diff options
Diffstat (limited to 'net/ceph')
| -rw-r--r-- | net/ceph/Kconfig | 3 | ||||
| -rw-r--r-- | net/ceph/auth_x.c | 2 | ||||
| -rw-r--r-- | net/ceph/ceph_common.c | 58 | ||||
| -rw-r--r-- | net/ceph/debugfs.c | 14 | ||||
| -rw-r--r-- | net/ceph/messenger.c | 24 | ||||
| -rw-r--r-- | net/ceph/messenger_v1.c | 56 | ||||
| -rw-r--r-- | net/ceph/messenger_v2.c | 263 | ||||
| -rw-r--r-- | net/ceph/mon_client.c | 2 | ||||
| -rw-r--r-- | net/ceph/osdmap.c | 18 |
9 files changed, 218 insertions, 222 deletions
diff --git a/net/ceph/Kconfig b/net/ceph/Kconfig index 0aa21fcbf6ec..ea60e3ef0834 100644 --- a/net/ceph/Kconfig +++ b/net/ceph/Kconfig @@ -6,8 +6,7 @@ config CEPH_LIB select CRYPTO_AES select CRYPTO_CBC select CRYPTO_GCM - select CRYPTO_HMAC - select CRYPTO_SHA256 + select CRYPTO_LIB_SHA256 select CRYPTO select KEYS default n diff --git a/net/ceph/auth_x.c b/net/ceph/auth_x.c index b71b1635916e..a21c157daf7d 100644 --- a/net/ceph/auth_x.c +++ b/net/ceph/auth_x.c @@ -631,6 +631,7 @@ static int handle_auth_session_key(struct ceph_auth_client *ac, u64 global_id, /* connection secret */ ceph_decode_32_safe(p, end, len, e_inval); + ceph_decode_need(p, end, len, e_inval); dout("%s connection secret blob len %d\n", __func__, len); if (len > 0) { dp = *p + ceph_x_encrypt_offset(); @@ -648,6 +649,7 @@ static int handle_auth_session_key(struct ceph_auth_client *ac, u64 global_id, /* service tickets */ ceph_decode_32_safe(p, end, len, e_inval); + ceph_decode_need(p, end, len, e_inval); dout("%s service tickets blob len %d\n", __func__, len); if (len > 0) { ret = ceph_x_proc_ticket_reply(ac, &th->session_key, diff --git a/net/ceph/ceph_common.c b/net/ceph/ceph_common.c index 4c6441536d55..e734e57be083 100644 --- a/net/ceph/ceph_common.c +++ b/net/ceph/ceph_common.c @@ -786,41 +786,52 @@ void ceph_reset_client_addr(struct ceph_client *client) EXPORT_SYMBOL(ceph_reset_client_addr); /* - * true if we have the mon map (and have thus joined the cluster) - */ -static bool have_mon_and_osd_map(struct ceph_client *client) -{ - return client->monc.monmap && client->monc.monmap->epoch && - client->osdc.osdmap && client->osdc.osdmap->epoch; -} - -/* * mount: join the ceph cluster, and open root directory. */ -int __ceph_open_session(struct ceph_client *client, unsigned long started) +int __ceph_open_session(struct ceph_client *client) { - unsigned long timeout = client->options->mount_timeout; - long err; + DEFINE_WAIT_FUNC(wait, woken_wake_function); + long timeout = ceph_timeout_jiffies(client->options->mount_timeout); + bool have_monmap, have_osdmap; + int err; /* open session, and wait for mon and osd maps */ err = ceph_monc_open_session(&client->monc); if (err < 0) return err; - while (!have_mon_and_osd_map(client)) { - if (timeout && time_after_eq(jiffies, started + timeout)) - return -ETIMEDOUT; + add_wait_queue(&client->auth_wq, &wait); + for (;;) { + mutex_lock(&client->monc.mutex); + err = client->auth_err; + have_monmap = client->monc.monmap && client->monc.monmap->epoch; + mutex_unlock(&client->monc.mutex); + + down_read(&client->osdc.lock); + have_osdmap = client->osdc.osdmap && client->osdc.osdmap->epoch; + up_read(&client->osdc.lock); + + if (err || (have_monmap && have_osdmap)) + break; + + if (signal_pending(current)) { + err = -ERESTARTSYS; + break; + } + + if (!timeout) { + err = -ETIMEDOUT; + break; + } /* wait */ dout("mount waiting for mon_map\n"); - err = wait_event_interruptible_timeout(client->auth_wq, - have_mon_and_osd_map(client) || (client->auth_err < 0), - ceph_timeout_jiffies(timeout)); - if (err < 0) - return err; - if (client->auth_err < 0) - return client->auth_err; + timeout = wait_woken(&wait, TASK_INTERRUPTIBLE, timeout); } + remove_wait_queue(&client->auth_wq, &wait); + + if (err) + return err; pr_info("client%llu fsid %pU\n", ceph_client_gid(client), &client->fsid); @@ -833,12 +844,11 @@ EXPORT_SYMBOL(__ceph_open_session); int ceph_open_session(struct ceph_client *client) { int ret; - unsigned long started = jiffies; /* note the start time */ dout("open_session start\n"); mutex_lock(&client->mount_mutex); - ret = __ceph_open_session(client, started); + ret = __ceph_open_session(client); mutex_unlock(&client->mount_mutex); return ret; diff --git a/net/ceph/debugfs.c b/net/ceph/debugfs.c index 2110439f8a24..83c270bce63c 100644 --- a/net/ceph/debugfs.c +++ b/net/ceph/debugfs.c @@ -36,8 +36,9 @@ static int monmap_show(struct seq_file *s, void *p) int i; struct ceph_client *client = s->private; + mutex_lock(&client->monc.mutex); if (client->monc.monmap == NULL) - return 0; + goto out_unlock; seq_printf(s, "epoch %d\n", client->monc.monmap->epoch); for (i = 0; i < client->monc.monmap->num_mon; i++) { @@ -48,6 +49,9 @@ static int monmap_show(struct seq_file *s, void *p) ENTITY_NAME(inst->name), ceph_pr_addr(&inst->addr)); } + +out_unlock: + mutex_unlock(&client->monc.mutex); return 0; } @@ -56,13 +60,14 @@ static int osdmap_show(struct seq_file *s, void *p) int i; struct ceph_client *client = s->private; struct ceph_osd_client *osdc = &client->osdc; - struct ceph_osdmap *map = osdc->osdmap; + struct ceph_osdmap *map; struct rb_node *n; + down_read(&osdc->lock); + map = osdc->osdmap; if (map == NULL) - return 0; + goto out_unlock; - down_read(&osdc->lock); seq_printf(s, "epoch %u barrier %u flags 0x%x\n", map->epoch, osdc->epoch_barrier, map->flags); @@ -131,6 +136,7 @@ static int osdmap_show(struct seq_file *s, void *p) seq_printf(s, "]\n"); } +out_unlock: up_read(&osdc->lock); return 0; } diff --git a/net/ceph/messenger.c b/net/ceph/messenger.c index d1b5705dc0c6..70b25f4ecba6 100644 --- a/net/ceph/messenger.c +++ b/net/ceph/messenger.c @@ -252,7 +252,8 @@ int __init ceph_msgr_init(void) * The number of active work items is limited by the number of * connections, so leave @max_active at default. */ - ceph_msgr_wq = alloc_workqueue("ceph-msgr", WQ_MEM_RECLAIM, 0); + ceph_msgr_wq = alloc_workqueue("ceph-msgr", + WQ_MEM_RECLAIM | WQ_PERCPU, 0); if (ceph_msgr_wq) return 0; @@ -459,7 +460,7 @@ int ceph_tcp_connect(struct ceph_connection *con) set_sock_callbacks(sock, con); con_sock_state_connecting(con); - ret = kernel_connect(sock, (struct sockaddr *)&ss, sizeof(ss), + ret = kernel_connect(sock, (struct sockaddr_unsized *)&ss, sizeof(ss), O_NONBLOCK); if (ret == -EINPROGRESS) { dout("connect %s EINPROGRESS sk_state = %u\n", @@ -1524,7 +1525,7 @@ static void con_fault_finish(struct ceph_connection *con) * in case we faulted due to authentication, invalidate our * current tickets so that we can get new ones. */ - if (con->v1.auth_retry) { + if (!ceph_msgr2(from_msgr(con->msgr)) && con->v1.auth_retry) { dout("auth_retry %d, invalidating\n", con->v1.auth_retry); if (con->ops->invalidate_authorizer) con->ops->invalidate_authorizer(con); @@ -1714,9 +1715,10 @@ static void clear_standby(struct ceph_connection *con) { /* come back from STANDBY? */ if (con->state == CEPH_CON_S_STANDBY) { - dout("clear_standby %p and ++connect_seq\n", con); + dout("clear_standby %p\n", con); con->state = CEPH_CON_S_PREOPEN; - con->v1.connect_seq++; + if (!ceph_msgr2(from_msgr(con->msgr))) + con->v1.connect_seq++; WARN_ON(ceph_con_flag_test(con, CEPH_CON_F_WRITE_PENDING)); WARN_ON(ceph_con_flag_test(con, CEPH_CON_F_KEEPALIVE_PENDING)); } @@ -1792,9 +1794,9 @@ void ceph_msg_revoke(struct ceph_msg *msg) WARN_ON(con->state != CEPH_CON_S_OPEN); dout("%s con %p msg %p was sending\n", __func__, con, msg); if (ceph_msgr2(from_msgr(con->msgr))) - ceph_con_v2_revoke(con); + ceph_con_v2_revoke(con, msg); else - ceph_con_v1_revoke(con); + ceph_con_v1_revoke(con, msg); ceph_msg_put(con->out_msg); con->out_msg = NULL; } else { @@ -2109,11 +2111,13 @@ int ceph_con_in_msg_alloc(struct ceph_connection *con, return ret; } -void ceph_con_get_out_msg(struct ceph_connection *con) +struct ceph_msg *ceph_con_get_out_msg(struct ceph_connection *con) { struct ceph_msg *msg; - BUG_ON(list_empty(&con->out_queue)); + if (list_empty(&con->out_queue)) + return NULL; + msg = list_first_entry(&con->out_queue, struct ceph_msg, list_head); WARN_ON(msg->con != con); @@ -2140,7 +2144,7 @@ void ceph_con_get_out_msg(struct ceph_connection *con) * message or in case of a fault. */ WARN_ON(con->out_msg); - con->out_msg = ceph_msg_get(msg); + return con->out_msg = ceph_msg_get(msg); } /* diff --git a/net/ceph/messenger_v1.c b/net/ceph/messenger_v1.c index 0cb61c76b9b8..c9e002d96319 100644 --- a/net/ceph/messenger_v1.c +++ b/net/ceph/messenger_v1.c @@ -169,10 +169,9 @@ static void prepare_message_data(struct ceph_msg *msg, u32 data_len) * Prepare footer for currently outgoing message, and finish things * off. Assumes out_kvec* are already valid.. we just add on to the end. */ -static void prepare_write_message_footer(struct ceph_connection *con) +static void prepare_write_message_footer(struct ceph_connection *con, + struct ceph_msg *m) { - struct ceph_msg *m = con->out_msg; - m->footer.flags |= CEPH_MSG_FOOTER_COMPLETE; dout("prepare_write_message_footer %p\n", con); @@ -192,9 +191,9 @@ static void prepare_write_message_footer(struct ceph_connection *con) /* * Prepare headers for the next outgoing message. */ -static void prepare_write_message(struct ceph_connection *con) +static void prepare_write_message(struct ceph_connection *con, + struct ceph_msg *m) { - struct ceph_msg *m; u32 crc; con_out_kvec_reset(con); @@ -210,9 +209,6 @@ static void prepare_write_message(struct ceph_connection *con) &con->v1.out_temp_ack); } - ceph_con_get_out_msg(con); - m = con->out_msg; - dout("prepare_write_message %p seq %lld type %d len %d+%d+%zd\n", m, con->out_seq, le16_to_cpu(m->hdr.type), le32_to_cpu(m->hdr.front_len), le32_to_cpu(m->hdr.middle_len), @@ -231,31 +227,31 @@ static void prepare_write_message(struct ceph_connection *con) /* fill in hdr crc and finalize hdr */ crc = crc32c(0, &m->hdr, offsetof(struct ceph_msg_header, crc)); - con->out_msg->hdr.crc = cpu_to_le32(crc); - memcpy(&con->v1.out_hdr, &con->out_msg->hdr, sizeof(con->v1.out_hdr)); + m->hdr.crc = cpu_to_le32(crc); + memcpy(&con->v1.out_hdr, &m->hdr, sizeof(con->v1.out_hdr)); /* fill in front and middle crc, footer */ crc = crc32c(0, m->front.iov_base, m->front.iov_len); - con->out_msg->footer.front_crc = cpu_to_le32(crc); + m->footer.front_crc = cpu_to_le32(crc); if (m->middle) { crc = crc32c(0, m->middle->vec.iov_base, m->middle->vec.iov_len); - con->out_msg->footer.middle_crc = cpu_to_le32(crc); + m->footer.middle_crc = cpu_to_le32(crc); } else - con->out_msg->footer.middle_crc = 0; + m->footer.middle_crc = 0; dout("%s front_crc %u middle_crc %u\n", __func__, - le32_to_cpu(con->out_msg->footer.front_crc), - le32_to_cpu(con->out_msg->footer.middle_crc)); - con->out_msg->footer.flags = 0; + le32_to_cpu(m->footer.front_crc), + le32_to_cpu(m->footer.middle_crc)); + m->footer.flags = 0; /* is there a data payload? */ - con->out_msg->footer.data_crc = 0; + m->footer.data_crc = 0; if (m->data_length) { - prepare_message_data(con->out_msg, m->data_length); + prepare_message_data(m, m->data_length); con->v1.out_more = 1; /* data + footer will follow */ } else { /* no, queue up footer too and be done */ - prepare_write_message_footer(con); + prepare_write_message_footer(con, m); } ceph_con_flag_set(con, CEPH_CON_F_WRITE_PENDING); @@ -462,9 +458,9 @@ out: * 0 -> socket full, but more to do * <0 -> error */ -static int write_partial_message_data(struct ceph_connection *con) +static int write_partial_message_data(struct ceph_connection *con, + struct ceph_msg *msg) { - struct ceph_msg *msg = con->out_msg; struct ceph_msg_data_cursor *cursor = &msg->cursor; bool do_datacrc = !ceph_test_opt(from_msgr(con->msgr), NOCRC); u32 crc; @@ -516,7 +512,7 @@ static int write_partial_message_data(struct ceph_connection *con) else msg->footer.flags |= CEPH_MSG_FOOTER_NOCRC; con_out_kvec_reset(con); - prepare_write_message_footer(con); + prepare_write_message_footer(con, msg); return 1; /* must return > 0 to indicate success */ } @@ -1472,6 +1468,7 @@ bad_tag: */ int ceph_con_v1_try_write(struct ceph_connection *con) { + struct ceph_msg *msg; int ret = 1; dout("try_write start %p state %d\n", con, con->state); @@ -1518,14 +1515,15 @@ more: } /* msg pages? */ - if (con->out_msg) { + msg = con->out_msg; + if (msg) { if (con->v1.out_msg_done) { - ceph_msg_put(con->out_msg); + ceph_msg_put(msg); con->out_msg = NULL; /* we're done with this one */ goto do_next; } - ret = write_partial_message_data(con); + ret = write_partial_message_data(con, msg); if (ret == 1) goto more; /* we need to send the footer, too! */ if (ret == 0) @@ -1545,8 +1543,8 @@ do_next: goto more; } /* is anything else pending? */ - if (!list_empty(&con->out_queue)) { - prepare_write_message(con); + if ((msg = ceph_con_get_out_msg(con)) != NULL) { + prepare_write_message(con, msg); goto more; } if (con->in_seq > con->in_seq_acked) { @@ -1564,10 +1562,8 @@ out: return ret; } -void ceph_con_v1_revoke(struct ceph_connection *con) +void ceph_con_v1_revoke(struct ceph_connection *con, struct ceph_msg *msg) { - struct ceph_msg *msg = con->out_msg; - WARN_ON(con->v1.out_skip); /* footer */ if (con->v1.out_msg_done) { diff --git a/net/ceph/messenger_v2.c b/net/ceph/messenger_v2.c index 5483b4eed94e..833e57849c1d 100644 --- a/net/ceph/messenger_v2.c +++ b/net/ceph/messenger_v2.c @@ -709,7 +709,7 @@ static int setup_crypto(struct ceph_connection *con, dout("%s con %p con_mode %d session_key_len %d con_secret_len %d\n", __func__, con, con->v2.con_mode, session_key_len, con_secret_len); - WARN_ON(con->v2.hmac_tfm || con->v2.gcm_tfm || con->v2.gcm_req); + WARN_ON(con->v2.hmac_key_set || con->v2.gcm_tfm || con->v2.gcm_req); if (con->v2.con_mode != CEPH_CON_MODE_CRC && con->v2.con_mode != CEPH_CON_MODE_SECURE) { @@ -723,22 +723,8 @@ static int setup_crypto(struct ceph_connection *con, return 0; /* auth_none */ } - noio_flag = memalloc_noio_save(); - con->v2.hmac_tfm = crypto_alloc_shash("hmac(sha256)", 0, 0); - memalloc_noio_restore(noio_flag); - if (IS_ERR(con->v2.hmac_tfm)) { - ret = PTR_ERR(con->v2.hmac_tfm); - con->v2.hmac_tfm = NULL; - pr_err("failed to allocate hmac tfm context: %d\n", ret); - return ret; - } - - ret = crypto_shash_setkey(con->v2.hmac_tfm, session_key, - session_key_len); - if (ret) { - pr_err("failed to set hmac key: %d\n", ret); - return ret; - } + hmac_sha256_preparekey(&con->v2.hmac_key, session_key, session_key_len); + con->v2.hmac_key_set = true; if (con->v2.con_mode == CEPH_CON_MODE_CRC) { WARN_ON(con_secret_len); @@ -793,38 +779,26 @@ static int setup_crypto(struct ceph_connection *con, return 0; /* auth_x, secure mode */ } -static int ceph_hmac_sha256(struct ceph_connection *con, - const struct kvec *kvecs, int kvec_cnt, u8 *hmac) +static void ceph_hmac_sha256(struct ceph_connection *con, + const struct kvec *kvecs, int kvec_cnt, + u8 hmac[SHA256_DIGEST_SIZE]) { - SHASH_DESC_ON_STACK(desc, con->v2.hmac_tfm); /* tfm arg is ignored */ - int ret; + struct hmac_sha256_ctx ctx; int i; - dout("%s con %p hmac_tfm %p kvec_cnt %d\n", __func__, con, - con->v2.hmac_tfm, kvec_cnt); + dout("%s con %p hmac_key_set %d kvec_cnt %d\n", __func__, con, + con->v2.hmac_key_set, kvec_cnt); - if (!con->v2.hmac_tfm) { + if (!con->v2.hmac_key_set) { memset(hmac, 0, SHA256_DIGEST_SIZE); - return 0; /* auth_none */ - } - - desc->tfm = con->v2.hmac_tfm; - ret = crypto_shash_init(desc); - if (ret) - goto out; - - for (i = 0; i < kvec_cnt; i++) { - ret = crypto_shash_update(desc, kvecs[i].iov_base, - kvecs[i].iov_len); - if (ret) - goto out; + return; /* auth_none */ } - ret = crypto_shash_final(desc, hmac); - -out: - shash_desc_zero(desc); - return ret; /* auth_x, both plain and secure modes */ + /* auth_x, both plain and secure modes */ + hmac_sha256_init(&ctx, &con->v2.hmac_key); + for (i = 0; i < kvec_cnt; i++) + hmac_sha256_update(&ctx, kvecs[i].iov_base, kvecs[i].iov_len); + hmac_sha256_final(&ctx, hmac); } static void gcm_inc_nonce(struct ceph_gcm_nonce *nonce) @@ -1087,13 +1061,16 @@ static int decrypt_control_remainder(struct ceph_connection *con) static int process_v2_sparse_read(struct ceph_connection *con, struct page **pages, int spos) { - struct ceph_msg_data_cursor *cursor = &con->v2.in_cursor; + struct ceph_msg_data_cursor cursor; int ret; + ceph_msg_data_cursor_init(&cursor, con->in_msg, + con->in_msg->sparse_read_total); + for (;;) { char *buf = NULL; - ret = con->ops->sparse_read(con, cursor, &buf); + ret = con->ops->sparse_read(con, &cursor, &buf); if (ret <= 0) return ret; @@ -1111,11 +1088,11 @@ static int process_v2_sparse_read(struct ceph_connection *con, } else { struct bio_vec bv; - get_bvec_at(cursor, &bv); + get_bvec_at(&cursor, &bv); len = min_t(int, len, bv.bv_len); memcpy_page(bv.bv_page, bv.bv_offset, spage, soff, len); - ceph_msg_data_advance(cursor, len); + ceph_msg_data_advance(&cursor, len); } spos += len; ret -= len; @@ -1455,17 +1432,14 @@ static int prepare_auth_request_more(struct ceph_connection *con, static int prepare_auth_signature(struct ceph_connection *con) { void *buf; - int ret; buf = alloc_conn_buf(con, head_onwire_len(SHA256_DIGEST_SIZE, con_secure(con))); if (!buf) return -ENOMEM; - ret = ceph_hmac_sha256(con, con->v2.in_sign_kvecs, - con->v2.in_sign_kvec_cnt, CTRL_BODY(buf)); - if (ret) - return ret; + ceph_hmac_sha256(con, con->v2.in_sign_kvecs, con->v2.in_sign_kvec_cnt, + CTRL_BODY(buf)); return prepare_control(con, FRAME_TAG_AUTH_SIGNATURE, buf, SHA256_DIGEST_SIZE); @@ -1564,8 +1538,7 @@ static int prepare_keepalive2(struct ceph_connection *con) struct timespec64 now; ktime_get_real_ts64(&now); - dout("%s con %p timestamp %lld.