From df9f86faf3ee610527ed02031fe7dd3c8b752e44 Mon Sep 17 00:00:00 2001 From: Sage Weil Date: Mon, 1 Nov 2010 15:49:23 -0700 Subject: ceph: fix small seq message skipping If the client gets out of sync with the server message sequence number, we normally skip low seq messages (ones we already received). The skip code was also incrementing the expected seq, such that all subsequent messages also appeared old and got skipped, and an eventual timeout on the osd connection. This resulted in some lagging requests and console messages like [233480.882885] ceph: skipping osd22 10.138.138.13:6804 seq 2016, expected 2017 [233480.882919] ceph: skipping osd22 10.138.138.13:6804 seq 2017, expected 2018 [233480.882963] ceph: skipping osd22 10.138.138.13:6804 seq 2018, expected 2019 [233480.883488] ceph: skipping osd22 10.138.138.13:6804 seq 2019, expected 2020 [233485.219558] ceph: skipping osd22 10.138.138.13:6804 seq 2020, expected 2021 [233485.906595] ceph: skipping osd22 10.138.138.13:6804 seq 2021, expected 2022 [233490.379536] ceph: skipping osd22 10.138.138.13:6804 seq 2022, expected 2023 [233495.523260] ceph: skipping osd22 10.138.138.13:6804 seq 2023, expected 2024 [233495.923194] ceph: skipping osd22 10.138.138.13:6804 seq 2024, expected 2025 [233500.534614] ceph: tid 6023602 timed out on osd22, will reset osd Reported-by: Theodore Ts'o Signed-off-by: Sage Weil --- net/ceph/messenger.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'net/ceph/messenger.c') diff --git a/net/ceph/messenger.c b/net/ceph/messenger.c index 0e8157ee5d4..d379abf873b 100644 --- a/net/ceph/messenger.c +++ b/net/ceph/messenger.c @@ -1532,14 +1532,13 @@ static int read_partial_message(struct ceph_connection *con) /* verify seq# */ seq = le64_to_cpu(con->in_hdr.seq); if ((s64)seq - (s64)con->in_seq < 1) { - pr_info("skipping %s%lld %s seq %lld, expected %lld\n", + pr_info("skipping %s%lld %s seq %lld expected %lld\n", ENTITY_NAME(con->peer_name), ceph_pr_addr(&con->peer_addr.in_addr), seq, con->in_seq + 1); con->in_base_pos = -front_len - middle_len - data_len - sizeof(m->footer); con->in_tag = CEPH_MSGR_TAG_READY; - con->in_seq++; return 0; } else if ((s64)seq - (s64)con->in_seq > 1) { pr_err("read_partial_message bad seq %lld expected %lld\n", -- cgit v1.2.3-18-g5258 From c5c6b19d4b8f5431fca05f28ae9e141045022149 Mon Sep 17 00:00:00 2001 From: Sage Weil Date: Tue, 9 Nov 2010 12:40:00 -0800 Subject: ceph: explicitly specify page alignment in network messages The alignment used for reading data into or out of pages used to be taken from the data_off field in the message header. This only worked as long as the page alignment matched the object offset, breaking direct io to non-page aligned offsets. Instead, explicitly specify the page alignment next to the page vector in the ceph_msg struct, and use that instead of the message header (which probably shouldn't be trusted). The alloc_msg callback is responsible for filling in this field properly when it sets up the page vector. Signed-off-by: Sage Weil --- net/ceph/messenger.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) (limited to 'net/ceph/messenger.c') diff --git a/net/ceph/messenger.c b/net/ceph/messenger.c index d379abf873b..1c7a2ec4f3c 100644 --- a/net/ceph/messenger.c +++ b/net/ceph/messenger.c @@ -540,8 +540,7 @@ static void prepare_write_message(struct ceph_connection *con) /* initialize page iterator */ con->out_msg_pos.page = 0; if (m->pages) - con->out_msg_pos.page_pos = - le16_to_cpu(m->hdr.data_off) & ~PAGE_MASK; + con->out_msg_pos.page_pos = m->page_alignment; else con->out_msg_pos.page_pos = 0; con->out_msg_pos.data_pos = 0; @@ -1491,7 +1490,7 @@ static int read_partial_message(struct ceph_connection *con) struct ceph_msg *m = con->in_msg; int ret; int to, left; - unsigned front_len, middle_len, data_len, data_off; + unsigned front_len, middle_len, data_len; int datacrc = con->msgr->nocrc; int skip; u64 seq; @@ -1527,7 +1526,6 @@ static int read_partial_message(struct ceph_connection *con) data_len = le32_to_cpu(con->in_hdr.data_len); if (data_len > CEPH_MSG_MAX_DATA_LEN) return -EIO; - data_off = le16_to_cpu(con->in_hdr.data_off); /* verify seq# */ seq = le64_to_cpu(con->in_hdr.seq); @@ -1575,7 +1573,7 @@ static int read_partial_message(struct ceph_connection *con) con->in_msg_pos.page = 0; if (m->pages) - con->in_msg_pos.page_pos = data_off & ~PAGE_MASK; + con->in_msg_pos.page_pos = m->page_alignment; else con->in_msg_pos.page_pos = 0; con->in_msg_pos.data_pos = 0; @@ -2300,6 +2298,7 @@ struct ceph_msg *ceph_msg_new(int type, int front_len, gfp_t flags) /* data */ m->nr_pages = 0; + m->page_alignment = 0; m->pages = NULL; m->pagelist = NULL; m->bio = NULL; @@ -2369,6 +2368,7 @@ static struct ceph_msg *ceph_alloc_msg(struct ceph_connection *con, type, front_len); return NULL; } + msg->page_alignment = le16_to_cpu(hdr->data_off); } memcpy(&msg->hdr, &con->in_hdr, sizeof(con->in_hdr)); -- cgit v1.2.3-18-g5258 From d96c9043d1588f04c7f467167f653c07d83232d5 Mon Sep 17 00:00:00 2001 From: Sage Weil Date: Mon, 13 Dec 2010 20:30:28 -0800 Subject: ceph: fix msgr_init error path create_workqueue() returns NULL on failure. Signed-off-by: Sage Weil --- net/ceph/messenger.c | 8 +++----- 1 file changed, 3 insertions(+), 5 deletions(-) (limited to 'net/ceph/messenger.c') diff --git a/net/ceph/messenger.c b/net/ceph/messenger.c index 1c7a2ec4f3c..b6ff4a1519a 100644 --- a/net/ceph/messenger.c +++ b/net/ceph/messenger.c @@ -97,11 +97,9 @@ struct workqueue_struct *ceph_msgr_wq; int ceph_msgr_init(void) { ceph_msgr_wq = create_workqueue("ceph-msgr"); - if (IS_ERR(ceph_msgr_wq)) { - int ret = PTR_ERR(ceph_msgr_wq); - pr_err("msgr_init failed to create workqueue: %d\n", ret); - ceph_msgr_wq = NULL; - return ret; + if (!ceph_msgr_wq) { + pr_err("msgr_init failed to create workqueue\n"); + return -ENOMEM; } return 0; } -- cgit v1.2.3-18-g5258