From nobody Wed Oct 29 17:30:52 2025 Delivered-To: importer@patchew.org Received-SPF: pass (zoho.com: domain of gnu.org designates 208.118.235.17 as permitted sender) client-ip=208.118.235.17; envelope-from=qemu-devel-bounces+importer=patchew.org@nongnu.org; helo=lists.gnu.org; Authentication-Results: mx.zohomail.com; dkim=fail; spf=pass (zoho.com: domain of gnu.org designates 208.118.235.17 as permitted sender) smtp.mailfrom=qemu-devel-bounces+importer=patchew.org@nongnu.org; dmarc=fail(p=none dis=none) header.from=gmail.com Return-Path: Received: from lists.gnu.org (lists.gnu.org [208.118.235.17]) by mx.zohomail.com with SMTPS id 15246671516771013.0899999395801; Wed, 25 Apr 2018 07:39:11 -0700 (PDT) Received: from localhost ([::1]:37310 helo=lists.gnu.org) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1fBLZe-0001v5-KY for importer@patchew.org; Wed, 25 Apr 2018 10:39:10 -0400 Received: from eggs.gnu.org ([2001:4830:134:3::10]:33392) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1fBLXe-0000gc-GB for qemu-devel@nongnu.org; Wed, 25 Apr 2018 10:37:11 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1fBLXc-0008AJ-LS for qemu-devel@nongnu.org; Wed, 25 Apr 2018 10:37:06 -0400 Received: from mail-pf0-x242.google.com ([2607:f8b0:400e:c00::242]:38370) by eggs.gnu.org with esmtps (TLS1.0:RSA_AES_128_CBC_SHA1:16) (Exim 4.71) (envelope-from ) id 1fBLXc-0008A5-CN for qemu-devel@nongnu.org; Wed, 25 Apr 2018 10:37:04 -0400 Received: by mail-pf0-x242.google.com with SMTP id o76so10688172pfi.5 for ; Wed, 25 Apr 2018 07:37:04 -0700 (PDT) Received: from VM_127_91_centos.localdomain ([119.28.195.137]) by smtp.gmail.com with ESMTPSA id y29sm18244004pff.42.2018.04.25.07.37.01 (version=TLS1_2 cipher=ECDHE-RSA-AES128-GCM-SHA256 bits=128/128); Wed, 25 Apr 2018 07:37:02 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20161025; h=from:to:cc:subject:date:message-id:in-reply-to:references; bh=Ql+dRZCtZHYCN2tS0PM886LoJ19LT7BTN2dHi55o/Nw=; b=fo1WFuj1b1K81GqjkHBKl8TWPgQxxgaNhNqxdIDFekpJufsLwWD2FBZ2uB33S1BeFu qQq7P7+poTKTz7Jme5GCO2+uYWKN/TVh/DDJYpsS3u359gPXI45+0iFWEYS2gwKTYq7s ITeyJRvASNT+W0XHRMkAOUSyJRBcon6jg9nYt0IUwPonfA0qO7EEoC6jTAO5xmyyOx6+ Eb2O05ytbw1J7pXcfWKkS0VhEvfxFZ8a8Z5Yj7ynMIf3D2rovmAQXMV6cO3iu/gzRdPt 89KGPTeYGMNN9pevo9rHsyTp6WH/qxi9Gr+GlVHgBmUzCY7Ymebvfzc6oxB4OZBIPnbs Msog== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20161025; h=x-gm-message-state:from:to:cc:subject:date:message-id:in-reply-to :references; bh=Ql+dRZCtZHYCN2tS0PM886LoJ19LT7BTN2dHi55o/Nw=; b=txXuWux7ZNa+ATTU9ttOLXbLJtUNN3ZuWu9rnNbuMrkTuN6XXSUxolv0Bu7ss7K3T4 1jhQqcogAAd7/fXtsC1MPjyYU9uOMikRYnv2N8ZQdSw7TKWeZco+lkYcIP0WM1W3PtKN inzacGGUD+KnSOLd08vmxsLiQ4jQ0P6W5ziJUzQ2vDlEimr4y1lEDMTP7Vz/9+zmgm60 dnuzNiUeKJTJDmr9s/MzjzpPjZzM1WsT+FYfXtP2I8K585XgKRH6w9SCAXTZNW6tWvnE