%09ld\n", __func__, con, now.tv_sec, - now.tv_nsec); + dout("%s con %p timestamp %ptSp\n", __func__, con, &now); ceph_encode_timespec64(ts, &now); @@ -1589,10 +1562,11 @@ static int prepare_ack(struct ceph_connection *con) return prepare_control(con, FRAME_TAG_ACK, con->v2.out_buf, 8); } -static void prepare_epilogue_plain(struct ceph_connection *con, bool aborted) +static void prepare_epilogue_plain(struct ceph_connection *con, + struct ceph_msg *msg, bool aborted) { dout("%s con %p msg %p aborted %d crcs %u %u %u\n", __func__, con, - con->out_msg, aborted, con->v2.out_epil.front_crc, + msg, aborted, con->v2.out_epil.front_crc, con->v2.out_epil.middle_crc, con->v2.out_epil.data_crc); encode_epilogue_plain(con, aborted); @@ -1603,10 +1577,9 @@ static void prepare_epilogue_plain(struct ceph_connection *con, bool aborted) * For "used" empty segments, crc is -1. For unused (trailing) * segments, crc is 0. */ -static void prepare_message_plain(struct ceph_connection *con) +static void prepare_message_plain(struct ceph_connection *con, + struct ceph_msg *msg) { - struct ceph_msg *msg = con->out_msg; - prepare_head_plain(con, con->v2.out_buf, sizeof(struct ceph_msg_header2), NULL, 0, false); @@ -1647,7 +1620,7 @@ static void prepare_message_plain(struct ceph_connection *con) con->v2.out_state = OUT_S_QUEUE_DATA; } else { con->v2.out_epil.data_crc = 0; - prepare_epilogue_plain(con, false); + prepare_epilogue_plain(con, msg, false); con->v2.out_state = OUT_S_FINISH_MESSAGE; } } @@ -1659,7 +1632,8 @@ static void prepare_message_plain(struct ceph_connection *con) * allocate pages for the entire tail of the message (currently up * to ~32M) and two sgs arrays (up to ~256K each)... */ -static int prepare_message_secure(struct ceph_connection *con) +static int prepare_message_secure(struct ceph_connection *con, + struct ceph_msg *msg) { void *zerop = page_address(ceph_zero_page); struct sg_table enc_sgt = {}; @@ -1674,7 +1648,7 @@ static int prepare_message_secure(struct ceph_connection *con) if (ret) return ret; - tail_len = tail_onwire_len(con->out_msg, true); + tail_len = tail_onwire_len(msg, true); if (!tail_len) { /* * Empty message: once the head is written, @@ -1685,7 +1659,7 @@ static int prepare_message_secure(struct ceph_connection *con) } encode_epilogue_secure(con, false); - ret = setup_message_sgs(&sgt, con->out_msg, zerop, zerop, zerop, + ret = setup_message_sgs(&sgt, msg, zerop, zerop, zerop, &con->v2.out_epil, NULL, 0, false); if (ret) goto out; @@ -1714,7 +1688,7 @@ static int prepare_message_secure(struct ceph_connection *con) goto out; dout("%s con %p msg %p sg_cnt %d enc_page_cnt %d\n", __func__, con, - con->out_msg, sgt.orig_nents, enc_page_cnt); + msg, sgt.orig_nents, enc_page_cnt); con->v2.out_state = OUT_S_QUEUE_ENC_PAGE; out: @@ -1723,19 +1697,19 @@ out: return ret; } -static int prepare_message(struct ceph_connection *con) +static int prepare_message(struct ceph_connection *con, struct ceph_msg *msg) { int lens[] = { sizeof(struct ceph_msg_header2), - front_len(con->out_msg), - middle_len(con->out_msg), - data_len(con->out_msg) + front_len(msg), + middle_len(msg), + data_len(msg) }; struct ceph_frame_desc desc; int ret; dout("%s con %p msg %p logical %d+%d+%d+%d\n", __func__, con, - con->out_msg, lens[0], lens[1], lens[2], lens[3]); + msg, lens[0], lens[1], lens[2], lens[3]); if (con->in_seq > con->in_seq_acked) { dout("%s con %p in_seq_acked %llu -> %llu\n", __func__, con, @@ -1746,15 +1720,15 @@ static int prepare_message(struct ceph_connection *con) reset_out_kvecs(con); init_frame_desc(&desc, FRAME_TAG_MESSAGE, lens, 4); encode_preamble(&desc, con->v2.out_buf); - fill_header2(CTRL_BODY(con->v2.out_buf), &con->out_msg->hdr, + fill_header2(CTRL_BODY(con->v2.out_buf), &msg->hdr, con->in_seq_acked); if (con_secure(con)) { - ret = prepare_message_secure(con); + ret = prepare_message_secure(con, msg); if (ret) return ret; } else { - prepare_message_plain(con); + prepare_message_plain(con, msg); } ceph_con_flag_set(con, CEPH_CON_F_WRITE_PENDING); @@ -2460,10 +2434,8 @@ static int process_auth_signature(struct ceph_connection *con, return -EINVAL; } - ret = ceph_hmac_sha256(con, con->v2.out_sign_kvecs, - con->v2.out_sign_kvec_cnt, hmac); - if (ret) - return ret; + ceph_hmac_sha256(con, con->v2.out_sign_kvecs, con->v2.out_sign_kvec_cnt, + hmac); ceph_decode_need(&p, end, SHA256_DIGEST_SIZE, bad); if (crypto_memneq(p, hmac, SHA256_DIGEST_SIZE)) { @@ -2759,8 +2731,7 @@ static int process_keepalive2_ack(struct ceph_connection *con, ceph_decode_need(&p, end, sizeof(struct ceph_timespec), bad); ceph_decode_timespec64(&con->last_keepalive_ack, p); - dout("%s con %p timestamp %lld.