct cl_object *obj, struct cl_attr *attr) { - struct lov_object *lov = cl2lov(obj); - struct lov_layout_raid0 *r0 = lov_r0(lov); - struct lov_stripe_md *lsm = lov->u.raid0.lo_lsm; - struct ost_lvb *lvb = &lov_env_info(env)->lti_lvb; - __u64 kms; - int result = 0; + struct lov_object *lov = cl2lov(obj); + struct lov_layout_raid0 *r0 = lov_r0(lov); + struct cl_attr *lov_attr = &r0->lo_attr; + int result = 0; ENTRY; - if (!r0->lo_attr_valid) { - /* - * Fill LVB with attributes already initialized by the upper - * layer. - */ - cl_attr2lvb(lvb, attr); - kms = attr->cat_kms; - - /* - * XXX that should be replaced with a loop over sub-objects, - * doing cl_object_attr_get() on them. But for now, let's - * reuse old lov code. - */ - /* - * XXX take lsm spin-lock to keep lov_merge_lvb_kms() - * happy. It's not needed, because new code uses - * ->coh_attr_guard spin-lock to protect consistency of - * sub-object attributes. - */ - lov_stripe_lock(lsm); - result = lov_merge_lvb_kms(lsm, lvb, &kms); - lov_stripe_unlock(lsm); - if (result == 0) { - cl_lvb2attr(attr, lvb); - attr->cat_kms = kms; - r0->lo_attr_valid = 1; - r0->lo_attr = *attr; - } - } else - *attr = r0->lo_attr; - RETURN(result); + /* this is called w/o holding type guard mutex, so it must be inside + * an on going IO otherwise lsm may be replaced. + * LU-2117: it turns out there exists one exception. For mmaped files, + * the lock of those files may be requested in the other file's IO + * context, and this function is called in ccc_lock_state(), it will + * hit this assertion. + * Anyway, it's still okay to call attr_get w/o type guard as layout + * can't go if locks exist. */ + /* LASSERT(cfs_atomic_read(&lsm->lsm_refc) > 1); */ + + if (!r0->lo_attr_valid) { + struct lov_stripe_md *lsm = lov->lo_lsm; + struct ost_lvb *lvb = &lov_env_info(env)->lti_lvb; + __u64 kms = 0; + + memset(lvb, 0, sizeof(*lvb)); + /* XXX: timestamps can be negative by sanity:test_39m, + * how can it be? */ + lvb->lvb_atime = LLONG_MIN; + lvb->lvb_ctime = LLONG_MIN; + lvb->lvb_mtime = LLONG_MIN; + + /* + * XXX that should be replaced with a loop over sub-objects, + * doing cl_object_attr_get() on them. But for now, let's + * reuse old lov code. + */ + + /* + * XXX take lsm spin-lock to keep lov_merge_lvb_kms() + * happy. It's not needed, because new code uses + * ->coh_attr_guard spin-lock to protect consistency of + * sub-object attributes. + */ + lov_stripe_lock(lsm); + result = lov_merge_lvb_kms(lsm, lvb, &kms); + lov_stripe_unlock(lsm); + if (result == 0) { + cl_lvb2attr(lov_attr, lvb); + lov_attr->cat_kms = kms; + r0->lo_attr_valid = 1; + } + } + if (result == 0) { /* merge results */ + attr->cat_blocks = lov_attr->cat_blocks; + attr->cat_size = lov_attr->cat_size; + attr->cat_kms = lov_attr->cat_kms; + if (attr->cat_atime < lov_attr->cat_atime) + attr->cat_atime = lov_attr->cat_atime; + if (attr->cat_ctime < lov_attr->cat_ctime) + attr->cat_ctime = lov_attr->cat_ctime; + if (attr->cat_mtime < lov_attr->cat_mtime) + attr->cat_mtime = lov_attr->cat_mtime; + } + RETURN(result); } const static struct lov_layout_operations lov_dispatch[] = { @@ -402,7 +489,7 @@ const static struct lov_layout_operations lov_dispatch[] = { .llo_install = lov_install_empty, .llo_print = lov_print_empty, .llo_page_init = lov_page_init_empty, - .llo_lock_init = NULL, + .llo_lock_init = lov_lock_init_empty, .llo_io_init = lov_io_init_empty, .llo_getattr = lov_attr_get_empty }, @@ -433,18 +520,29 @@ const static struct lov_layout_operations lov_dispatch[] = { lov_dispatch[__llt].op(__VA_ARGS__); \ }) +static inline void lov_conf_freeze(struct lov_object *lov) +{ + if (lov->lo_owner != cfs_current()) + down_read(&lov->lo_type_guard); +} + +static inline void lov_conf_thaw(struct lov_object *lov) +{ + if (lov->lo_owner != cfs_current()) + up_read(&lov->lo_type_guard); +} + #define LOV_2DISPATCH_MAYLOCK(obj, op, lock, ...) \ ({ \ struct lov_object *__obj = (obj); \ int __lock = !!(lock); \ typeof(lov_dispatch[0].op(__VA_ARGS__)) __result; \ \ - __lock &= __obj->lo_owner != cfs_current(); \ if (__lock) \ - cfs_down_read(&__obj->lo_type_guard); \ + lov_conf_freeze(__obj); \ __result = LOV_2DISPATCH_NOLOCK(obj, op, __VA_ARGS__); \ if (__lock) \ - cfs_up_read(&__obj->lo_type_guard); \ + lov_conf_thaw(__obj); \ __result; \ }) @@ -459,58 +557,100 @@ do { \ struct lov_object *__obj = (obj); \ enum lov_layout_type __llt; \ \ - if (__obj->lo_owner != cfs_current()) \ - cfs_down_read(&__obj->lo_type_guard); \ + lov_conf_freeze(__obj); \ __llt = __obj->lo_type; \ LASSERT(0 <= __llt && __llt < ARRAY_SIZE(lov_dispatch)); \ lov_dispatch[__llt].op(__VA_ARGS__); \ - if (__obj->lo_owner != cfs_current()) \ - cfs_up_read(&__obj->lo_type_guard); \ + lov_conf_thaw(__obj); \ } while (0) -static int lov_layout_change(const struct lu_env *env, - struct lov_object *obj, enum lov_layout_type llt, - const struct cl_object_conf *conf) +static void lov_conf_lock(struct lov_object *lov) { - int result; - union lov_layout_state *state = &lov_env_info(env)->lti_state; - const struct lov_layout_operations *old_ops; - const struct lov_layout_operations *new_ops; + LASSERT(lov->lo_owner != cfs_current()); + down_write(&lov->lo_type_guard); + LASSERT(lov->lo_owner == NULL); + lov->lo_owner = cfs_current(); +} - LASSERT(0 <= obj->lo_type && obj->lo_type < ARRAY_SIZE(lov_dispatch)); - LASSERT(0 <= llt && llt < ARRAY_SIZE(lov_dispatch)); - ENTRY; +static void lov_conf_unlock(struct lov_object *lov) +{ + lov->lo_owner = NULL; + up_write(&lov->lo_type_guard); +} - old_ops = &lov_dispatch[obj->lo_type]; - new_ops = &lov_dispatch[llt]; - - result = new_ops->llo_init(env, lu2lov_dev(obj->lo_cl.co_lu.lo_dev), - obj, conf, state); - if (result == 0) { - struct cl_object_header *hdr = cl_object_header(&obj->lo_cl); - void *cookie; - struct lu_env *nested; - int refcheck; - - cookie = cl_env_reenter(); - nested = cl_env_get(&refcheck); - if (!IS_ERR(nested)) - cl_object_prune(nested, &obj->lo_cl); - else - result = PTR_ERR(nested); - cl_env_put(nested, &refcheck); - cl_env_reexit(cookie); - - old_ops->llo_fini(env, obj, &obj->u); - LASSERT(cfs_list_empty(&hdr->coh_locks)); - LASSERT(hdr->coh_tree.rnode == NULL); - LASSERT(hdr->coh_pages == 0); - - new_ops->llo_install(env, obj, state); - obj->lo_type = llt; - } else - new_ops->llo_fini(env, obj, state); - RETURN(result); +static int lov_layout_wait(const struct lu_env *env, struct lov_object *lov) +{ + struct l_wait_info lwi = { 0 }; + ENTRY; + + while (cfs_atomic_read(&lov->lo_active_ios) > 0) { + CDEBUG(D_INODE, "file:"DFID" wait for active IO, now: %d.