gjDkaRrBO3ZRMbf/wDsHiBg4TU1zC/QXIVoA0zwtmDSaQwiZk9FCxdAtxB/3NQHcnLXU uFoA== X-Gm-Message-State: ALQs6tB6P9d9HcHF86jZ/vcuUPGp9FCvcHnNtPBkkEl+OXDVuNqQ9aY+ L9xLh1WeC8adDYyFKvqwzC4= X-Google-Smtp-Source: AIpwx4/jsijUXxvjQqT3I00b326Uru9F+yybXmUhzDeSOJm+zfZdX9Dv5FP+jLzvUiMNLxeAFzlArw== X-Received: by 10.99.163.9 with SMTP id s9mr23650539pge.187.1524667023381; Wed, 25 Apr 2018 07:37:03 -0700 (PDT) From: Lidong Chen X-Google-Original-From: Lidong Chen To: quintela@redhat.com, dgilbert@redhat.com Date: Wed, 25 Apr 2018 22:35:33 +0800 Message-Id: <1524666934-8064-5-git-send-email-lidongchen@tencent.com> X-Mailer: git-send-email 1.8.3.1 In-Reply-To: <1524666934-8064-1-git-send-email-lidongchen@tencent.com> References: <1524666934-8064-1-git-send-email-lidongchen@tencent.com> X-detected-operating-system: by eggs.gnu.org: Genre and OS details not recognized. X-Received-From: 2607:f8b0:400e:c00::242 Subject: [Qemu-devel] [PATCH v2 4/5] migration: implement bi-directional RDMA QIOChannel X-BeenThere: qemu-devel@nongnu.org X-Mailman-Version: 2.1.21 Precedence: list List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Cc: galsha@mellanox.com, adido@mellanox.com, aviadye@mellanox.com, qemu-devel@nongnu.org, licq@mellanox.com, Lidong Chen Errors-To: qemu-devel-bounces+importer=patchew.org@nongnu.org Sender: "Qemu-devel" X-ZohoMail-DKIM: fail (Header signature does not verify) X-ZohoMail: RDKM_2 RSF_0 Z_629925259 SPT_0 Content-Transfer-Encoding: quoted-printable MIME-Version: 1.0 Content-Type: text/plain; charset="utf-8" This patch implements bi-directional RDMA QIOChannel. Because different threads may access RDMAQIOChannel concurrently, this patch use RCU to prote= ct it. Signed-off-by: Lidong Chen --- migration/rdma.c | 162 +++++++++++++++++++++++++++++++++++++++++++++++++--= ---- 1 file changed, 146 insertions(+), 16 deletions(-) diff --git a/migration/rdma.c b/migration/rdma.c index f5c1d02..0652224 100644 --- a/migration/rdma.c +++ b/migration/rdma.c @@ -86,6 +86,7 @@ static uint32_t known_capabilities =3D RDMA_CAPABILITY_PI= N_ALL; " to abort!"); \ rdma->error_reported =3D 1; \ } \ + rcu_read_unlock(); \ return rdma->error_state; \ } \ } while (0) @@ -405,6 +406,7 @@ struct QIOChannelRDMA { RDMAContext *rdma; QEMUFile *file; bool blocking; /* XXX we don't actually honour this yet */ + QemuMutex lock; }; =20 /* @@ -2635,12 +2637,29 @@ static ssize_t qio_channel_rdma_writev(QIOChannel *= ioc, { QIOChannelRDMA *rioc =3D QIO_CHANNEL_RDMA(ioc); QEMUFile *f =3D rioc->file; - RDMAContext *rdma =3D rioc->rdma; + RDMAContext *rdma; int ret; ssize_t done =3D 0; size_t i; size_t len =3D 0; =20 + rcu_read_lock(); + rdma =3D atomic_rcu_read(&rioc->rdma); + + if (!rdma) { + rcu_read_unlock(); + return -EIO; + } + + if (rdma->listen_id) { + rdma =3D rdma->return_path; + } + + if (!rdma) { + rcu_read_unlock(); + return -EIO; + } + CHECK_ERROR_STATE(); =20 /* @@ -2650,6 +2669,7 @@ static ssize_t qio_channel_rdma_writev(QIOChannel *io= c, ret =3D qemu_rdma_write_flush(f, rdma); if (ret < 0) { rdma->error_state =3D ret; + rcu_read_unlock(); return ret; } =20 @@ -2669,6 +2689,7 @@ static ssize_t qio_channel_rdma_writev(QIOChannel *io= c, =20 if (ret < 0) { rdma->error_state =3D ret; + rcu_read_unlock(); return ret; } =20 @@ -2677,6 +2698,7 @@ static ssize_t qio_channel_rdma_writev(QIOChannel *io= c, } } =20 + rcu_read_unlock(); return done; } =20 @@ -2710,12 +2732,29 @@ static ssize_t qio_channel_rdma_readv(QIOChannel *i= oc, Error **errp) { QIOChannelRDMA *rioc =3D QIO_CHANNEL_RDMA(ioc); - RDMAContext *rdma =3D rioc->rdma; + RDMAContext *rdma; RDMAControlHeader head; int ret =3D 0; ssize_t i; size_t done =3D 0; =20 + rcu_read_lock(); + rdma =3D atomic_rcu_read(&rioc->rdma); + + if (!rdma) { + rcu_read_unlock(); + return -EIO; + } + + if (!rdma->listen_id) { + rdma =3D rdma->return_path; + } + + if (!rdma) { + rcu_read_unlock(); + return -EIO; + } + CHECK_ERROR_STATE(); =20 for (i =3D 0; i < niov; i++) { @@ -2727,7 +2766,7 @@ static ssize_t qio_channel_rdma_readv(QIOChannel *ioc, * were given and dish out the bytes until we run * out of bytes. */ - ret =3D qemu_rdma_fill(rioc->rdma, data, want, 0); + ret =3D qemu_rdma_fill(rdma, data, want, 0); done +=3D ret; want -=3D ret; /* Got what we needed, so go to next iovec */ @@ -2749,25 +2788,28 @@ static ssize_t qio_channel_rdma_readv(QIOChannel *i= oc, =20 if (ret < 0) { rdma->error_state =3D ret; + rcu_read_unlock(); return ret; } =20 /* * SEND was received with new bytes, now try again. */ - ret =3D qemu_rdma_fill(rioc->rdma, data, want, 0); + ret =3D qemu_rdma_fill(rdma, data, want, 0); done +=3D ret; want -=3D ret; =20 /* Still didn't get enough, so lets just return */ if (want) { if (done =3D=3D 0) { + rcu_read_unlock(); return QIO_CHANNEL_ERR_BLOCK; } else { break; } } } + rcu_read_unlock(); return done; } =20 @@ -2823,6 +2865,16 @@ qio_channel_rdma_source_prepare(GSource *source, GIOCondition cond =3D 0; *timeout =3D -1; =20 + if ((rdma->listen_id && rsource->condition =3D=3D G_IO_OUT) || + (!rdma->listen_id && rsource->condition =3D=3D G_IO_IN)) { + rdma =3D rdma->return_path; + } + + if (!rdma) { + error_report("RDMAContext is NULL when prepare Gsource"); + return FALSE; + } + if (rdma->wr_data[0].control_len) { cond |=3D G_IO_IN; } @@ -2838,6 +2890,16 @@ qio_channel_rdma_source_check(GSource *source) RDMAContext *rdma =3D rsource->rioc->rdma; GIOCondition cond =3D 0; =20 + if ((rdma->listen_id && rsource->condition =3D=3D G_IO_OUT) || + (!rdma->listen_id && rsource->condition =3D=3D G_IO_IN)) { + rdma =3D rdma->return_path; + } + + if (!rdma) { + error_report("RDMAContext