%09ld\n", __func__, con, - con->last_keepalive_ack.tv_sec, con->last_keepalive_ack.tv_nsec); + dout("%s con %p timestamp %ptSp\n", __func__, con, &con->last_keepalive_ack); return 0; @@ -3184,20 +3155,20 @@ int ceph_con_v2_try_read(struct ceph_connection *con) } } -static void queue_data(struct ceph_connection *con) +static void queue_data(struct ceph_connection *con, struct ceph_msg *msg) { struct bio_vec bv; con->v2.out_epil.data_crc = -1; - ceph_msg_data_cursor_init(&con->v2.out_cursor, con->out_msg, - data_len(con->out_msg)); + ceph_msg_data_cursor_init(&con->v2.out_cursor, msg, + data_len(msg)); get_bvec_at(&con->v2.out_cursor, &bv); set_out_bvec(con, &bv, true); con->v2.out_state = OUT_S_QUEUE_DATA_CONT; } -static void queue_data_cont(struct ceph_connection *con) +static void queue_data_cont(struct ceph_connection *con, struct ceph_msg *msg) { struct bio_vec bv; @@ -3218,7 +3189,7 @@ static void queue_data_cont(struct ceph_connection *con) * we are done. */ reset_out_kvecs(con); - prepare_epilogue_plain(con, false); + prepare_epilogue_plain(con, msg, false); con->v2.out_state = OUT_S_FINISH_MESSAGE; } @@ -3250,7 +3221,7 @@ static void queue_enc_page(struct ceph_connection *con) con->v2.out_state = OUT_S_FINISH_MESSAGE; } -static void queue_zeros(struct ceph_connection *con) +static void queue_zeros(struct ceph_connection *con, struct ceph_msg *msg) { dout("%s con %p out_zero %d\n", __func__, con, con->v2.out_zero); @@ -3267,7 +3238,7 @@ static void queue_zeros(struct ceph_connection *con) * Once it's written, we are done patching up for the revoke. */ reset_out_kvecs(con); - prepare_epilogue_plain(con, true); + prepare_epilogue_plain(con, msg, true); con->v2.out_state = OUT_S_FINISH_MESSAGE; } @@ -3294,6 +3265,7 @@ static void finish_message(struct ceph_connection *con) static int populate_out_iter(struct ceph_connection *con) { + struct ceph_msg *msg; int ret; dout("%s con %p state %d out_state %d\n", __func__, con, con->state, @@ -3309,18 +3281,18 @@ static int populate_out_iter(struct ceph_connection *con) switch (con->v2.out_state) { case OUT_S_QUEUE_DATA: WARN_ON(!con->out_msg); - queue_data(con); + queue_data(con, con->out_msg); goto populated; case OUT_S_QUEUE_DATA_CONT: WARN_ON(!con->out_msg); - queue_data_cont(con); + queue_data_cont(con, con->out_msg); goto populated; case OUT_S_QUEUE_ENC_PAGE: queue_enc_page(con); goto populated; case OUT_S_QUEUE_ZEROS: WARN_ON(con->out_msg); /* revoked */ - queue_zeros(con); + queue_zeros(con, con->out_msg); goto populated; case OUT_S_FINISH_MESSAGE: finish_message(con); @@ -3339,9 +3311,8 @@ static int populate_out_iter(struct ceph_connection *con) pr_err("prepare_keepalive2 failed: %d\n", ret); return ret; } - } else if (!list_empty(&con->out_queue)) { - ceph_con_get_out_msg(con); - ret = prepare_message(con); + } else if ((msg = ceph_con_get_out_msg(con)) != NULL) { + ret = prepare_message(con, msg); if (ret) { pr_err("prepare_message failed: %d\n", ret); return ret; @@ -3453,17 +3424,18 @@ static u32 crc32c_zeros(u32 crc, int zero_len) return crc; } -static void prepare_zero_front(struct ceph_connection *con, int resid) +static void prepare_zero_front(struct ceph_connection *con, + struct ceph_msg *msg, int resid) { int sent; - WARN_ON(!resid || resid > front_len(con->out_msg)); - sent = front_len(con->out_msg) - resid; + WARN_ON(!resid || resid > front_len(msg)); + sent = front_len(msg) - resid; dout("%s con %p sent %d resid %d\n", __func__, con, sent, resid); if (sent) { con->v2.out_epil.front_crc = - crc32c(-1, con->out_msg->front.iov_base, sent); + crc32c(-1, msg->front.iov_base, sent); con->v2.out_epil.front_crc = crc32c_zeros(con->v2.out_epil.front_crc, resid); } else { @@ -3474,17 +3446,18 @@ static void prepare_zero_front(struct ceph_connection *con, int resid) out_zero_add(con, resid); } -static void prepare_zero_middle(struct ceph_connection *con, int resid) +static void prepare_zero_middle(struct ceph_connection *con, + struct ceph_msg *msg, int resid) { int sent; - WARN_ON(!resid || resid > middle_len(con->out_msg)); - sent = middle_len(con->out_msg) - resid; + WARN_ON(!resid || resid > middle_len(msg)); + sent = middle_len(msg) - resid; dout("%s con %p sent %d resid %d\n", __func__, con, sent, resid); if (sent) { con->v2.out_epil.middle_crc = - crc32c(-1, con->out_msg->middle->vec.iov_base, sent); + crc32c(-1, msg->middle->vec.iov_base, sent); con->v2.out_epil.middle_crc = crc32c_zeros(con->v2.out_epil.middle_crc, resid); } else { @@ -3495,61 +3468,64 @@ static void prepare_zero_middle(struct ceph_connection *con, int