From 02b2f549d502b46e68b97ea1452fb8853b3327dd Mon Sep 17 00:00:00 2001 From: Dongsheng Yang Date: Tue, 18 Dec 2018 04:31:48 -0500 Subject: libceph: allow setting abort_on_full for rbd Introduce a new option abort_on_full, default to false. Then we can get -ENOSPC when the pool is full, or reaches quota. [ Don't show abort_on_full in /proc/mounts. ] Signed-off-by: Dongsheng Yang Reviewed-by: Ilya Dryomov Signed-off-by: Ilya Dryomov --- net/ceph/ceph_common.c | 11 ++++++++++- net/ceph/debugfs.c | 2 +- net/ceph/osd_client.c | 4 ++-- 3 files changed, 13 insertions(+), 4 deletions(-) (limited to 'net/ceph') diff --git a/net/ceph/ceph_common.c b/net/ceph/ceph_common.c index 87afb9ec4c68..9cab80207ced 100644 --- a/net/ceph/ceph_common.c +++ b/net/ceph/ceph_common.c @@ -255,6 +255,7 @@ enum { Opt_nocephx_sign_messages, Opt_tcp_nodelay, Opt_notcp_nodelay, + Opt_abort_on_full, }; static match_table_t opt_tokens = { @@ -280,6 +281,7 @@ static match_table_t opt_tokens = { {Opt_nocephx_sign_messages, "nocephx_sign_messages"}, {Opt_tcp_nodelay, "tcp_nodelay"}, {Opt_notcp_nodelay, "notcp_nodelay"}, + {Opt_abort_on_full, "abort_on_full"}, {-1, NULL} }; @@ -535,6 +537,10 @@ ceph_parse_options(char *options, const char *dev_name, opt->flags &= ~CEPH_OPT_TCP_NODELAY; break; + case Opt_abort_on_full: + opt->flags |= CEPH_OPT_ABORT_ON_FULL; + break; + default: BUG_ON(token); } @@ -549,7 +555,8 @@ out: } EXPORT_SYMBOL(ceph_parse_options); -int ceph_print_client_options(struct seq_file *m, struct ceph_client *client) +int ceph_print_client_options(struct seq_file *m, struct ceph_client *client, + bool show_all) { struct ceph_options *opt = client->options; size_t pos = m->count; @@ -574,6 +581,8 @@ int ceph_print_client_options(struct seq_file *m, struct ceph_client *client) seq_puts(m, "nocephx_sign_messages,"); if ((opt->flags & CEPH_OPT_TCP_NODELAY) == 0) seq_puts(m, "notcp_nodelay,"); + if (show_all && (opt->flags & CEPH_OPT_ABORT_ON_FULL)) + seq_puts(m, "abort_on_full,"); if (opt->mount_timeout != CEPH_MOUNT_TIMEOUT_DEFAULT) seq_printf(m, "mount_timeout=%d,", diff --git a/net/ceph/debugfs.c b/net/ceph/debugfs.c index 02952605d121..46f65709a6ff 100644 --- a/net/ceph/debugfs.c +++ b/net/ceph/debugfs.c @@ -375,7 +375,7 @@ static int client_options_show(struct seq_file *s, void *p) struct ceph_client *client = s->private; int ret; - ret = ceph_print_client_options(s, client); + ret = ceph_print_client_options(s, client, true); if (ret) return ret; diff --git a/net/ceph/osd_client.c b/net/ceph/osd_client.c index d23a9f81f3d7..fa9530dd876e 100644 --- a/net/ceph/osd_client.c +++ b/net/ceph/osd_client.c @@ -2315,7 +2315,7 @@ again: (ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) || pool_full(osdc, req->r_t.base_oloc.pool))) { dout("req %p full/pool_full\n", req); - if (osdc->abort_on_full) { + if (ceph_test_opt(osdc->client, ABORT_ON_FULL)) { err = -ENOSPC; } else { pr_warn_ratelimited("FULL or reached pool quota\n"); @@ -2545,7 +2545,7 @@ static void ceph_osdc_abort_on_full(struct ceph_osd_client *osdc) { bool victims = false; - if (osdc->abort_on_full && + if (ceph_test_opt(osdc->client, ABORT_ON_FULL) && (ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) || have_pool_full(osdc))) for_each_request(osdc, abort_on_full_fn, &victims); } -- cgit v1.2.3 From 4aac9228d16458cedcfd90c7fb37211cf3653ac3 Mon Sep 17 00:00:00 2001 From: Ilya Dryomov Date: Mon, 14 Jan 2019 21:13:10 +0100 Subject: libceph: avoid KEEPALIVE_PENDING races in ceph_con_keepalive() con_fault() can transition the connection into STANDBY right after ceph_con_keepalive() clears STANDBY in clear_standby(): libceph user thread ceph-msgr worker ceph_con_keepalive() mutex_lock(&con->mutex) clear_standby(con) mutex_unlock(&con->mutex) mutex_lock(&con->mutex) con_fault() ... if KEEPALIVE_PENDING isn't set set state to STANDBY ... mutex_unlock(&con->mutex) set KEEPALIVE_PENDING set WRITE_PENDING This triggers warnings in clear_standby() when either ceph_con_send() or ceph_con_keepalive() get to clearing STANDBY next time. I don't see a reason to condition queue_con() call on the previous value of KEEPALIVE_PENDING, so move the setting of KEEPALIVE_PENDING into the critical section -- unlike WRITE_PENDING, KEEPALIVE_PENDING could have been a non-atomic flag. Reported-by: syzbot+acdeb633f6211ccdf886@syzkaller.appspotmail.com Signed-off-by: Ilya Dryomov Tested-by: Myungho Jung --- net/ceph/messenger.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'net/ceph') diff --git a/net/ceph/messenger.c b/net/ceph/messenger.c index d5718284db57..3661cdd927f1 100644 --- a/net/ceph/messenger.c +++ b/net/ceph/messenger.c @@ -3206,9 +3206,10 @@ void ceph_con_keepalive(struct ceph_connection *con) dout("con_keepalive %p\n", con); mutex_lock(&con->mutex); clear_standby(con); + con_flag_set(con, CON_FLAG_KEEPALIVE_PENDING); mutex_unlock(&con->mutex); - if (con_flag_test_and_set(con, CON_FLAG_KEEPALIVE_PENDING) == 0 && - con_flag_test_and_set(con, CON_FLAG_WRITE_PENDING) == 0) + + if (con_flag_test_and_set(con, CON_FLAG_WRITE_PENDING) == 0) queue_con(con); } EXPORT_SYMBOL(ceph_con_keepalive); -- cgit v1.2.3