From patchwork Wed Sep 27 11:45:13 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 8bit X-Patchwork-Submitter: David Marchand X-Patchwork-Id: 132000 X-Patchwork-Delegate: thomas@monjalon.net Return-Path: X-Original-To: patchwork@inbox.dpdk.org Delivered-To: patchwork@inbox.dpdk.org Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by inbox.dpdk.org (Postfix) with ESMTP id 3006842651; Wed, 27 Sep 2023 13:45:30 +0200 (CEST) Received: from mails.dpdk.org (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id 1ECAE402E4; Wed, 27 Sep 2023 13:45:30 +0200 (CEST) Received: from us-smtp-delivery-124.mimecast.com (us-smtp-delivery-124.mimecast.com [170.10.133.124]) by mails.dpdk.org (Postfix) with ESMTP id 4C72B402D3 for ; Wed, 27 Sep 2023 13:45:28 +0200 (CEST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=redhat.com; s=mimecast20190719; t=1695815127; h=from:from:reply-to:subject:subject:date:date:message-id:message-id: to:to:cc:cc:mime-version:mime-version:content-type:content-type: content-transfer-encoding:content-transfer-encoding: in-reply-to:in-reply-to:references:references; bh=9y5MYAl0N+RzfQbjuRsYDe6ZD7mY6o9mRHixPU7pHt8=; b=UGsK7PKB+7IZdf5ROrJF8oKkteV9ejuxrmqLbf0xPs24hM5aM++C1PidhYaTKl4NiOnMoV rSb3w+Q5OnCCfWBwsVRLLP5oYHeTTaW1UpHG7X2K+B+LycIk1FZQKwDwofGHyHX25JlkOD LrNCW+u3V01SEXDVWzGuEhJMNtHuAdU= Received: from mimecast-mx02.redhat.com (mx-ext.redhat.com [66.187.233.73]) by relay.mimecast.com with ESMTP with STARTTLS (version=TLSv1.2, cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id us-mta-628-FJgvy9zPPf622JPXJF2yvg-1; Wed, 27 Sep 2023 07:45:22 -0400 X-MC-Unique: FJgvy9zPPf622JPXJF2yvg-1 Received: from smtp.corp.redhat.com (int-mx10.intmail.prod.int.rdu2.redhat.com [10.11.54.10]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by mimecast-mx02.redhat.com (Postfix) with ESMTPS id 04908280FEC5; Wed, 27 Sep 2023 11:45:22 +0000 (UTC) Received: from dmarchan.redhat.com (unknown [10.45.224.48]) by smtp.corp.redhat.com (Postfix) with ESMTP id 450F1492C37; Wed, 27 Sep 2023 11:45:20 +0000 (UTC) From: David Marchand To: dev@dpdk.org Cc: probb@iol.unh.edu, =?utf-8?q?Morten_Br=C3=B8rup?= , Thomas Monjalon , Ferruh Yigit , Andrew Rybchenko Subject: [PATCH v4 1/3] ethdev: protect shared memory accesses under one lock Date: Wed, 27 Sep 2023 13:45:13 +0200 Message-ID: <20230927114515.1245213-2-david.marchand@redhat.com> In-Reply-To: <20230927114515.1245213-1-david.marchand@redhat.com> References: <20230818091321.2404089-1-david.marchand@redhat.com> <20230927114515.1245213-1-david.marchand@redhat.com> MIME-Version: 1.0 X-Scanned-By: MIMEDefang 3.1 on 10.11.54.10 X-Mimecast-Spam-Score: 0 X-Mimecast-Originator: redhat.com X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org ethdev currently uses two locks to protect access around eth_dev_shared_data: - one (process local) for avoiding multiple threads to reserve/lookup the eth_dev_shared_data memzone, - one (shared with other processes) for protecting port allocation/destruction, A next change will make it possible for eth_dev_shared_data memzone to be freed during a DPDK application lifetime. Accessing its content must then be protected for concurrent accesses and with this new protection, existing locks become unneeded. Simplify the logic and put everything under a single lock in DPDK shared memory config (which cannot be freed during a DPDK application lifetime). Signed-off-by: David Marchand Acked-by: Morten Brørup --- Changes since v3: - updated commitlog, --- lib/eal/common/eal_common_mcfg.c | 6 +++ lib/eal/common/eal_memcfg.h | 1 + lib/eal/include/rte_eal_memconfig.h | 4 ++ lib/eal/version.map | 1 + lib/ethdev/ethdev_driver.c | 30 ++++++++------- lib/ethdev/ethdev_private.c | 9 ----- lib/ethdev/ethdev_private.h | 9 +++-- lib/ethdev/rte_ethdev.c | 60 ++++++++++++++++------------- 8 files changed, 68 insertions(+), 52 deletions(-) diff --git a/lib/eal/common/eal_common_mcfg.c