resid) out_zero_add(con, resid); } -static void prepare_zero_data(struct ceph_connection *con) +static void prepare_zero_data(struct ceph_connection *con, + struct ceph_msg *msg) { dout("%s con %p\n", __func__, con); - con->v2.out_epil.data_crc = crc32c_zeros(-1, data_len(con->out_msg)); - out_zero_add(con, data_len(con->out_msg)); + con->v2.out_epil.data_crc = crc32c_zeros(-1, data_len(msg)); + out_zero_add(con, data_len(msg)); } -static void revoke_at_queue_data(struct ceph_connection *con) +static void revoke_at_queue_data(struct ceph_connection *con, + struct ceph_msg *msg) { int boundary; int resid; - WARN_ON(!data_len(con->out_msg)); + WARN_ON(!data_len(msg)); WARN_ON(!iov_iter_is_kvec(&con->v2.out_iter)); resid = iov_iter_count(&con->v2.out_iter); - boundary = front_len(con->out_msg) + middle_len(con->out_msg); + boundary = front_len(msg) + middle_len(msg); if (resid > boundary) { resid -= boundary; WARN_ON(resid > MESSAGE_HEAD_PLAIN_LEN); dout("%s con %p was sending head\n", __func__, con); - if (front_len(con->out_msg)) - prepare_zero_front(con, front_len(con->out_msg)); - if (middle_len(con->out_msg)) - prepare_zero_middle(con, middle_len(con->out_msg)); - prepare_zero_data(con); + if (front_len(msg)) + prepare_zero_front(con, msg, front_len(msg)); + if (middle_len(msg)) + prepare_zero_middle(con, msg, middle_len(msg)); + prepare_zero_data(con, msg); WARN_ON(iov_iter_count(&con->v2.out_iter) != resid); con->v2.out_state = OUT_S_QUEUE_ZEROS; return; } - boundary = middle_len(con->out_msg); + boundary = middle_len(msg); if (resid > boundary) { resid -= boundary; dout("%s con %p was sending front\n", __func__, con); - prepare_zero_front(con, resid); - if (middle_len(con->out_msg)) - prepare_zero_middle(con, middle_len(con->out_msg)); - prepare_zero_data(con); - queue_zeros(con); + prepare_zero_front(con, msg, resid); + if (middle_len(msg)) + prepare_zero_middle(con, msg, middle_len(msg)); + prepare_zero_data(con, msg); + queue_zeros(con, msg); return; } WARN_ON(!resid); dout("%s con %p was sending middle\n", __func__, con); - prepare_zero_middle(con, resid); - prepare_zero_data(con); - queue_zeros(con); + prepare_zero_middle(con, msg, resid); + prepare_zero_data(con, msg); + queue_zeros(con, msg); } -static void revoke_at_queue_data_cont(struct ceph_connection *con) +static void revoke_at_queue_data_cont(struct ceph_connection *con, + struct ceph_msg *msg) { int sent, resid; /* current piece of data */ - WARN_ON(!data_len(con->out_msg)); + WARN_ON(!data_len(msg)); WARN_ON(!iov_iter_is_bvec(&con->v2.out_iter)); resid = iov_iter_count(&con->v2.out_iter); WARN_ON(!resid || resid > con->v2.out_bvec.bv_len); @@ -3568,10 +3544,11 @@ static void revoke_at_queue_data_cont(struct ceph_connection *con) con->v2.out_iter.count -= resid; out_zero_add(con, con->v2.out_cursor.total_resid); - queue_zeros(con); + queue_zeros(con, msg); } -static void revoke_at_finish_message(struct ceph_connection *con) +static void revoke_at_finish_message(struct ceph_connection *con, + struct ceph_msg *msg) { int boundary; int resid; @@ -3579,39 +3556,39 @@ static void revoke_at_finish_message(struct ceph_connection *con) WARN_ON(!iov_iter_is_kvec(&con->v2.out_iter)); resid = iov_iter_count(&con->v2.out_iter); - if (!front_len(con->out_msg) && !middle_len(con->out_msg) && - !data_len(con->out_msg)) { + if (!front_len(msg) && !middle_len(msg) && + !data_len(msg)) { WARN_ON(!resid || resid > MESSAGE_HEAD_PLAIN_LEN); dout("%s con %p was sending head (empty message) - noop\n", __func__, con); return; } - boundary = front_len(con->out_msg) + middle_len(con->out_msg) + + boundary = front_len(msg) + middle_len(msg) + CEPH_EPILOGUE_PLAIN_LEN; if (resid > boundary) { resid -= boundary; WARN_ON(resid > MESSAGE_HEAD_PLAIN_LEN); dout("%s con %p was sending head\n", __func__, con); - if (front_len(con->out_msg)) - prepare_zero_front(con, front_len(con->out_msg)); - if (middle_len(con->out_msg)) - prepare_zero_middle(con, middle_len(con->out_msg)); + if (front_len(msg)) + prepare_zero_front(con, msg, front_len(msg)); + if (middle_len(msg)) + prepare_zero_middle(con, msg, middle_len(msg)); con->v2.out_iter.count -= CEPH_EPILOGUE_PLAIN_LEN; WARN_ON(iov_iter_count(&con->v2.out_iter) != resid); con->v2.out_state = OUT_S_QUEUE_ZEROS; return; } - boundary = middle_len(con->out_msg) + CEPH_EPILOGUE_PLAIN_LEN; + boundary = middle_len(msg) + CEPH_EPILOGUE_PLAIN_LEN; if (resid > boundary) { resid -= boundary; dout("%s con %p was sending