is NULL when check Gsource"); + return FALSE; + } + if (rdma->wr_data[0].control_len) { cond |=3D G_IO_IN; } @@ -2856,6 +2918,16 @@ qio_channel_rdma_source_dispatch(GSource *source, RDMAContext *rdma =3D rsource->rioc->rdma; GIOCondition cond =3D 0; =20 + if ((rdma->listen_id && rsource->condition =3D=3D G_IO_OUT) || + (!rdma->listen_id && rsource->condition =3D=3D G_IO_IN)) { + rdma =3D rdma->return_path; + } + + if (!rdma) { + error_report("RDMAContext is NULL when dispatch Gsource"); + return FALSE; + } + if (rdma->wr_data[0].control_len) { cond |=3D G_IO_IN; } @@ -2905,15 +2977,29 @@ static int qio_channel_rdma_close(QIOChannel *ioc, Error **errp) { QIOChannelRDMA *rioc =3D QIO_CHANNEL_RDMA(ioc); + RDMAContext *rdma; trace_qemu_rdma_close(); - if (rioc->rdma) { - if (!rioc->rdma->error_state) { - rioc->rdma->error_state =3D qemu_file_get_error(rioc->file); - } - qemu_rdma_cleanup(rioc->rdma); - g_free(rioc->rdma); - rioc->rdma =3D NULL; + + qemu_mutex_lock(&rioc->lock); + rdma =3D rioc->rdma; + if (!rdma) { + qemu_mutex_unlock(&rioc->lock); + return 0; + } + atomic_rcu_set(&rioc->rdma, NULL); + qemu_mutex_unlock(&rioc->lock); + + if (!rdma->error_state) { + rdma->error_state =3D qemu_file_get_error(rioc->file); + } + qemu_rdma_cleanup(rdma); + + if (rdma->return_path) { + qemu_rdma_cleanup(rdma->return_path); + g_free(rdma->return_path); } + + g_free(rdma); return 0; } =20 @@ -2956,12 +3042,21 @@ static size_t qemu_rdma_save_page(QEMUFile *f, void= *opaque, size_t size, uint64_t *bytes_sent) { QIOChannelRDMA *rioc =3D QIO_CHANNEL_RDMA(opaque); - RDMAContext *rdma =3D rioc->rdma; + RDMAContext *rdma; int ret; =20 + rcu_read_lock(); + rdma =3D atomic_rcu_read(&rioc->rdma); + + if (!rdma) { + rcu_read_unlock(); + return -EIO; + } + CHECK_ERROR_STATE(); =20 if (migrate_get_current()->state =3D=3D MIGRATION_STATUS_POSTCOPY_ACTI= VE) { + rcu_read_unlock(); return RAM_SAVE_CONTROL_NOT_SUPP; } =20 @@ -3046,9 +3141,11 @@ static size_t qemu_rdma_save_page(QEMUFile *f, void = *opaque, } } =20 + rcu_read_unlock(); return RAM_SAVE_CONTROL_DELAYED; err: rdma->error_state =3D ret; + rcu_read_unlock(); return ret; } =20 @@ -3224,8 +3321,8 @@ static int qemu_rdma_registration_handle(QEMUFile *f,= void *opaque) RDMAControlHeader blocks =3D { .type =3D RDMA_CONTROL_RAM_BLOCKS_RESUL= T, .repeat =3D 1 }; QIOChannelRDMA *rioc =3D QIO_CHANNEL_RDMA(opaque); - RDMAContext *rdma =3D rioc->rdma; - RDMALocalBlocks *local =3D &rdma->local_ram_blocks; + RDMAContext *rdma; + RDMALocalBlocks *local; RDMAControlHeader head; RDMARegister *reg, *registers; RDMACompress *comp; @@ -3238,8 +3335,17 @@ static int qemu_rdma_registration_handle(QEMUFile *f= , void *opaque) int count =3D 0; int i =3D 0; =20 + rcu_read_lock(); + rdma =3D atomic_rcu_read(&rioc->rdma); + + if (!rdma) { + rcu_read_unlock(); + return -EIO; + } + CHECK_ERROR_STATE(); =20 + local =3D &rdma->local_ram_blocks; do { trace_qemu_rdma_registration_handle_wait(); =20 @@ -3469,6 +3575,7 @@ out: if (ret < 0) { rdma->error_state =3D ret; } + rcu_read_unlock(); return ret; } =20 @@ -3525,11 +3632,19 @@ static int qemu_rdma_registration_start(QEMUFile *f= , void *opaque, uint64_t flags, void *data) { QIOChannelRDMA *rioc =3D QIO_CHANNEL_RDMA(opaque); - RDMAContext *rdma =3D rioc->rdma; + RDMAContext *rdma; + + rcu_read_lock(); + rdma =3D atomic_rcu_read(&rioc->rdma); + if (!rdma) { + rcu_read_unlock(); + return -EIO; + } =20 CHECK_ERROR_STATE(); =20 if (migrate_get_current()->state =3D=3D MIGRATION_STATUS_POSTCOPY_ACTI= VE) { + rcu_read_unlock(); return 0; } =20 @@ -3537,6 +3652,7 @@ static int qemu_rdma_registration_start(QEMUFile *f, = void *opaque, qemu_put_be64(f, RAM_SAVE_FLAG_HOOK); qemu_fflush(f); =20 + rcu_read_unlock(); return 0; } =20 @@ -3549,13 +3665,21 @@ static int qemu_rdma_registration_stop(QEMUFile *f,= void *opaque, { Error *local_err =3D NULL, **errp =3D &local_err; QIOChannelRDMA *rioc =3D QIO_CHANNEL_RDMA(opaque); - RDMAContext *rdma =3D rioc->rdma; + RDMAContext *rdma; RDMAControlHeader head =3D { .len =3D 0, .repeat =3D 1 }; int ret =3D 0; =20 + rcu_read_lock(); + rdma =3D atomic_rcu_read(&rioc->rdma); + if (!rdma) { + rcu_read_unlock(); + return -EIO; + } + CHECK_ERROR_STATE(); =20 if (migrate_get_current()->state =3D=3D MIGRATION_STATUS_POSTCOPY_ACTI= VE) { + rcu_read_unlock(); return 0; } =20 @@ -3587,6 +3711,7 @@ static int qemu_rdma_registration_stop(QEMUFile *f, v= oid *opaque, qemu_rdma_reg_whole_ram_blocks : NULL); if (ret < 0) { ERROR(errp, "receiving remote info!"); + rcu_read_unlock(); return ret; } =20 @@ -3610,6 +3735,7 @@ static int qemu_rdma_registration_stop(QEMUFile *f, v= oid *opaque, "not identical on both the source and destination.= ", local->nb_blocks, nb_dest_blocks); rdma->error_state =3D -EINVAL; + rcu_read_unlock(); return -EINVAL; } =20 @@ -3626,6 +3752,7 @@ static int qemu_rdma_registration_stop(QEMUFile *f, v= oid *opaque, local->block[i].length, rdma->dest_blocks[i].length); rdma->error_state =3D -EINVAL; + rcu_read_unlock(); return -EINVAL; } local->block[i].remote_host_addr =3D @@ -3643,9 +3770,11 @@ static int qemu_rdma_registration_stop(QEMUFile *f, = void *opaque, goto err; } =20 + rcu_read_unlock(); return 0; err: rdma->error_state =3D ret; + rcu_read_unlock(); return ret; } =20 @@ -3707,6 +3836,7 @@ static QEMUFile *qemu_fopen_rdma(RDMAContext *rdma, c= onst char *mode) =20 rioc =3D QIO_CHANNEL_RDMA(object_new(TYPE_QIO_CHANNEL_RDMA)); rioc->rdma =3D rdma; + qemu_mutex_init(&rioc->lock); =20 if (mode[0] =3D=3D 'w') { rioc->file =3D qemu_fopen_channel_output(QIO_CHANNEL(rioc)); --=20 1.8.3.1