b/lib/eal/common/eal_common_mcfg.c index b60d41f7b6..2a785e74c4 100644 --- a/lib/eal/common/eal_common_mcfg.c +++ b/lib/eal/common/eal_common_mcfg.c @@ -177,6 +177,12 @@ rte_mcfg_timer_unlock(void) rte_spinlock_unlock(rte_mcfg_timer_get_lock()); } +rte_spinlock_t * +rte_mcfg_ethdev_get_lock(void) +{ + return &rte_eal_get_configuration()->mem_config->ethdev_lock; +} + bool rte_mcfg_get_single_file_segments(void) { diff --git a/lib/eal/common/eal_memcfg.h b/lib/eal/common/eal_memcfg.h index 8889ba063f..d5c63e2f4d 100644 --- a/lib/eal/common/eal_memcfg.h +++ b/lib/eal/common/eal_memcfg.h @@ -37,6 +37,7 @@ struct rte_mem_config { rte_rwlock_t qlock; /**< used by tailqs for thread safety. */ rte_rwlock_t mplock; /**< used by mempool library for thread safety. */ rte_spinlock_t tlock; /**< used by timer library for thread safety. */ + rte_spinlock_t ethdev_lock; /**< used by ethdev library. */ rte_rwlock_t memory_hotplug_lock; /**< Indicates whether memory hotplug request is in progress. */ diff --git a/lib/eal/include/rte_eal_memconfig.h b/lib/eal/include/rte_eal_memconfig.h index c527f9aa29..0b1d0d4ff0 100644 --- a/lib/eal/include/rte_eal_memconfig.h +++ b/lib/eal/include/rte_eal_memconfig.h @@ -39,6 +39,10 @@ __rte_internal rte_spinlock_t * rte_mcfg_timer_get_lock(void); +__rte_internal +rte_spinlock_t * +rte_mcfg_ethdev_get_lock(void); + /** * Lock the internal EAL shared memory configuration for shared access. */ diff --git a/lib/eal/version.map b/lib/eal/version.map index 915057b325..e00a844805 100644 --- a/lib/eal/version.map +++ b/lib/eal/version.map @@ -452,6 +452,7 @@ INTERNAL { rte_intr_vec_list_free; rte_intr_vec_list_index_get; rte_intr_vec_list_index_set; + rte_mcfg_ethdev_get_lock; rte_mcfg_mem_get_lock; rte_mcfg_mempool_get_lock; rte_mcfg_tailq_get_lock; diff --git a/lib/ethdev/ethdev_driver.c b/lib/ethdev/ethdev_driver.c index 30db839a77..c92cd4b947 100644 --- a/lib/ethdev/ethdev_driver.c +++ b/lib/ethdev/ethdev_driver.c @@ -45,6 +45,7 @@ eth_dev_allocated(const char *name) static uint16_t eth_dev_find_free_port(void) + __rte_exclusive_locks_required(rte_mcfg_ethdev_get_lock()) { uint16_t i; @@ -61,6 +62,7 @@ eth_dev_find_free_port(void) static struct rte_eth_dev * eth_dev_get(uint16_t port_id) + __rte_exclusive_locks_required(rte_mcfg_ethdev_get_lock()) { struct rte_eth_dev *eth_dev = &rte_eth_devices[port_id]; @@ -87,10 +89,10 @@ rte_eth_dev_allocate(const char *name) return NULL; } - eth_dev_shared_data_prepare(); + /* Synchronize port creation between primary and secondary processes. */ + rte_spinlock_lock(rte_mcfg_ethdev_get_lock()); - /* Synchronize port creation between primary and secondary threads. */ - rte_spinlock_lock(ð_dev_shared_data->ownership_lock); + eth_dev_shared_data_prepare(); if (eth_dev_allocated(name) != NULL) { RTE_ETHDEV_LOG(ERR, @@ -114,7 +116,7 @@ rte_eth_dev_allocate(const char *name) pthread_mutex_init(ð_dev->data->flow_ops_mutex, NULL); unlock: - rte_spinlock_unlock(ð_dev_shared_data->ownership_lock); + rte_spinlock_unlock(rte_mcfg_ethdev_get_lock()); return eth_dev; } @@ -124,13 +126,13 @@ rte_eth_dev_allocated(const char *name) { struct rte_eth_dev *ethdev; - eth_dev_shared_data_prepare(); + rte_spinlock_lock(rte_mcfg_ethdev_get_lock()); - rte_spinlock_lock(ð_dev_shared_data->ownership_lock); + eth_dev_shared_data_prepare(); ethdev = eth_dev_allocated(name); - rte_spinlock_unlock(ð_dev_shared_data->ownership_lock); + rte_spinlock_unlock(rte_mcfg_ethdev_get_lock()); return ethdev; } @@ -146,10 +148,10 @@ rte_eth_dev_attach_secondary(const char *name) uint16_t i; struct rte_eth_dev *eth_dev = NULL; - eth_dev_shared_data_prepare(); - /* Synchronize port attachment to primary port creation and release. */ - rte_spinlock_lock(ð_dev_shared_data->ownership_lock); + rte_spinlock_lock(rte_mcfg_ethdev_get_lock()); + + eth_dev_shared_data_prepare(); for (i = 0; i < RTE_MAX_ETHPORTS; i++) { if (strcmp(eth_dev_shared_data->data[i].name, name) == 0) @@ -164,7 +166,7 @@ rte_eth_dev_attach_secondary(const char *name) RTE_ASSERT(eth_dev->data->port_id == i); } - rte_spinlock_unlock(ð_dev_shared_data->ownership_lock); + rte_spinlock_unlock(rte_mcfg_ethdev_get_lock()); return eth_dev; } @@ -220,7 +222,9 @@ rte_eth_dev_release_port(struct rte_eth_dev *eth_dev) if (eth_dev == NULL) return -EINVAL; + rte_spinlock_lock(rte_mcfg_ethdev_get_lock()); eth_dev_shared_data_prepare(); + rte_spinlock_unlock(rte_mcfg_ethdev_get_lock()); if (eth_dev->state != RTE_ETH_DEV_UNUSED) rte_eth_dev_callback_process(eth_dev, @@ -228,7 +232,7 @@ rte_eth_dev_release_port(struct rte_eth_dev *eth_dev) eth_dev_fp_ops_reset(rte_eth_fp_ops + eth_dev->data->port_id); - rte_spinlock_lock(ð_dev_shared_data->ownership_lock); + rte_spinlock_lock(rte_mcfg_ethdev_get_lock()); eth_dev->state = RTE_ETH_DEV_UNUSED; eth_dev->device = NULL; @@ -252,7 +256,7 @@ rte_eth_dev_release_port(struct rte_eth_dev *eth_dev) memset(eth_dev->data, 0, sizeof(struct rte_eth_dev_data)); } - rte_spinlock_unlock(ð_dev_shared_data->ownership_lock); + rte_spinlock_unlock(rte_mcfg_ethdev_get_lock()); return 0; } diff --git a/lib/ethdev/ethdev_private.c b/lib/ethdev/ethdev_private.c index 14ec8c6ccf..6756625729 100644 --- a/lib/ethdev/ethdev_private.c +++ b/lib/ethdev/ethdev_private.c @@ -11,12 +11,8 @@ static const char *MZ_RTE_ETH_DEV_DATA = "rte_eth_dev_data"; -/* Shared memory between primary and secondary processes. */ struct eth_dev_shared *eth_dev_shared_data; -/* spinlock for shared data allocation */ -static rte_spinlock_t eth_dev_shared_data_lock = RTE_SPINLOCK_INITIALIZER; - /* spinlock for eth device callbacks */ rte_spinlock_t eth_dev_cb_lock = RTE_SPINLOCK_INITIALIZER; @@ -328,8 +324,6 @@ eth_dev_shared_data_prepare(void) const unsigned int flags = 0; const struct rte_memzone *mz; - rte_spinlock_lock(ð_dev_shared_data_lock); - if (eth_dev_shared_data == NULL) { if (rte_eal_process_type() == RTE_PROC_PRIMARY) { /* Allocate port data and ownership shared memory. */ @@ -345,13 +339,10 @@ eth_dev_shared_data_prepare(void) if (rte_eal_process_type() == RTE_PROC_PRIMARY) { eth_dev_shared_data->next_owner_id = RTE_ETH_DEV_NO_OWNER + 1; - rte_spinlock_init(ð_dev_shared_data->ownership_lock); memset(eth_dev_shared_data->data, 0, sizeof(eth_dev_shared_data->data)); } } - - rte_spinlock_unlock(ð_dev_shared_data_lock); } void diff --git a/lib/ethdev/ethdev_private.h b/lib/ethdev/ethdev_private.h index acb4b335c8..f7706e6a95 100644 --- a/lib/ethdev/ethdev_private.h +++ b/lib/ethdev/ethdev_private.h @@ -7,6 +7,7 @@ #include +#include #include #include @@ -14,11 +15,12 @@ struct eth_dev_shared { uint64_t next_owner_id; - rte_spinlock_t ownership_lock; struct rte_eth_dev_data data[RTE_MAX_ETHPORTS]; }; -extern struct eth_dev_shared *eth_dev_shared_data; +/* Shared memory between primary and secondary processes. */ +extern struct eth_dev_shared *eth_dev_shared_data + __rte_guarded_by(rte_mcfg_ethdev_get_lock()); /** * The user application callback description. @@ -65,7 +67,8 @@ void eth_dev_fp_ops_setup(struct rte_eth_fp_ops *fpo, const struct rte_eth_dev *dev); -void eth_dev_shared_data_prepare(void); +void eth_dev_shared_data_prepare(void) + __rte_exclusive_locks_required(rte_mcfg_ethdev_get_lock()); void eth_dev_rxq_release(struct rte_eth_dev *dev, uint16_t qid); void eth_dev_txq_release(struct rte_eth_dev *dev, uint16_t qid); diff --git a/lib/ethdev/rte_ethdev.c b/lib/ethdev/rte_ethdev.c index 46eaed6467..e0d22c27c6 100644 --- a/lib/ethdev/rte_ethdev.c +++ b/lib/ethdev/rte_ethdev.c @@ -409,6 +409,7 @@ rte_eth_dev_is_valid_port(uint16_t port_id) static int eth_is_valid_owner_id(uint64_t owner_id) + __rte_exclusive_locks_required(rte_mcfg_ethdev_get_lock()) { if (owner_id == RTE_ETH_DEV_NO_OWNER || eth_dev_shared_data->next_owner_id <= owner_id) @@ -437,13 +438,12 @@ rte_eth_dev_owner_new(uint64_t *owner_id) return -EINVAL; } - eth_dev_shared_data_prepare(); - - rte_spinlock_lock(ð_dev_shared_data->ownership_lock); + rte_spinlock_lock(rte_mcfg_ethdev_get_lock()); + eth_dev_shared_data_prepare(); *owner_id = eth_dev_shared_data->next_owner_id++; - rte_spinlock_unlock(ð_dev_shared_data->ownership_lock); + rte_spinlock_unlock(rte_mcfg_ethdev_get_lock()); rte_ethdev_trace_owner_new(*owner_id); @@ -453,6 +453,7 @@ rte_eth_dev_owner_new(uint64_t *owner_id) static int eth_dev_owner_set(const uint16_t port_id, const uint64_t old_owner_id, const struct rte_eth_dev_owner *new_owner) + __rte_exclusive_locks_required(rte_mcfg_ethdev_get_lock()) { struct rte_eth_dev *ethdev = &rte_eth_devices[port_id]; struct rte_eth_dev_owner *port_owner; @@ -503,13 +504,12 @@ rte_eth_dev_owner_set(const uint16_t port_id, { int ret; - eth_dev_shared_data_prepare(); - - rte_spinlock_lock(ð_dev_shared_data->ownership_lock); + rte_spinlock_lock(rte_mcfg_ethdev_get_lock()); + eth_dev_shared_data_prepare(); ret = eth_dev_owner_set(port_id, RTE_ETH_DEV_NO_OWNER, owner); - rte_spinlock_unlock(ð_dev_shared_data->ownership_lock); + rte_spinlock_unlock(rte_mcfg_ethdev_get_lock()); rte_ethdev_trace_owner_set(port_id, owner, ret); @@ -523,13 +523,12 @@ rte_eth_dev_owner_unset(const uint16_t port_id, const uint64_t owner_id) {.id = RTE_ETH_DEV_NO_OWNER, .name = ""}; int ret; - eth_dev_shared_data_prepare(); - - rte_spinlock_lock(ð_dev_shared_data->ownership_lock); + rte_spinlock_lock(rte_mcfg_ethdev_get_lock()); + eth_dev_shared_data_prepare(); ret = eth_dev_owner_set(port_id, owner_id, &new_owner); - rte_spinlock_unlock(ð_dev_shared_data->ownership_lock); + rte_spinlock_unlock(rte_mcfg_ethdev_get_lock()); rte_ethdev_trace_owner_unset(port_id, owner_id, ret); @@ -542,10 +541,9 @@ rte_eth_dev_owner_delete(const uint64_t owner_id) uint16_t port_id; int ret = 0; - eth_dev_shared_data_prepare(); - - rte_spinlock_lock(ð_dev_shared_data->ownership_lock); + rte_spinlock_lock(rte_mcfg_ethdev_get_lock()); + eth_dev_shared_data_prepare(); if (eth_is_valid_owner_id(owner_id)) { for (port_id = 0; port_id < RTE_MAX_ETHPORTS; port_id++) { struct rte_eth_dev_data *data = @@ -564,7 +562,7 @@ rte_eth_dev_owner_delete(const uint64_t owner_id) ret = -EINVAL; } - rte_spinlock_unlock(ð_dev_shared_data->ownership_lock); + rte_spinlock_unlock(rte_mcfg_ethdev_get_lock()); rte_ethdev_trace_owner_delete(owner_id, ret); @@ -591,11 +589,12 @@ rte_eth_dev_owner_get(const uint16_t port_id, struct rte_eth_dev_owner *owner) return -EINVAL; } - eth_dev_shared_data_prepare(); + rte_spinlock_lock(rte_mcfg_ethdev_get_lock()); - rte_spinlock_lock(ð_dev_shared_data->ownership_lock); + eth_dev_shared_data_prepare(); rte_memcpy(owner, ðdev->data->owner, sizeof(*owner)); - rte_spinlock_unlock(ð_dev_shared_data->ownership_lock); + + rte_spinlock_unlock(rte_mcfg_ethdev_get_lock()); rte_ethdev_trace_owner_get(port_id, owner); @@ -675,9 +674,12 @@ rte_eth_dev_get_name_by_port(uint16_t port_id, char *name) return -EINVAL; } + rte_spinlock_lock(rte_mcfg_ethdev_get_lock()); /* shouldn't check 'rte_eth_devices[i].data', * because it might be overwritten by VDEV PMD */ tmp = eth_dev_shared_data->data[port_id].name; + rte_spinlock_unlock(rte_mcfg_ethdev_get_lock()); + strcpy(name, tmp); rte_ethdev_trace_get_name_by_port(port_id, name); @@ -688,6 +690,7 @@ rte_eth_dev_get_name_by_port(uint16_t port_id, char *name) int rte_eth_dev_get_port_by_name(const char *name, uint16_t *port_id) { + int ret = -ENODEV; uint16_t pid; if (name == NULL) { @@ -701,16 +704,19 @@ rte_eth_dev_get_port_by_name(const char *name, uint16_t *port_id) return -EINVAL; } - RTE_ETH_FOREACH_VALID_DEV(pid) - if (!strcmp(name, eth_dev_shared_data->data[pid].name)) { - *port_id = pid; - - rte_ethdev_trace_get_port_by_name(name, *port_id); + rte_spinlock_lock(rte_mcfg_ethdev_get_lock()); + RTE_ETH_FOREACH_VALID_DEV(pid) { + if (strcmp(name, eth_dev_shared_data->data[pid].name) != 0) + continue; - return 0; - } + *port_id = pid; + rte_ethdev_trace_get_port_by_name(name, *port_id); + ret = 0; + break; + } + rte_spinlock_unlock(rte_mcfg_ethdev_get_lock()); - return -ENODEV; + return ret; } int From patchwork Wed Sep 27 11:45:14 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: David Marchand X-Patchwork-Id: 132001 X-Patchwork-Delegate: thomas@monjalon.net Return-Path: X-Original-To: patchwork@inbox.dpdk.org Delivered-To: patchwork@inbox.dpdk.org Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by inbox.dpdk.org (Postfix) with ESMTP id 4DD1D42651; Wed, 27 Sep 2023 13:45:35 +0200 (CEST) Received: from mails.dpdk.org (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id 4E8DD402F0; Wed, 27 Sep 2023 13:45:32 +0200 (CEST) Received: from us-smtp-delivery-124.mimecast.com (us-smtp-delivery-124.mimecast.com [170.10.133.124]) by mails.dpdk.org (Postfix) with ESMTP id 6AD92402E9 for ; Wed, 27 Sep 2023 13:45:30 +0200 (CEST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=redhat.com; s=mimecast20190719; t=1695815129; h=from:from:reply-to:subject:subject:date:date:message-id:message-id: to:to:cc:cc:mime-version:mime-version:content-type:content-type: content-transfer-encoding:content-transfer-encoding: in-reply-to:in-reply-to:references:references; bh=t3QZkLxMozZYwtm6l/e7tTv4v5o+z9+056wyl+Q7ylA=; b=HuaKoyGtmRtEsf/JWg9Oy1WO4rWho52q4qp2Bdvh7h3QJBDfb6QH7T1clilHkGYNrQyNmg BajOyDmFUeMbyt8r7eBR3ys4S7nXZP8BAG+J7M4pSwLPIoej8rdzA4c1hhKoGk5h/5IgYw +mHKn3pLz4+Z+WBXOWZILGmj4SW3R9o= Received: from mimecast-mx02.redhat.com (mimecast-mx02.redhat.com [66.187.233.88]) by relay.mimecast.com with ESMTP with STARTTLS (version=TLSv1.2, cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id us-mta-10-eeuOHZ7dOBq2xBamYVbJ8A-1; Wed, 27 Sep 2023 07:45:25 -0400 X-MC-Unique: eeuOHZ7dOBq2xBamYVbJ8A-1 Received: from smtp.corp.redhat.com (int-mx08.intmail.prod.int.rdu2.redhat.com [10.11.54.8]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by mimecast-mx02.redhat.com (Postfix) with ESMTPS id 333A580D69A; Wed, 27 Sep 2023 11:45:25 +0000 (UTC) Received: from dmarchan.redhat.com (unknown [10.45.224.48]) by smtp.corp.redhat.com (Postfix) with ESMTP id 21159C15BB8; Wed, 27 Sep 2023 11:45:23 +0000 (UTC) From: David Marchand To: dev@dpdk.org Cc: probb@iol.unh.edu, Thomas Monjalon , Ferruh Yigit , Andrew Rybchenko Subject: [PATCH v4 2/3] ethdev: avoid panicking in absence of ethdev shared data Date: Wed, 27 Sep 2023 13:45:14 +0200 Message-ID: <20230927114515.1245213-3-david.marchand@redhat.com> In-Reply-To: <20230927114515.1245213-1-david.marchand@redhat.com> References: <20230818091321.2404089-1-david.marchand@redhat.com> <20230927114515.1245213-1-david.marchand@redhat.com> MIME-Version: 1.0 X-Scanned-By: MIMEDefang 3.1 on 10.11.54.8 X-Mimecast-Spam-Score: 0 X-Mimecast-Originator: redhat.com X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org This is a preparation step before freeing the ethdev shared data memzone. Previously, because the primary process never freed the memzone, a secondary process could assume this memzone was present. But in the next commit, this will change. Make eth_dev_shared_data_prepare() report whether the memzone is available so that upper level API can react accordingly. Signed-off-by: David Marchand --- lib/ethdev/ethdev_driver.c | 23 ++++++++++++++----- lib/ethdev/ethdev_private.c | 10 +++++--- lib/ethdev/ethdev_private.h | 2 +- lib/ethdev/ethdev_trace.h | 6 +++-- lib/ethdev/rte_ethdev.c | 46 +++++++++++++++++++++++++------------ 5 files changed, 60 insertions(+), 27 deletions(-) diff --git a/lib/ethdev/ethdev_driver.c b/lib/ethdev/ethdev_driver.c index c92cd4b947..b339e325a0 100644 --- a/lib/ethdev/ethdev_driver.c +++ b/lib/ethdev/ethdev_driver.c @@ -92,7 +92,8 @@ rte_eth_dev_allocate(const char *name) /* Synchronize port creation between primary and secondary processes. */ rte_spinlock_lock(rte_mcfg_ethdev_get_lock()); - eth_dev_shared_data_prepare(); + if (eth_dev_shared_data_prepare() == NULL) + goto unlock; if (eth_dev_allocated(name) != NULL) { RTE_ETHDEV_LOG(ERR, @@ -128,9 +129,10 @@ rte_eth_dev_allocated(const char *name) rte_spinlock_lock(rte_mcfg_ethdev_get_lock()); - eth_dev_shared_data_prepare(); - - ethdev = eth_dev_allocated(name); + if (eth_dev_shared_data_prepare() != NULL) + ethdev = eth_dev_allocated(name); + else + ethdev = NULL; rte_spinlock_unlock(rte_mcfg_ethdev_get_lock()); @@ -151,7 +153,8 @@ rte_eth_dev_attach_secondary(const char *name) /* Synchronize port attachment to primary port creation and release. */ rte_spinlock_lock(rte_mcfg_ethdev_get_lock()); - eth_dev_shared_data_prepare(); + if (eth_dev_shared_data_prepare() == NULL) + goto unlock; for (i = 0; i < RTE_MAX_ETHPORTS; i++) { if (strcmp(eth_dev_shared_data->data[i].name, name) == 0) @@ -166,6 +169,7 @@ rte_eth_dev_attach_secondary(const char *name) RTE_ASSERT(eth_dev->data->port_id == i); } +unlock: rte_spinlock_unlock(rte_mcfg_ethdev_get_lock()); return eth_dev; } @@ -219,12 +223,19 @@ rte_eth_dev_probing_finish(struct rte_eth_dev *dev) int rte_eth_dev_release_port(struct rte_eth_dev *eth_dev) { + int ret; + if (eth_dev == NULL) return -EINVAL; rte_spinlock_lock(rte_mcfg_ethdev_get_lock()); - eth_dev_shared_data_prepare(); + if (eth_dev_shared_data_prepare() == NULL) + ret = -EINVAL; + else + ret = 0; rte_spinlock_unlock(rte_mcfg_ethdev_get_lock()); + if (ret != 0) + return ret; if (eth_dev->state != RTE_ETH_DEV_UNUSED) rte_eth_dev_callback_process(eth_dev, diff --git a/lib/ethdev/ethdev_private.c b/lib/ethdev/ethdev_private.c index 6756625729..911de1e595 100644 --- a/lib/ethdev/ethdev_private.c +++ b/lib/ethdev/ethdev_private.c @@ -318,7 +318,7 @@ rte_eth_call_tx_callbacks(uint16_t port_id, uint16_t queue_id, return nb_pkts; } -void +void * eth_dev_shared_data_prepare(void) { const unsigned int flags = 0; @@ -332,8 +332,10 @@ eth_dev_shared_data_prepare(void) rte_socket_id(), flags); } else mz = rte_memzone_lookup(MZ_RTE_ETH_DEV_DATA); - if (mz == NULL) - rte_panic("Cannot allocate ethdev shared data\n"); + if (mz == NULL) { + RTE_ETHDEV_LOG(ERR, "Cannot allocate ethdev shared data\n"); + goto out; + } eth_dev_shared_data = mz->addr; if (rte_eal_process_type() == RTE_PROC_PRIMARY) { @@ -343,6 +345,8 @@ eth_dev_shared_data_prepare(void) sizeof(eth_dev_shared_data->data)); } } +out: + return eth_dev_shared_data; } void diff --git a/lib/ethdev/ethdev_private.h b/lib/ethdev/ethdev_private.h index f7706e6a95..1572da7b48 100644 --- a/lib/ethdev/ethdev_private.h +++ b/lib/ethdev/ethdev_private.h @@ -67,7 +67,7 @@ void eth_dev_fp_ops_setup(struct rte_eth_fp_ops *fpo, const struct rte_eth_dev *dev); -void eth_dev_shared_data_prepare(void) +void *eth_dev_shared_data_prepare(void) __rte_exclusive_locks_required(rte_mcfg_ethdev_get_lock()); void eth_dev_rxq_release(struct rte_eth_dev *dev, uint16_t qid); diff --git a/lib/ethdev/ethdev_trace.h b/lib/ethdev/ethdev_trace.h index 423e71236e..e367d29c3a 100644 --- a/lib/ethdev/ethdev_trace.h +++ b/lib/ethdev/ethdev_trace.h @@ -112,8 +112,9 @@ RTE_TRACE_POINT( RTE_TRACE_POINT( rte_ethdev_trace_owner_new, - RTE_TRACE_POINT_ARGS(uint64_t owner_id), + RTE_TRACE_POINT_ARGS(uint64_t owner_id, int ret), rte_trace_point_emit_u64(owner_id); + rte_trace_point_emit_int(ret); ) RTE_TRACE_POINT( @@ -377,10 +378,11 @@ RTE_TRACE_POINT( RTE_TRACE_POINT( rte_ethdev_trace_owner_get, RTE_TRACE_POINT_ARGS(uint16_t port_id, - const struct rte_eth_dev_owner *owner), + const struct rte_eth_dev_owner *owner, int ret), rte_trace_point_emit_u16(port_id); rte_trace_point_emit_u64(owner->id); rte_trace_point_emit_string(owner->name); + rte_trace_point_emit_int(ret); ) RTE_TRACE_POINT( diff --git a/lib/ethdev/rte_ethdev.c b/lib/ethdev/rte_ethdev.c index e0d22c27c6..5c9495ecfe 100644 --- a/lib/ethdev/rte_ethdev.c +++ b/lib/ethdev/rte_ethdev.c @@ -388,7 +388,7 @@ rte_eth_find_next_sibling(uint16_t port_id, uint16_t ref_port_id) static bool