From patchwork Mon Jan 6 18:58:15 2020 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Gabriel Krisman Bertazi X-Patchwork-Id: 11319895 Return-Path: Received: from mail.kernel.org (pdx-korg-mail-1.web.codeaurora.org [172.30.200.123]) by pdx-korg-patchwork-2.web.codeaurora.org (Postfix) with ESMTP id B15DA14BD for ; Mon, 6 Jan 2020 18:58:37 +0000 (UTC) Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by mail.kernel.org (Postfix) with ESMTP id 9A3B72072A for ; Mon, 6 Jan 2020 18:58:35 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1726797AbgAFS6c (ORCPT ); Mon, 6 Jan 2020 13:58:32 -0500 Received: from bhuna.collabora.co.uk ([46.235.227.227]:35916 "EHLO bhuna.collabora.co.uk" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1726657AbgAFS6b (ORCPT ); Mon, 6 Jan 2020 13:58:31 -0500 Received: from localhost (unknown [IPv6:2610:98:8005::147]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) (Authenticated sender: krisman) by bhuna.collabora.co.uk (Postfix) with ESMTPSA id 09F3029196A; Mon, 6 Jan 2020 18:58:28 +0000 (GMT) From: Gabriel Krisman Bertazi To: gregkh@linuxfoundation.org Cc: rafael@kernel.org, lduncan@suse.com, cleech@redhat.com, jejb@linux.ibm.com, martin.petersen@oracle.com, open-iscsi@googlegroups.com, linux-kernel@vger.kernel.org, linux-scsi@vger.kernel.org, Gabriel Krisman Bertazi , kernel@collabora.com Subject: [PATCH 1/3] drivers: base: Support atomic version of attribute_container_device_trigger Date: Mon, 6 Jan 2020 13:58:15 -0500 Message-Id: <20200106185817.640331-2-krisman@collabora.com> X-Mailer: git-send-email 2.24.1 In-Reply-To: <20200106185817.640331-1-krisman@collabora.com> References: <20200106185817.640331-1-krisman@collabora.com> MIME-Version: 1.0 Sender: linux-scsi-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-scsi@vger.kernel.org attribute_container_device_trigger invokes callbacks that may fail for one or more classdev's, for instance, the transport_add_class_device callback, called during transport creation, does memory allocation. This information, though, is not propagated to upper layers, and any driver using the attribute_container_device_trigger API will not know whether any, some, or all callbacks succeeded. This patch implements a safe version of this dispatcher, to either succeed all the callbacks or revert to the original state. Signed-off-by: Gabriel Krisman Bertazi Reviewed-by: Greg Kroah-Hartman --- drivers/base/attribute_container.c | 103 ++++++++++++++++++++++++++++ include/linux/attribute_container.h | 7 ++ 2 files changed, 110 insertions(+) diff --git a/drivers/base/attribute_container.c b/drivers/base/attribute_container.c index 20736aaa0e69..f7bd0f4db13d 100644 --- a/drivers/base/attribute_container.c +++ b/drivers/base/attribute_container.c @@ -236,6 +236,109 @@ attribute_container_remove_device(struct device *dev, mutex_unlock(&attribute_container_mutex); } +static int +do_attribute_container_device_trigger_safe(struct device *dev, + struct attribute_container *cont, + int (*fn)(struct attribute_container *, + struct device *, struct device *), + int (*undo)(struct attribute_container *, + struct device *, struct device *)) +{ + int ret; + struct internal_container *ic, *failed; + struct klist_iter iter; + + if (attribute_container_no_classdevs(cont)) + return fn(cont, dev, NULL); + + klist_for_each_entry(ic, &cont->containers, node, &iter) { + if (dev == ic->classdev.parent) { + ret = fn(cont, dev, &ic->classdev); + if (ret) { + failed = ic; + klist_iter_exit(&iter); + goto fail; + } + } + } + return 0; + +fail: + if (!undo) + return ret; + + /* Attempt to