front\n", __func__, con); - prepare_zero_front(con, resid); - if (middle_len(con->out_msg)) - prepare_zero_middle(con, middle_len(con->out_msg)); + prepare_zero_front(con, msg, resid); + if (middle_len(msg)) + prepare_zero_middle(con, msg, middle_len(msg)); con->v2.out_iter.count -= CEPH_EPILOGUE_PLAIN_LEN; - queue_zeros(con); + queue_zeros(con, msg); return; } @@ -3619,9 +3596,9 @@ static void revoke_at_finish_message(struct ceph_connection *con) if (resid > boundary) { resid -= boundary; dout("%s con %p was sending middle\n", __func__, con); - prepare_zero_middle(con, resid); + prepare_zero_middle(con, msg, resid); con->v2.out_iter.count -= CEPH_EPILOGUE_PLAIN_LEN; - queue_zeros(con); + queue_zeros(con, msg); return; } @@ -3629,7 +3606,7 @@ static void revoke_at_finish_message(struct ceph_connection *con) dout("%s con %p was sending epilogue - noop\n", __func__, con); } -void ceph_con_v2_revoke(struct ceph_connection *con) +void ceph_con_v2_revoke(struct ceph_connection *con, struct ceph_msg *msg) { WARN_ON(con->v2.out_zero); @@ -3642,13 +3619,13 @@ void ceph_con_v2_revoke(struct ceph_connection *con) switch (con->v2.out_state) { case OUT_S_QUEUE_DATA: - revoke_at_queue_data(con); + revoke_at_queue_data(con, msg); break; case OUT_S_QUEUE_DATA_CONT: - revoke_at_queue_data_cont(con); + revoke_at_queue_data_cont(con, msg); break; case OUT_S_FINISH_MESSAGE: - revoke_at_finish_message(con); + revoke_at_finish_message(con, msg); break; default: WARN(1, "bad out_state %d", con->v2.out_state); @@ -3814,10 +3791,8 @@ void ceph_con_v2_reset_protocol(struct ceph_connection *con) memzero_explicit(&con->v2.in_gcm_nonce, CEPH_GCM_IV_LEN); memzero_explicit(&con->v2.out_gcm_nonce, CEPH_GCM_IV_LEN); - if (con->v2.hmac_tfm) { - crypto_free_shash(con->v2.hmac_tfm); - con->v2.hmac_tfm = NULL; - } + memzero_explicit(&con->v2.hmac_key, sizeof(con->v2.hmac_key)); + con->v2.hmac_key_set = false; if (con->v2.gcm_req) { aead_request_free(con->v2.gcm_req); con->v2.gcm_req = NULL; diff --git a/net/ceph/mon_client.c b/net/ceph/mon_client.c index ab66b599ac47..c227ececa925 100644 --- a/net/ceph/mon_client.c +++ b/net/ceph/mon_client.c @@ -314,7 +314,7 @@ static void __schedule_delayed(struct ceph_mon_client *monc) delay = CEPH_MONC_PING_INTERVAL; dout("__schedule_delayed after %lu\n", delay); - mod_delayed_work(system_wq, &monc->delayed_work, + mod_delayed_work(system_percpu_wq, &monc->delayed_work, round_jiffies_relative(delay)); } diff --git a/net/ceph/osdmap.c b/net/ceph/osdmap.c index 295098873861..d245fa508e1c 100644 --- a/net/ceph/osdmap.c +++ b/net/ceph/osdmap.c @@ -1504,8 +1504,6 @@ static int decode_new_primary_temp(void **p, void *end, u32 ceph_get_primary_affinity(struct ceph_osdmap *map, int osd) { - BUG_ON(osd >= map->max_osd); - if (!map->osd_primary_affinity) return CEPH_OSD_DEFAULT_PRIMARY_AFFINITY; @@ -1514,8 +1512,6 @@ u32 ceph_get_primary_affinity(struct ceph_osdmap *map, int osd) static int set_primary_affinity(struct ceph_osdmap *map, int osd, u32 aff) { - BUG_ON(osd >= map->max_osd); - if (!map->osd_primary_affinity) { int i; @@ -1577,6 +1573,8 @@ static int decode_new_primary_affinity(void **p, void *end, ceph_decode_32_safe(p, end, osd, e_inval); ceph_decode_32_safe(p, end, aff, e_inval); + if (osd >= map->max_osd) + goto e_inval; ret = set_primary_affinity(map, osd, aff); if (ret) @@ -1879,7 +1877,9 @@ static int decode_new_up_state_weight(void **p, void *end, u8 struct_v, ceph_decode_need(p, end, 2*sizeof(u32), e_inval); osd = ceph_decode_32(p); w = ceph_decode_32(p); - BUG_ON(osd >= map->max_osd); + if (osd >= map->max_osd) + goto e_inval; + osdmap_info(map, "osd%d weight 0x%x %s\n", osd, w, w == CEPH_OSD_IN ? "(in)" : (w == CEPH_OSD_OUT ? "(out)" : "")); @@ -1905,13 +1905,15 @@ static int decode_new_up_state_weight(void **p, void *end, u8 struct_v, u32 xorstate; osd = ceph_decode_32(p); + if (osd >= map->max_osd) + goto e_inval; + if (struct_v >= 5) xorstate = ceph_decode_32(p); else xorstate = ceph_decode_8(p); if (xorstate == 0) xorstate = CEPH_OSD_UP; - BUG_ON(osd >= map->max_osd); if ((map->osd_state[osd] & CEPH_OSD_UP) && (xorstate & CEPH_OSD_UP)) osdmap_info(map, "osd%d down\n", osd); @@ -1937,7 +1939,9 @@ static int decode_new_up_state_weight(void **p, void *end, u8 struct_v, struct ceph_entity_addr addr; osd = ceph_decode_32(p); - BUG_ON(osd >= map->max_osd); + if (osd >= map->max_osd) + goto e_inval; + if (struct_v >= 7) ret = ceph_decode_entity_addrvec(p, end, msgr2, &addr); else |