\n", + PFID(lu_object_fid(lov2lu(lov))), + cfs_atomic_read(&lov->lo_active_ios)); + + l_wait_event(lov->lo_waitq, + cfs_atomic_read(&lov->lo_active_ios) == 0, &lwi); + } + RETURN(0); +} + +static int lov_layout_change(const struct lu_env *unused, + struct lov_object *lov, + const struct cl_object_conf *conf) +{ + int result; + enum lov_layout_type llt = LLT_EMPTY; + union lov_layout_state *state = &lov->u; + const struct lov_layout_operations *old_ops; + const struct lov_layout_operations *new_ops; + + struct cl_object_header *hdr = cl_object_header(&lov->lo_cl); + void *cookie; + struct lu_env *env; + int refcheck; + ENTRY; + + LASSERT(0 <= lov->lo_type && lov->lo_type < ARRAY_SIZE(lov_dispatch)); + + if (conf->u.coc_md != NULL && conf->u.coc_md->lsm != NULL) + llt = LLT_RAID0; /* only raid0 is supported. */ + LASSERT(0 <= llt && llt < ARRAY_SIZE(lov_dispatch)); + + cookie = cl_env_reenter(); + env = cl_env_get(&refcheck); + if (IS_ERR(env)) { + cl_env_reexit(cookie); + RETURN(PTR_ERR(env)); + } + + old_ops = &lov_dispatch[lov->lo_type]; + new_ops = &lov_dispatch[llt]; + + result = old_ops->llo_delete(env, lov, &lov->u); + if (result == 0) { + old_ops->llo_fini(env, lov, &lov->u); + + LASSERT(cfs_atomic_read(&lov->lo_active_ios) == 0); + LASSERT(hdr->coh_tree.rnode == NULL); + LASSERT(hdr->coh_pages == 0); + + lov->lo_type = LLT_EMPTY; + result = new_ops->llo_init(env, + lu2lov_dev(lov->lo_cl.co_lu.lo_dev), + lov, conf, state); + if (result == 0) { + new_ops->llo_install(env, lov, state); + lov->lo_type = llt; + } else { + new_ops->llo_delete(env, lov, state); + new_ops->llo_fini(env, lov, state); + /* this file becomes an EMPTY file. */ + } + } + + cl_env_put(env, &refcheck); + cl_env_reexit(cookie); + RETURN(result); } /***************************************************************************** @@ -525,12 +665,16 @@ int lov_object_init(const struct lu_env *env, struct lu_object *obj, struct lov_device *dev = lu2lov_dev(obj->lo_dev); struct lov_object *lov = lu2lov(obj); const struct cl_object_conf *cconf = lu2cl_conf(conf); - union lov_layout_state *set = &lov_env_info(env)->lti_state; + union lov_layout_state *set = &lov->u; const struct lov_layout_operations *ops; int result; ENTRY; - cfs_init_rwsem(&lov->lo_type_guard); + init_rwsem(&lov->lo_type_guard); + cfs_atomic_set(&lov->lo_active_ios, 0); + cfs_waitq_init(&lov->lo_waitq); + + cl_object_page_init(lu2cl(obj), sizeof(struct lov_page)); /* no locking is necessary, as object is being created */ lov->lo_type = cconf->u.coc_md->lsm != NULL ? LLT_RAID0 : LLT_EMPTY; @@ -538,32 +682,57 @@ int lov_object_init(const struct lu_env *env, struct lu_object *obj, result = ops->llo_init(env, dev, lov, cconf, set); if (result == 0) ops->llo_install(env, lov, set); - else - ops->llo_fini(env, lov, set); RETURN(result); } static int lov_conf_set(const struct lu_env *env, struct cl_object *obj, const struct cl_object_conf *conf) { - struct lov_object *lov = cl2lov(obj); - int result; - - ENTRY; - /* - * Currently only LLT_EMPTY -> LLT_RAID0 transition is supported. - */ - LASSERT(lov->lo_owner != cfs_current()); - cfs_down_write(&lov->lo_type_guard); - LASSERT(lov->lo_owner == NULL); - lov->lo_owner = cfs_current(); - if (lov->lo_type == LLT_EMPTY && conf->u.coc_md->lsm != NULL) - result = lov_layout_change(env, lov, LLT_RAID0, conf); - else - result = -EOPNOTSUPP; - lov->lo_owner = NULL; - cfs_up_write(&lov->lo_type_guard); - RETURN(result); + struct lov_stripe_md *lsm = NULL; + struct lov_object *lov = cl2lov(obj); + int result = 0; + ENTRY; + + lov_conf_lock(lov); + if (conf->coc_opc == OBJECT_CONF_INVALIDATE) { + lov->lo_layout_invalid = true; + GOTO(out, result = 0); + } + + if (conf->coc_opc == OBJECT_CONF_WAIT) { + if (lov->lo_layout_invalid && + cfs_atomic_read(&lov->lo_active_ios) > 0) { + lov_conf_unlock(lov); + result = lov_layout_wait(env, lov); + lov_conf_lock(lov); + } + GOTO(out, result); + } + + LASSERT(conf->coc_opc == OBJECT_CONF_SET); + + if (conf->u.coc_md != NULL) + lsm = conf->u.coc_md->lsm; + if ((lsm == NULL && lov->lo_lsm == NULL) || + (lsm != NULL && lov->lo_lsm != NULL && + lov->lo_lsm->lsm_layout_gen == lsm->lsm_layout_gen)) { + /* same version of layout */ + lov->lo_layout_invalid = false; + GOTO(out, result = 0); + } + + /* will change layout - check if there still exists active IO. */ + if (cfs_atomic_read(&lov->lo_active_ios) > 0) { + lov->lo_layout_invalid = true; + GOTO(out, result = -EBUSY); + } + + lov->lo_layout_invalid = lov_layout_change(env, lov, conf); + EXIT; + +out: + lov_conf_unlock(lov); + RETURN(result); } static void lov_object_delete(const struct lu_env *env, struct lu_object *obj) @@ -589,14 +758,14 @@ static void lov_object_free(const struct lu_env *env, struct lu_object *obj) static int lov_object_print(const struct lu_env *env, void *cookie, lu_printer_t p, const struct lu_object *o) { - return LOV_2DISPATCH(lu2lov(o), llo_print, env, cookie, p, o); + return LOV_2DISPATCH_NOLOCK(lu2lov(o), llo_print, env, cookie, p, o); } -struct cl_page *lov_page_init(const struct lu_env *env, struct cl_object *obj, - struct cl_page *page, cfs_page_t *vmpage) +int lov_page_init(const struct lu_env *env, struct cl_object *obj, + struct cl_page *page, cfs_page_t *vmpage) { - return LOV_2DISPATCH(cl2lov(obj), - llo_page_init, env, obj, page, vmpage); + return LOV_2DISPATCH_NOLOCK(cl2lov(obj), + llo_page_init, env, obj, page, vmpage); } /** @@ -604,23 +773,11 @@ struct cl_page *lov_page_init(const struct lu_env *env, struct cl_object *obj, * layer. Dispatches to the appropriate layout io initialization method. */ int lov_io_init(const struct lu_env *env, struct cl_object *obj, - struct cl_io *io) + struct cl_io *io) { - CL_IO_SLICE_CLEAN(lov_env_io(env), lis_cl); - /* - * Do not take lock in case of CIT_MISC io, because - * - * - if this is an io for a glimpse, then we don't care; - * - * - if this not a glimpse (writepage or lock cancellation), then - * layout change cannot happen because a page or a lock - * already exist; and - * - * - lock ordering (lock mutex nests within layout rw-semaphore) - * is obeyed in case of lock cancellation. - */ - return LOV_2DISPATCH_MAYLOCK(cl2lov(obj), llo_io_init, - io->ci_type != CIT_MISC, env, obj, io); + CL_IO_SLICE_CLEAN(lov_env_io(env), lis_cl); + return LOV_2DISPATCH_MAYLOCK(cl2lov(obj), llo_io_init, + !io->ci_ignore_layout, env, obj, io); } /** @@ -646,9 +803,11 @@ static int lov_attr_set(const struct lu_env *env, struct cl_object *obj, } int lov_lock_init(const struct lu_env *env, struct cl_object *obj, - struct cl_lock *lock, const struct cl_io *io) + struct cl_lock *lock, const struct cl_io *io) { - return LOV_2DISPATCH(cl2lov(obj), llo_lock_init, env, obj, lock, io); + /* No need to lock because we've taken one refcount of layout. */ + return LOV_2DISPATCH_NOLOCK(cl2lov(obj), llo_lock_init, env, obj, lock, + io); } static const struct cl_object_operations lov_ops = { @@ -694,4 +853,90 @@ struct lu_object *lov_object_alloc(const struct lu_env *env, RETURN(obj); } +struct lov_stripe_md *lov_lsm_addref(struct lov_object *lov) +{ + struct lov_stripe_md *lsm = NULL; + + lov_conf_freeze(lov); + if (lov->lo_lsm != NULL) { + lsm = lsm_addref(lov->lo_lsm); + CDEBUG(D_INODE, "lsm %p addref %d/%d by %p.\n", + lsm, cfs_atomic_read(&lsm->lsm_refc), + lov->lo_layout_invalid, cfs_current()); + } + lov_conf_thaw(lov); + return lsm; +} + +void lov_lsm_decref(struct lov_object *lov, struct lov_stripe_md *lsm) +{ + if (lsm == NULL) + return; + + CDEBUG(D_INODE, "lsm %p decref %d by %p.\n", + lsm, cfs_atomic_read(&lsm->lsm_refc), cfs_current()); + + lov_free_memmd(&lsm); +} + +struct lov_stripe_md *lov_lsm_get(struct cl_object *clobj) +{ + struct lu_object *luobj; + struct lov_stripe_md *lsm = NULL; + + if (clobj == NULL) + return NULL; + + luobj = lu_object_locate(&cl_object_header(clobj)->coh_lu, + &lov_device_type); + if (luobj != NULL) + lsm = lov_lsm_addref(lu2lov(luobj)); + return lsm; +} +EXPORT_SYMBOL(lov_lsm_get); + +void lov_lsm_put(struct cl_object *unused, struct lov_stripe_md *lsm) +{ + if (lsm != NULL) + lov_free_memmd(&lsm); +} +EXPORT_SYMBOL(lov_lsm_put); + +int lov_read_and_clear_async_rc(struct cl_object *clob) +{ + struct lu_object *luobj; + int rc = 0; + ENTRY; + + luobj = lu_object_locate(&cl_object_header(clob)->coh_lu, + &lov_device_type); + if (luobj != NULL) { + struct lov_object *lov = lu2lov(luobj); + + lov_conf_freeze(lov); + switch (lov->lo_type) { + case LLT_RAID0: { + struct lov_stripe_md *lsm; + int i; + + lsm = lov->lo_lsm; + LASSERT(lsm != NULL); + for (i = 0; i < lsm->lsm_stripe_count; i++) { + struct lov_oinfo *loi = lsm->lsm_oinfo[i]; + if (loi->loi_ar.ar_rc && !rc) + rc = loi->loi_ar.ar_rc; + loi->loi_ar.ar_rc = 0; + } + } + case LLT_EMPTY: + break; + default: + LBUG(); + } + lov_conf_thaw(lov); + } + RETURN(rc); +} +EXPORT_SYMBOL(lov_read_and_clear_async_rc); + /** @} lov */