From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from us-smtp-delivery-124.mimecast.com (us-smtp-delivery-124.mimecast.com [170.10.133.124]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by smtp.subspace.kernel.org (Postfix) with ESMTPS id 210581400A for ; Sun, 19 Nov 2023 16:38:26 +0000 (UTC) Authentication-Results: smtp.subspace.kernel.org; dmarc=pass (p=none dis=none) header.from=redhat.com Authentication-Results: smtp.subspace.kernel.org; spf=pass smtp.mailfrom=redhat.com Authentication-Results: smtp.subspace.kernel.org; dkim=pass (1024-bit key) header.d=redhat.com header.i=@redhat.com header.b="Sho/T343" DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=redhat.com; s=mimecast20190719; t=1700411905; h=from:from:reply-to:subject:subject:date:date:message-id:message-id: to:to:cc:cc:mime-version:mime-version:content-type:content-type: content-transfer-encoding:content-transfer-encoding: in-reply-to:in-reply-to:references:references; bh=tg5hFP9QXIUSPFfiYNBsI0baLFK4dmGqBn+poj7NQzk=; b=Sho/T3436wqfyYnyE9pZGZtWwtzf714RZmPR1HLMD2Eou/l8NFza85lF1TyG7/LQky2o/z 7D684jEaD8sDq3a3iQPz8qoWQzZfHPKKvBARw3cJdbftryJMNhS7tj9Zqo2SrXEsQwPG8V bZKLDrmENCShSxI6rLnR2B4397Qww2o= Received: from mimecast-mx02.redhat.com (mimecast-mx02.redhat.com [66.187.233.88]) by relay.mimecast.com with ESMTP with STARTTLS (version=TLSv1.3, cipher=TLS_AES_256_GCM_SHA384) id us-mta-564-a-U-2S6DO36rXnGnrBmhgg-1; Sun, 19 Nov 2023 11:38:23 -0500 X-MC-Unique: a-U-2S6DO36rXnGnrBmhgg-1 Received: from smtp.corp.redhat.com (int-mx02.intmail.prod.int.rdu2.redhat.com [10.11.54.2]) (using TLSv1.3 with cipher TLS_AES_256_GCM_SHA384 (256/256 bits) key-exchange X25519 server-signature RSA-PSS (2048 bits) server-digest SHA256) (No client certificate requested) by mimecast-mx02.redhat.com (Postfix) with ESMTPS id A1635828B21 for ; Sun, 19 Nov 2023 16:38:23 +0000 (UTC) Received: from fs-i40c-03.fs.lab.eng.bos.redhat.com (fs-i40c-03.fast.rdu2.eng.redhat.com [10.6.23.54]) by smtp.corp.redhat.com (Postfix) with ESMTP id 97A6D40C6EBB; Sun, 19 Nov 2023 16:38:23 +0000 (UTC) From: Alexander Aring To: teigland@redhat.com Cc: gfs2@lists.linux.dev, aahringo@redhat.com Subject: [PATCHv2 dlm/next 05/13] dlm: move root_list to ls_recover() stack Date: Sun, 19 Nov 2023 11:38:09 -0500 Message-Id: <20231119163817.751872-6-aahringo@redhat.com> In-Reply-To: <20231119163817.751872-1-aahringo@redhat.com> References: <20231119163817.751872-1-aahringo@redhat.com> Precedence: bulk X-Mailing-List: gfs2@lists.linux.dev List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 X-Scanned-By: MIMEDefang 3.4.1 on 10.11.54.2 X-Mimecast-Spam-Score: 0 X-Mimecast-Originator: redhat.com Content-Transfer-Encoding: 8bit Content-Type: text/plain; charset="US-ASCII"; x-default=true This patch moves the per lockspace ls_root_list list which is mainly used for snapshoting all dlm_rsb from a hash to a list to do recovery handling into the recovery function ls_recover() as stack variable. Doing that shows that there is no need for locking the ls_root_list which is created at the beginning of ls_recover() and destroyed at the ending of ls_recover(). In between only functionality is called doing read only access to the root_list stack variable. A special case is assigning the per lockspace ls_recover_dir_root_list variable to the stack variable. The ls_recover_dir_root_list get accessed by another concurrent process dlm_copy_master_names() during the time between ls_recover_dir_root_list is set and set to NULL again. This is done by a special distributed barrier functionality between dlm_recover_members_wait() and dlm_recover_directory_wait(). A comment was made to mention about this handling which might be changed to a better behaviour in future. However setting the ls_recover_dir_root_list to the stack variable and set it to NULL in this specific time will show us potential issues with the recovery handling if it's breaks. Signed-off-by: Alexander Aring --- fs/dlm/dir.c | 6 ++---- fs/dlm/dir.h | 3 ++- fs/dlm/dlm_internal.h | 6 ++---- fs/dlm/lock.c | 6 ++---- fs/dlm/lock.h | 2 +- fs/dlm/lockspace.c | 2 -- fs/dlm/recover.c | 30 ++++++++++-------------------- fs/dlm/recover.h | 8 +++++--- fs/dlm/recoverd.c | 35 +++++++++++++---------------------- 9 files changed, 37 insertions(+), 61 deletions(-) diff --git a/fs/dlm/dir.c b/fs/dlm/dir.c index 10753486049a..3da00c46cbb3 100644 --- a/fs/dlm/dir.c +++ b/fs/dlm/dir.c @@ -47,15 +47,13 @@ int dlm_dir_nodeid(struct dlm_rsb *r) return r->res_dir_nodeid; } -void dlm_recover_dir_nodeid(struct dlm_ls *ls) +void dlm_recover_dir_nodeid(struct dlm_ls *ls, const struct list_head *root_list) { struct dlm_rsb *r; - down_read(&ls->ls_root_sem); - list_for_each_entry(r, &ls->ls_root_list, res_root_list) { + list_for_each_entry(r, root_list, res_root_list) { r->res_dir_nodeid = dlm_hash2nodeid(ls, r->res_hash); } - up_read(&ls->ls_root_sem); } int dlm_recover_directory(struct dlm_ls *ls, uint64_t seq) diff --git a/fs/dlm/dir.h b/fs/dlm/dir.h index 39ecb69d7ef3..5b2a7ee3762d 100644 --- a/fs/dlm/dir.h +++ b/fs/dlm/dir.h @@ -14,7 +14,8 @@ int dlm_dir_nodeid(struct dlm_rsb *rsb); int dlm_hash2nodeid(struct dlm_ls *ls, uint32_t hash); -void dlm_recover_dir_nodeid(struct dlm_ls *ls); +void dlm_recover_dir_nodeid(struct dlm_ls *ls, + const struct list_head *root_list); int dlm_recover_directory(struct dlm_ls *ls, uint64_t seq); void dlm_copy_master_names(struct dlm_ls *ls, const char *inbuf, int inlen, char *outbuf, int outlen, int nodeid); diff --git a/fs/dlm/dlm_internal.h b/fs/dlm/dlm_internal.h index cb18f383acff..959f69fb2a52 100644 --- a/fs/dlm/dlm_internal.h +++ b/fs/dlm/dlm_internal.h @@ -644,10 +644,8 @@ struct dlm_ls { wait_queue_head_t ls_recover_lock_wait; spinlock_t ls_clear_proc_locks; - struct list_head ls_root_list; /* root resources */ - struct rw_semaphore ls_root_sem; /* protect root_list */ - struct list_head ls_masters_list; /* root resources */ - rwlock_t ls_masters_lock; /* protect root_list */ + struct list_head ls_masters_list; /* root resources */ + rwlock_t ls_masters_lock; /* protect root_list */ const struct dlm_lockspace_ops *ls_ops; void *ls_ops_arg; diff --git a/fs/dlm/lock.c b/fs/dlm/lock.c index 970b8499b66f..0218645e2f90 100644 --- a/fs/dlm/lock.c +++ b/fs/dlm/lock.c @@ -5185,7 +5185,7 @@ static void purge_dead_list(struct dlm_ls *ls, struct dlm_rsb *r, /* Get rid of locks held by nodes that are gone. */ -void dlm_recover_purge(struct dlm_ls *ls) +void dlm_recover_purge(struct dlm_ls *ls, const struct list_head *root_list) { struct dlm_rsb *r; struct dlm_member *memb; @@ -5204,8 +5204,7 @@ void dlm_recover_purge(struct dlm_ls *ls) if (!nodes_count) return; - down_write(&ls->ls_root_sem); - list_for_each_entry(r, &ls->ls_root_list, res_root_list) { + list_for_each_entry(r, root_list, res_root_list) { hold_rsb(r); lock_rsb(r); if (is_master(r)) { @@ -5220,7 +5219,6 @@ void dlm_recover_purge(struct dlm_ls *ls) unhold_rsb(r); cond_resched(); } - up_write(&ls->ls_root_sem); if (lkb_count) log_rinfo(ls, "dlm_recover_purge %u locks for %u nodes", diff --git a/fs/dlm/lock.h b/fs/dlm/lock.h index b54e2cbbe6e2..c8ff7780d3cc 100644 --- a/fs/dlm/lock.h +++ b/fs/dlm/lock.h @@ -31,7 +31,7 @@ int dlm_master_lookup(struct dlm_ls *ls, int from_nodeid, const char *name, int dlm_search_rsb_tree(struct rb_root *tree, const void *name, int len, struct dlm_rsb **r_ret); -void dlm_recover_purge(struct dlm_ls *ls); +void dlm_recover_purge(struct dlm_ls *ls, const struct list_head *root_list); void dlm_purge_mstcpy_locks(struct dlm_rsb *r); void dlm_recover_grant(struct dlm_ls *ls); int dlm_recover_waiters_post(struct dlm_ls *ls); diff --git a/fs/dlm/lockspace.c b/fs/dlm/lockspace.c index c427c76b5f07..da756e5c0f6c 100644 --- a/fs/dlm/lockspace.c +++ b/fs/dlm/lockspace.c @@ -580,8 +580,6 @@ static int new_lockspace(const char *name, const char *cluster, ls->ls_recover_list_count = 0; ls->ls_local_handle = ls; init_waitqueue_head(&ls->ls_wait_general); - INIT_LIST_HEAD(&ls->ls_root_list); - init_rwsem(&ls->ls_root_sem); INIT_LIST_HEAD(&ls->ls_masters_list); rwlock_init(&ls->ls_masters_lock); diff --git a/fs/dlm/recover.c b/fs/dlm/recover.c index 6abc283f8f36..172c6b73f37a 100644 --- a/fs/dlm/recover.c +++ b/fs/dlm/recover.c @@ -519,7 +519,8 @@ static int recover_master_static(struct dlm_rsb *r, unsigned int *count) * the correct dir node. */ -int dlm_recover_masters(struct dlm_ls *ls, uint64_t seq) +int dlm_recover_masters(struct dlm_ls *ls, uint64_t seq, + const struct list_head *root_list) { struct dlm_rsb *r; unsigned int total = 0; @@ -529,10 +530,8 @@ int dlm_recover_masters(struct dlm_ls *ls, uint64_t seq) log_rinfo(ls, "dlm_recover_masters"); - down_read(&ls->ls_root_sem); - list_for_each_entry(r, &ls->ls_root_list, res_root_list) { + list_for_each_entry(r, root_list, res_root_list) { if (dlm_recovery_stopped(ls)) { - up_read(&ls->ls_root_sem); error = -EINTR; goto out; } @@ -546,12 +545,9 @@ int dlm_recover_masters(struct dlm_ls *ls, uint64_t seq) cond_resched(); total++; - if (error) { - up_read(&ls->ls_root_sem); + if (error) goto out; - } } - up_read(&ls->ls_root_sem); log_rinfo(ls, "dlm_recover_masters %u of %u", count, total); @@ -656,13 +652,13 @@ static int recover_locks(struct dlm_rsb *r, uint64_t seq) return error; } -int dlm_recover_locks(struct dlm_ls *ls, uint64_t seq) +int dlm_recover_locks(struct dlm_ls *ls, uint64_t seq, + const struct list_head *root_list) { struct dlm_rsb *r; int error, count = 0; - down_read(&ls->ls_root_sem); - list_for_each_entry(r, &ls->ls_root_list, res_root_list) { + list_for_each_entry(r, root_list, res_root_list) { if (is_master(r)) { rsb_clear_flag(r, RSB_NEW_MASTER); continue; @@ -673,19 +669,15 @@ int dlm_recover_locks(struct dlm_ls *ls, uint64_t seq) if (dlm_recovery_stopped(ls)) { error = -EINTR; - up_read(&ls->ls_root_sem); goto out; } error = recover_locks(r, seq); - if (error) { - up_read(&ls->ls_root_sem); + if (error) goto out; - } count += r->res_recover_locks_count; } - up_read(&ls->ls_root_sem); log_rinfo(ls, "dlm_recover_locks %d out", count); @@ -854,13 +846,12 @@ static void recover_grant(struct dlm_rsb *r) rsb_set_flag(r, RSB_RECOVER_GRANT); } -void dlm_recover_rsbs(struct dlm_ls *ls) +void dlm_recover_rsbs(struct dlm_ls *ls, const struct list_head *root_list) { struct dlm_rsb *r; unsigned int count = 0; - down_read(&ls->ls_root_sem); - list_for_each_entry(r, &ls->ls_root_list, res_root_list) { + list_for_each_entry(r, root_list, res_root_list) { lock_rsb(r); if (is_master(r)) { if (rsb_flag(r, RSB_RECOVER_CONVERT)) @@ -881,7 +872,6 @@ void dlm_recover_rsbs(struct dlm_ls *ls) rsb_clear_flag(r, RSB_NEW_MASTER2); unlock_rsb(r); } - up_read(&ls->ls_root_sem); if (count) log_rinfo(ls, "dlm_recover_rsbs %d done", count); diff --git a/fs/dlm/recover.h b/fs/dlm/recover.h index 0b54550ee055..efc79a6e577d 100644 --- a/fs/dlm/recover.h +++ b/fs/dlm/recover.h @@ -19,12 +19,14 @@ int dlm_recover_members_wait(struct dlm_ls *ls, uint64_t seq); int dlm_recover_directory_wait(struct dlm_ls *ls, uint64_t seq); int dlm_recover_locks_wait(struct dlm_ls *ls, uint64_t seq); int dlm_recover_done_wait(struct dlm_ls *ls, uint64_t seq); -int dlm_recover_masters(struct dlm_ls *ls, uint64_t seq); +int dlm_recover_masters(struct dlm_ls *ls, uint64_t seq, + const struct list_head *root_list); int dlm_recover_master_reply(struct dlm_ls *ls, const struct dlm_rcom *rc); -int dlm_recover_locks(struct dlm_ls *ls, uint64_t seq); +int dlm_recover_locks(struct dlm_ls *ls, uint64_t seq, + const struct list_head *root_list); void dlm_recovered_lock(struct dlm_rsb *r); void dlm_clear_toss(struct dlm_ls *ls); -void dlm_recover_rsbs(struct dlm_ls *ls); +void dlm_recover_rsbs(struct dlm_ls *ls, const struct list_head *root_list); #endif /* __RECOVER_DOT_H__ */ diff --git a/fs/dlm/recoverd.c b/fs/dlm/recoverd.c index dfce8fc6a783..e5649201ba23 100644 --- a/fs/dlm/recoverd.c +++ b/fs/dlm/recoverd.c @@ -62,23 +62,17 @@ static void dlm_release_masters_list(struct dlm_ls *ls) write_unlock(&ls->ls_masters_lock); } -static void dlm_create_root_list(struct dlm_ls *ls) +static void dlm_create_root_list(struct dlm_ls *ls, struct list_head *root_list) { struct rb_node *n; struct dlm_rsb *r; int i; - down_write(&ls->ls_root_sem); - if (!list_empty(&ls->ls_root_list)) { - log_error(ls, "root list not empty"); - goto out; - } - for (i = 0; i < ls->ls_rsbtbl_size; i++) { spin_lock_bh(&ls->ls_rsbtbl[i].lock); for (n = rb_first(&ls->ls_rsbtbl[i].keep); n; n = rb_next(n)) { r = rb_entry(n, struct dlm_rsb, res_hashnode); - list_add(&r->res_root_list, &ls->ls_root_list); + list_add(&r->res_root_list, root_list); dlm_hold_rsb(r); } @@ -86,20 +80,16 @@ static void dlm_create_root_list(struct dlm_ls *ls) log_error(ls, "%s toss not empty", __func__); spin_unlock_bh(&ls->ls_rsbtbl[i].lock); } - out: - up_write(&ls->ls_root_sem); } -static void dlm_release_root_list(struct dlm_ls *ls) +static void dlm_release_root_list(struct list_head *root_list) { struct dlm_rsb *r, *safe; - down_write(&ls->ls_root_sem); - list_for_each_entry_safe(r, safe, &ls->ls_root_list, res_root_list) { + list_for_each_entry_safe(r, safe, root_list, res_root_list) { list_del_init(&r->res_root_list); dlm_put_rsb(r); } - up_write(&ls->ls_root_sem); } /* If the start for which we're re-enabling locking (seq) has been superseded @@ -131,6 +121,7 @@ static int enable_locking(struct dlm_ls *ls, uint64_t seq) static int ls_recover(struct dlm_ls *ls, struct dlm_recover *rv) { + LIST_HEAD(root_list); unsigned long start; int error, neg = 0; @@ -147,7 +138,7 @@ static int ls_recover(struct dlm_ls *ls, struct dlm_recover *rv) * routines. */ - dlm_create_root_list(ls); + dlm_create_root_list(ls, &root_list); /* * Add or remove nodes from the lockspace's ls_nodes list. @@ -163,7 +154,7 @@ static int ls_recover(struct dlm_ls *ls, struct dlm_recover *rv) goto fail; } - dlm_recover_dir_nodeid(ls); + dlm_recover_dir_nodeid(ls, &root_list); /* Create a snapshot of all active rsbs were we are the master of. * During the barrier between dlm_recover_members_wait() and @@ -241,14 +232,14 @@ static int ls_recover(struct dlm_ls *ls, struct dlm_recover *rv) * Clear lkb's for departed nodes. */ - dlm_recover_purge(ls); + dlm_recover_purge(ls, &root_list); /* * Get new master nodeid's for rsb's that were mastered on * departed nodes. */ - error = dlm_recover_masters(ls, rv->seq); + error = dlm_recover_masters(ls, rv->seq, &root_list); if (error) { log_rinfo(ls, "dlm_recover_masters error %d", error); goto fail; @@ -258,7 +249,7 @@ static int ls_recover(struct dlm_ls *ls, struct dlm_recover *rv) * Send our locks on remastered rsb's to the new masters. */ - error = dlm_recover_locks(ls, rv->seq); + error = dlm_recover_locks(ls, rv->seq, &root_list); if (error) { log_rinfo(ls, "dlm_recover_locks error %d", error); goto fail; @@ -281,7 +272,7 @@ static int ls_recover(struct dlm_ls *ls, struct dlm_recover *rv) * settings. */ - dlm_recover_rsbs(ls); + dlm_recover_rsbs(ls, &root_list); } else { /* * Other lockspace members may be going through the "neg" steps @@ -297,7 +288,7 @@ static int ls_recover(struct dlm_ls *ls, struct dlm_recover *rv) } } - dlm_release_root_list(ls); + dlm_release_root_list(&root_list); /* * Purge directory-related requests that are saved in requestqueue. @@ -347,7 +338,7 @@ static int ls_recover(struct dlm_ls *ls, struct dlm_recover *rv) return 0; fail: - dlm_release_root_list(ls); + dlm_release_root_list(&root_list); mutex_unlock(&ls->ls_recoverd_active); return error; -- 2.39.3