undo the work partially done. */ + klist_for_each_entry(ic, &cont->containers, node, &iter) { + if (ic == failed) { + klist_iter_exit(&iter); + break; + } + if (dev == ic->classdev.parent) + undo(cont, dev, &ic->classdev); + } + return ret; +} + +/** + * attribute_container_device_trigger_safe - execute a trigger for each + * matching classdev or fail all of them. + * + * @dev: The generic device to run the trigger for + * @fn the function to execute for each classdev. + * @undo A function to undo the work previously done in case of error + * + * This function is a safe version of + * attribute_container_device_trigger. It stops on the first error and + * undo the partial work that has been done, on previous classdev. It + * is guaranteed that either they all succeeded, or none of them + * succeeded. + */ +int +attribute_container_device_trigger_safe(struct device *dev, + int (*fn)(struct attribute_container *, + struct device *, + struct device *), + int (*undo)(struct attribute_container *, + struct device *, + struct device *)) +{ + struct attribute_container *cont, *failed = NULL; + int ret = 0; + + mutex_lock(&attribute_container_mutex); + + list_for_each_entry(cont, &attribute_container_list, node) { + + if (!cont->match(cont, dev)) + continue; + + ret = do_attribute_container_device_trigger_safe(dev, cont, + fn, undo); + if (ret) { + failed = cont; + break; + } + } + + if (ret && !WARN_ON(!undo)) { + list_for_each_entry(cont, &attribute_container_list, node) { + + if (failed == cont) + break; + + if (!cont->match(cont, dev)) + continue; + + do_attribute_container_device_trigger_safe(dev, cont, + undo, NULL); + } + } + + mutex_unlock(&attribute_container_mutex); + return ret; + +} + /** * attribute_container_device_trigger - execute a trigger for each matching classdev * diff --git a/include/linux/attribute_container.h b/include/linux/attribute_container.h index d12bb2153cd6..e4004d1e6725 100644 --- a/include/linux/attribute_container.h +++ b/include/linux/attribute_container.h @@ -54,6 +54,13 @@ void attribute_container_device_trigger(struct device *dev, int (*fn)(struct attribute_container *, struct device *, struct device *)); +int attribute_container_device_trigger_safe(struct device *dev, + int (*fn)(struct attribute_container *, + struct device *, + struct device *), + int (*undo)(struct attribute_container *, + struct device *, + struct device *)); void attribute_container_trigger(struct device *dev, int (*fn)(struct attribute_container *, struct device *)); From patchwork Mon Jan 6 18:58:16 2020 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Gabriel Krisman Bertazi X-Patchwork-Id: 11319899 Return-Path: Received: from mail.kernel.org (pdx-korg-mail-1.web.codeaurora.org [172.30.200.123]) by pdx-korg-patchwork-2.web.codeaurora.org (Postfix) with ESMTP id 141271398 for ; Mon, 6 Jan 2020 18:58:44 +0000 (UTC) Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by mail.kernel.org (Postfix) with ESMTP id F1823214D8 for ; Mon, 6 Jan 2020 18:58:43 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1726893AbgAFS6g (ORCPT ); Mon, 6 Jan 2020 13:58:36 -0500 Received: from bhuna.collabora.co.uk ([46.235.227.227]:35924 "EHLO bhuna.collabora.co.uk" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1726657AbgAFS6e (ORCPT ); Mon, 6 Jan 2020 13:58:34 -0500 Received: from localhost (unknown [IPv6:2610:98:8005::147]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) (Authenticated sender: krisman) by bhuna.collabora.co.uk (Postfix) with ESMTPSA id 6AD4429166F; Mon, 6 Jan 2020 18:58:32 +0000 (GMT) From: Gabriel Krisman Bertazi To: gregkh@linuxfoundation.org Cc: rafael@kernel.org, lduncan@suse.com, cleech@redhat.com, jejb@linux.ibm.com, martin.petersen@oracle.com, open-iscsi@googlegroups.com, linux-kernel@vger.kernel.org, linux-scsi@vger.kernel.org, Gabriel Krisman Bertazi , kernel@collabora.com Subject: [PATCH 2/3] drivers: base: Propagate errors through the transport component Date: Mon, 6 Jan 2020 13:58:16 -0500 Message-Id: <20200106185817.640331-3-krisman@collabora.com> X-Mailer: git-send-email 2.24.1 In-Reply-To: <20200106185817.640331-1-krisman@collabora.com> References: <20200106185817.640331-1-krisman@collabora.com> MIME-Version: 1.0 Sender: linux-scsi-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-scsi@vger.kernel.org The transport registration may fail. Make sure the errors are propagated to the callers. Signed-off-by: Gabriel Krisman Bertazi Reviewed-by: Greg Kroah-Hartman --- drivers/base/transport_class.c | 11 ++++++++--- include/linux/transport_class.h | 6 +++--- 2 files changed, 11 insertions(+), 6 deletions(-) diff --git a/drivers/base/transport_class.c b/drivers/base/transport_class.c index 5ed86ded6e6b..ccc86206e508 100644 --- a/drivers/base/transport_class.c +++ b/drivers/base/transport_class.c @@ -30,6 +30,10 @@ #include #include +static int transport_remove_classdev(struct attribute_container *cont, + struct device *dev, + struct device *classdev); + /** * transport_class_register - register an initial transport class * @@ -172,10 +176,11 @@ static int transport_add_class_device(struct attribute_container *cont, * routine is simply a trigger point used to add the device to the * system and register attributes for it. */ - -void transport_add_device(struct device *dev) +int transport_add_device(struct device *dev) { - attribute_container_device_trigger(dev, transport_add_class_device); + return attribute_container_device_trigger_safe(dev, + transport_add_class_device, + transport_remove_classdev); } EXPORT_SYMBOL_GPL(transport_add_device); diff --git a/include/linux/transport_class.h b/include/linux/transport_class.h index a9c59761927b..63076fb835e3 100644 --- a/include/linux/transport_class.h +++ b/include/linux/transport_class.h @@ -62,16 +62,16 @@ struct transport_container { container_of(x, struct transport_container, ac) void transport_remove_device(struct device *); -void transport_add_device(struct device *); +int transport_add_device(struct device *); void transport_setup_device(struct device *); void transport_configure_device(struct device *); void transport_destroy_device(struct device *); -static inline void +static inline int transport_register_device(struct device *dev) { transport_setup_device(dev); - transport_add_device(dev); + return transport_add_device(dev); } static inline void From patchwork Mon Jan 6 18:58:17 2020 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Gabriel Krisman Bertazi X-Patchwork-Id: 11319897 Return-Path: Received: from mail.kernel.org (pdx-korg-mail-1.web.codeaurora.org [172.30.200.123]) by pdx-korg-patchwork-2.web.codeaurora.org (Postfix) with ESMTP id A9A8214BD for ; Mon, 6 Jan 2020 18:58:38 +0000 (UTC) Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by mail.kernel.org (Postfix) with ESMTP id 8C78F2072C for ; Mon, 6 Jan 2020 18:58:38 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1726937AbgAFS6h (ORCPT ); Mon, 6 Jan 2020 13:58:37 -0500 Received: from bhuna.collabora.co.uk ([46.235.227.227]:35948 "EHLO bhuna.collabora.co.uk" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1726769AbgAFS6h (ORCPT ); Mon, 6 Jan 2020 13:58:37 -0500 Received: from localhost (unknown [IPv6:2610:98:8005::147]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) (Authenticated sender: krisman) by bhuna.collabora.co.uk (Postfix) with ESMTPSA id 5375929196F; Mon, 6 Jan 2020 18:58:35 +0000 (GMT) From: Gabriel Krisman Bertazi To: gregkh@linuxfoundation.org Cc: rafael@kernel.org, lduncan@suse.com, cleech@redhat.com, jejb@linux.ibm.com, martin.petersen@oracle.com, open-iscsi@googlegroups.com, linux-kernel@vger.kernel.org, linux-scsi@vger.kernel.org, Gabriel Krisman Bertazi , kernel@collabora.com Subject: [PATCH 3/3] iscsi: Fail session and connection on transport registration failure Date: Mon, 6 Jan 2020 13:58:17 -0500 Message-Id: <20200106185817.640331-4-krisman@collabora.com> X-Mailer: git-send-email 2.24.1 In-Reply-To: <20200106185817.640331-1-krisman@collabora.com> References: <20200106185817.640331-1-krisman@collabora.com> MIME-Version: 1.0 Sender: linux-scsi-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-scsi@vger.kernel.org If the transport cannot be registered, the session/connection creation needs to be failed early to let the initiator know. Otherwise, the system will have an outstanding connection that cannot be used nor removed by open-iscsi. The result is similar to the error below, triggered by injecting a failure in the transport's registration path. openiscsi reports success: root@debian-vm:~# iscsiadm -m node -T iqn:lun1 -p 127.0.0.1 -l Logging in to [iface: default, target: iqn:lun1, portal: 127.0.0.1,3260] Login to [iface: default, target: iqn:lun1, portal:127.0.0.1,3260] successful. But cannot remove the session afterwards, since the kernel is in an inconsistent state. root@debian-vm:~# iscsiadm -m node -T iqn:lun1 -p 127.0.0.1 -u iscsiadm: No matching sessions found Signed-off-by: Gabriel Krisman Bertazi --- drivers/scsi/scsi_transport_iscsi.c | 18 ++++++++++++++++-- 1 file changed, 16 insertions(+), 2 deletions(-) diff --git a/drivers/scsi/scsi_transport_iscsi.c b/drivers/scsi/scsi_transport_iscsi.c index ed8d9709b9b9..2b732b7a9a5b 100644 --- a/drivers/scsi/scsi_transport_iscsi.c +++ b/drivers/scsi/scsi_transport_iscsi.c @@ -2093,7 +2093,12 @@ int iscsi_add_session(struct iscsi_cls_session *session, unsigned int target_id) "could not register session's dev\n"); goto release_ida; } - transport_register_device(&session->dev); + err = transport_register_device(&session->dev); + if (err) { + iscsi_cls_session_printk(KERN_ERR, session, + "could not register transport's dev\n"); + goto release_dev; + } spin_lock_irqsave(&sesslock, flags); list_add(&session->sess_list, &sesslist); @@ -2103,6 +2108,8 @@ int iscsi_add_session(struct iscsi_cls_session *session, unsigned int target_id) ISCSI_DBG_TRANS_SESSION(session, "Completed session adding\n"); return 0; +release_dev: + device_del(&session->dev); release_ida: if (session->ida_used) ida_simple_remove(&iscsi_sess_ida, session->target_id); @@ -2263,7 +2270,12 @@ iscsi_create_conn(struct iscsi_cls_session *session, int dd_size, uint32_t cid) "register connection's dev\n"); goto release_parent_ref; } - transport_register_device(&conn->dev); + err = transport_register_device(&conn->dev); + if (err) { + iscsi_cls_session_printk(KERN_ERR, session, "could not " + "register transport's dev\n"); + goto release_conn_ref; + } spin_lock_irqsave(&connlock, flags); list_add(&conn->conn_list, &connlist); @@ -2272,6 +2284,8 @@ iscsi_create_conn(struct iscsi_cls_session *session, int dd_size, uint32_t cid) ISCSI_DBG_TRANS_CONN(conn, "Completed conn creation\n"); return conn; +release_conn_ref: + put_device(&conn->dev); release_parent_ref: put_device(&session->dev); free_conn: