From patchwork Tue Apr 5 19:19:37 2022 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Jeff Layton X-Patchwork-Id: 12802360 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by smtp.lore.kernel.org (Postfix) with ESMTP id 1049EC433F5 for ; Wed, 6 Apr 2022 04:16:42 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1445687AbiDFEOs (ORCPT ); Wed, 6 Apr 2022 00:14:48 -0400 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:35496 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1573554AbiDETWj (ORCPT ); Tue, 5 Apr 2022 15:22:39 -0400 Received: from ams.source.kernel.org (ams.source.kernel.org [145.40.68.75]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id 632473E0DB; Tue, 5 Apr 2022 12:20:40 -0700 (PDT) Received: from smtp.kernel.org (relay.kernel.org [52.25.139.140]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by ams.source.kernel.org (Postfix) with ESMTPS id 0AB56B81F6B; Tue, 5 Apr 2022 19:20:39 +0000 (UTC) Received: by smtp.kernel.org (Postfix) with ESMTPSA id 0503FC385A1; Tue, 5 Apr 2022 19:20:36 +0000 (UTC) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=kernel.org; s=k20201202; t=1649186437; bh=gFp0s2+yOcXsxdx+8FwxESNn106vn66cOvnAIMvKYsk=; h=From:To:Cc:Subject:Date:In-Reply-To:References:From; b=VsI1yGO5xbYAMRP+i5E4Cubrv3Faw4rfQwY+423rZ6IyILXWi/UVtF9IXBaZ74/cY YZEUqMyQk4EMV9V3S3jSi2IAM7DqaPSbNcDfW0Ejr/z/0dJdfQAQ0ixi+kvNt514kE bgAahZ/++1hP+6TyivNQ8X4bEpOxYz4r8ptcfSeE+kHI56pwrnkiLWYe1MXkJcO6Xu VSlJPqyYzNm+VMd+4EQX7NUfEiruPnZ+46ahMdmpGqjGjqmaCijz55i2ea1VlGKnjF jyTCJPP6/RZac+VTbaqI8Kv149wFgV6jpJ8trlqdSOYGunrAt5aD4ANDjKiu8jyDVw Q3NtEK2o+UCZg== From: Jeff Layton To: idryomov@gmail.com, xiubli@redhat.com Cc: ceph-devel@vger.kernel.org, linux-fsdevel@vger.kernel.org, linux-fscrypt@vger.kernel.org, linux-kernel@vger.kernel.org, lhenriques@suse.de Subject: [PATCH v13 06/59] libceph: add sparse read support to msgr1 Date: Tue, 5 Apr 2022 15:19:37 -0400 Message-Id: <20220405192030.178326-7-jlayton@kernel.org> X-Mailer: git-send-email 2.35.1 In-Reply-To: <20220405192030.178326-1-jlayton@kernel.org> References: <20220405192030.178326-1-jlayton@kernel.org> MIME-Version: 1.0 Precedence: bulk List-ID: X-Mailing-List: ceph-devel@vger.kernel.org Add 2 new fields to ceph_connection_v1_info to track the necessary info in sparse reads. Skip initializing the cursor for a sparse read. Break out read_partial_message_section into a wrapper around a new read_partial_message_chunk function that doesn't zero out the crc first. Add new helper functions to drive receiving into the destinations provided by the sparse_read state machine. Signed-off-by: Jeff Layton --- include/linux/ceph/messenger.h | 4 ++ net/ceph/messenger_v1.c | 98 +++++++++++++++++++++++++++++++--- 2 files changed, 94 insertions(+), 8 deletions(-) diff --git a/include/linux/ceph/messenger.h b/include/linux/ceph/messenger.h index 7f09a4213834..f4adbfee56d5 100644 --- a/include/linux/ceph/messenger.h +++ b/include/linux/ceph/messenger.h @@ -337,6 +337,10 @@ struct ceph_connection_v1_info { int in_base_pos; /* bytes read */ + /* sparse reads */ + struct kvec in_sr_kvec; /* current location to receive into */ + u64 in_sr_len; /* amount of data in this extent */ + /* message in temps */ u8 in_tag; /* protocol control byte */ struct ceph_msg_header in_hdr; diff --git a/net/ceph/messenger_v1.c b/net/ceph/messenger_v1.c index 6b014eca3a13..bf385e458a01 100644 --- a/net/ceph/messenger_v1.c +++ b/net/ceph/messenger_v1.c @@ -160,9 +160,9 @@ static size_t sizeof_footer(struct ceph_connection *con) static void prepare_message_data(struct ceph_msg *msg, u32 data_len) { - /* Initialize data cursor */ - - ceph_msg_data_cursor_init(&msg->cursor, msg, data_len); + /* Initialize data cursor if it's not a sparse read */ + if (!msg->sparse_read) + ceph_msg_data_cursor_init(&msg->cursor, msg, data_len); } /* @@ -967,9 +967,9 @@ static void process_ack(struct ceph_connection *con) prepare_read_tag(con); } -static int read_partial_message_section(struct ceph_connection *con, - struct kvec *section, - unsigned int sec_len, u32 *crc) +static int read_partial_message_chunk(struct ceph_connection *con, + struct kvec *section, + unsigned int sec_len, u32 *crc) { int ret, left; @@ -985,11 +985,91 @@ static int read_partial_message_section(struct ceph_connection *con, section->iov_len += ret; } if (section->iov_len == sec_len) - *crc = crc32c(0, section->iov_base, section->iov_len); + *crc = crc32c(*crc, section->iov_base, section->iov_len); return 1; } +static inline int read_partial_message_section(struct ceph_connection *con, + struct kvec *section, + unsigned int sec_len, u32 *crc) +{ + *crc = 0; + return read_partial_message_chunk(con, section, sec_len, crc); +} + +static int read_sparse_msg_extent(struct ceph_connection *con, u32 *crc) +{ + struct ceph_msg_data_cursor *cursor = &con->in_msg->cursor; + bool do_bounce = ceph_test_opt(from_msgr(con->msgr), RXBOUNCE); + + if (do_bounce && unlikely(!con->bounce_page)) { + con->bounce_page = alloc_page(GFP_NOIO); + if (!con->bounce_page) { + pr_err("failed to allocate bounce page\n"); + return -ENOMEM; + } + } + + while (cursor->sr_resid > 0) { + struct page *page, *rpage; + size_t off, len; + int ret; + + page = ceph_msg_data_next(cursor, &off, &len, NULL); + rpage = do_bounce ? con->bounce_page : page; + + /* clamp to what remains in extent */ + len = min_t(int, len, cursor->sr_resid); + ret = ceph_tcp_recvpage(con->sock, rpage, (int)off, len); + if (ret <= 0) + return ret; + *crc = ceph_crc32c_page(*crc, rpage, off, ret); + ceph_msg_data_advance(cursor, (size_t)ret); + cursor->sr_resid -= ret; + if (do_bounce) + memcpy_page(page, off, rpage, off, ret); + } + return 1; +} + +static int read_sparse_msg_data(struct ceph_connection *con) +{ + struct ceph_msg_data_cursor *cursor = &con->in_msg->cursor; + bool do_datacrc = !ceph_test_opt(from_msgr(con->msgr), NOCRC); + u32 crc = 0; + int ret = 1; + + if (do_datacrc) + crc = con->in_data_crc; + + do { + if (con->v1.in_sr_kvec.iov_base) + ret = read_partial_message_chunk(con, + &con->v1.in_sr_kvec, + con->v1.in_sr_len, + &crc); + else if (cursor->sr_resid > 0) + ret = read_sparse_msg_extent(con, &crc); + + if (ret <= 0) { + if (do_datacrc) + con->in_data_crc = crc; + return ret; + } + + memset(&con->v1.in_sr_kvec, 0, sizeof(con->v1.in_sr_kvec)); + ret = con->ops->sparse_read(con, cursor, + (char **)&con->v1.in_sr_kvec.iov_base); + con->v1.in_sr_len = ret; + } while (ret > 0); + + if (do_datacrc) + con->in_data_crc = crc; + + return ret < 0 ? ret : 1; /* must return > 0 to indicate success */ +} + static int read_partial_msg_data(struct ceph_connection *con) { struct ceph_msg_data_cursor *cursor = &con->in_msg->cursor; @@ -1180,7 +1260,9 @@ static int read_partial_message(struct ceph_connection *con) if (!m->num_data_items) return -EIO; - if (ceph_test_opt(from_msgr(con->msgr), RXBOUNCE)) + if (m->sparse_read) + ret = read_sparse_msg_data(con); + else if (ceph_test_opt(from_msgr(con->msgr), RXBOUNCE)) ret = read_partial_msg_data_bounce(con); else ret = read_partial_msg_data(con);