eth_dev_is_allocated(const struct rte_eth_dev *ethdev) { - return ethdev->data->name[0] != '\0'; + return ethdev->data != NULL && ethdev->data->name[0] != '\0'; } int @@ -433,6 +433,8 @@ rte_eth_find_next_owned_by(uint16_t port_id, const uint64_t owner_id) int rte_eth_dev_owner_new(uint64_t *owner_id) { + int ret; + if (owner_id == NULL) { RTE_ETHDEV_LOG(ERR, "Cannot get new owner ID to NULL\n"); return -EINVAL; @@ -440,14 +442,18 @@ rte_eth_dev_owner_new(uint64_t *owner_id) rte_spinlock_lock(rte_mcfg_ethdev_get_lock()); - eth_dev_shared_data_prepare(); - *owner_id = eth_dev_shared_data->next_owner_id++; + if (eth_dev_shared_data_prepare() != NULL) { + *owner_id = eth_dev_shared_data->next_owner_id++; + ret = 0; + } else { + ret = -ENOMEM; + } rte_spinlock_unlock(rte_mcfg_ethdev_get_lock()); - rte_ethdev_trace_owner_new(*owner_id); + rte_ethdev_trace_owner_new(*owner_id, ret); - return 0; + return ret; } static int @@ -506,8 +512,10 @@ rte_eth_dev_owner_set(const uint16_t port_id, rte_spinlock_lock(rte_mcfg_ethdev_get_lock()); - eth_dev_shared_data_prepare(); - ret = eth_dev_owner_set(port_id, RTE_ETH_DEV_NO_OWNER, owner); + if (eth_dev_shared_data_prepare() != NULL) + ret = eth_dev_owner_set(port_id, RTE_ETH_DEV_NO_OWNER, owner); + else + ret = -ENOMEM; rte_spinlock_unlock(rte_mcfg_ethdev_get_lock()); @@ -525,8 +533,10 @@ rte_eth_dev_owner_unset(const uint16_t port_id, const uint64_t owner_id) rte_spinlock_lock(rte_mcfg_ethdev_get_lock()); - eth_dev_shared_data_prepare(); - ret = eth_dev_owner_set(port_id, owner_id, &new_owner); + if (eth_dev_shared_data_prepare() != NULL) + ret = eth_dev_owner_set(port_id, owner_id, &new_owner); + else + ret = -ENOMEM; rte_spinlock_unlock(rte_mcfg_ethdev_get_lock()); @@ -543,8 +553,9 @@ rte_eth_dev_owner_delete(const uint64_t owner_id) rte_spinlock_lock(rte_mcfg_ethdev_get_lock()); - eth_dev_shared_data_prepare(); - if (eth_is_valid_owner_id(owner_id)) { + if (eth_dev_shared_data_prepare() == NULL) { + ret = -ENOMEM; + } else if (eth_is_valid_owner_id(owner_id)) { for (port_id = 0; port_id < RTE_MAX_ETHPORTS; port_id++) { struct rte_eth_dev_data *data = rte_eth_devices[port_id].data; @@ -573,6 +584,7 @@ int rte_eth_dev_owner_get(const uint16_t port_id, struct rte_eth_dev_owner *owner) { struct rte_eth_dev *ethdev; + int ret; RTE_ETH_VALID_PORTID_OR_ERR_RET(port_id, -ENODEV); ethdev = &rte_eth_devices[port_id]; @@ -591,14 +603,18 @@ rte_eth_dev_owner_get(const uint16_t port_id, struct rte_eth_dev_owner *owner) rte_spinlock_lock(rte_mcfg_ethdev_get_lock()); - eth_dev_shared_data_prepare(); - rte_memcpy(owner, ðdev->data->owner, sizeof(*owner)); + if (eth_dev_shared_data_prepare() != NULL) { + rte_memcpy(owner, ðdev->data->owner, sizeof(*owner)); + ret = 0; + } else { + ret = -ENOMEM; + } rte_spinlock_unlock(rte_mcfg_ethdev_get_lock()); - rte_ethdev_trace_owner_get(port_id, owner); + rte_ethdev_trace_owner_get(port_id, owner, ret); - return 0; + return ret; } int From patchwork Wed Sep 27 11:45:15 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 8bit X-Patchwork-Submitter: David Marchand X-Patchwork-Id: 132002 X-Patchwork-Delegate: thomas@monjalon.net Return-Path: X-Original-To: patchwork@inbox.dpdk.org Delivered-To: patchwork@inbox.dpdk.org Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by inbox.dpdk.org (Postfix) with ESMTP id 9D69542651; Wed, 27 Sep 2023 13:45:43 +0200 (CEST) Received: from mails.dpdk.org (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id E5C1240A77; Wed, 27 Sep 2023 13:45:33 +0200 (CEST) Received: from us-smtp-delivery-124.mimecast.com (us-smtp-delivery-124.mimecast.com [170.10.129.124]) by mails.dpdk.org (Postfix) with ESMTP id C69A6402D3 for ; Wed, 27 Sep 2023 13:45:32 +0200 (CEST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=redhat.com; s=mimecast20190719; t=1695815132; h=from:from:reply-to:subject:subject:date:date:message-id:message-id: to:to:cc:cc:mime-version:mime-version:content-type:content-type: content-transfer-encoding:content-transfer-encoding: in-reply-to:in-reply-to:references:references; bh=7FFBDXEwdl2KJTlmVks+gN6W2Btkmts3wnu5w1aBpPM=; b=U3e7cGRNcSJSAgH6HO0kvs8iFGeoKnONOzzD6/awRLWZcdnmuPOQ3VKZnWslTT769zQQLo R/echb7RAjSF/qH0mI7a8WDMBoT/2cLkvLIw/K8k9RHfHK3IAt/YyYj4ZtmZaXMC6aazJg Esyt1iKP2xnlEPtuehc7BvZMcshkY6w= Received: from mimecast-mx02.redhat.com (mimecast-mx02.redhat.com [66.187.233.88]) by relay.mimecast.com with ESMTP with STARTTLS (version=TLSv1.2, cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id us-mta-219-dCDp7TGNM4-0VPOZO8zC0w-1; Wed, 27 Sep 2023 07:45:29 -0400 X-MC-Unique: dCDp7TGNM4-0VPOZO8zC0w-1 Received: from smtp.corp.redhat.com (int-mx10.intmail.prod.int.rdu2.redhat.com [10.11.54.10]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by mimecast-mx02.redhat.com (Postfix) with ESMTPS id 7062B8032F6; Wed, 27 Sep 2023 11:45:28 +0000 (UTC) Received: from dmarchan.redhat.com (unknown [10.45.224.48]) by smtp.corp.redhat.com (Postfix) with ESMTP id 186B8492C37; Wed, 27 Sep 2023 11:45:26 +0000 (UTC) From: David Marchand To: dev@dpdk.org Cc: probb@iol.unh.edu, =?utf-8?q?Morten_Br=C3=B8rup?= , Thomas Monjalon , Ferruh Yigit , Andrew Rybchenko , Anatoly Burakov Subject: [PATCH v4 3/3] ethdev: cleanup shared data with the last port Date: Wed, 27 Sep 2023 13:45:15 +0200 Message-ID: <20230927114515.1245213-4-david.marchand@redhat.com> In-Reply-To: <20230927114515.1245213-1-david.marchand@redhat.com> References: <20230818091321.2404089-1-david.marchand@redhat.com> <20230927114515.1245213-1-david.marchand@redhat.com> MIME-Version: 1.0 X-Scanned-By: MIMEDefang 3.1 on 10.11.54.10 X-Mimecast-Spam-Score: 0 X-Mimecast-Originator: redhat.com X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org If no port is allocated and no port owner is still registered, ethdev from a primary process may release the memzone used to store port data. This makes it possible for the DPDK memory allocator to release associated resources back to the OS. Signed-off-by: David Marchand Acked-by: Morten Brørup --- Changes since v2: - tracked owners count and and avoided releasing shared mem if some owner is still registered, --- lib/ethdev/ethdev_driver.c | 6 ++++++ lib/ethdev/ethdev_private.c | 21 ++++++++++++++++++++- lib/ethdev/ethdev_private.h | 4 ++++ lib/ethdev/rte_ethdev.c | 3 +++ 4 files changed, 33 insertions(+), 1 deletion(-) diff --git a/lib/ethdev/ethdev_driver.c b/lib/ethdev/ethdev_driver.c index b339e325a0..fff4b7b4cd 100644 --- a/lib/ethdev/ethdev_driver.c +++ b/lib/ethdev/ethdev_driver.c @@ -115,6 +115,8 @@ rte_eth_dev_allocate(const char *name) eth_dev->data->backer_port_id = RTE_MAX_ETHPORTS; eth_dev->data->mtu = RTE_ETHER_MTU; pthread_mutex_init(ð_dev->data->flow_ops_mutex, NULL); + RTE_ASSERT(rte_eal_process_type() == RTE_PROC_PRIMARY); + eth_dev_shared_data->allocated_ports++; unlock: rte_spinlock_unlock(rte_mcfg_ethdev_get_lock()); @@ -265,6 +267,10 @@ rte_eth_dev_release_port(struct rte_eth_dev *eth_dev) rte_free(eth_dev->data->dev_private); pthread_mutex_destroy(ð_dev->data->flow_ops_mutex); memset(eth_dev->data, 0, sizeof(struct rte_eth_dev_data)); + eth_dev->data = NULL; + + eth_dev_shared_data->allocated_ports--; + eth_dev_shared_data_release(); } rte_spinlock_unlock(rte_mcfg_ethdev_get_lock()); diff --git a/lib/ethdev/ethdev_private.c b/lib/ethdev/ethdev_private.c index 911de1e595..aea9112cc2 100644 --- a/lib/ethdev/ethdev_private.c +++ b/lib/ethdev/ethdev_private.c @@ -11,6 +11,7 @@ static const char *MZ_RTE_ETH_DEV_DATA = "rte_eth_dev_data"; +static const struct rte_memzone *eth_dev_shared_mz; struct eth_dev_shared *eth_dev_shared_data; /* spinlock for eth device callbacks */ @@ -324,7 +325,7 @@ eth_dev_shared_data_prepare(void) const unsigned int flags = 0; const struct rte_memzone *mz; - if (eth_dev_shared_data == NULL) { + if (eth_dev_shared_mz == NULL) { if (rte_eal_process_type() == RTE_PROC_PRIMARY) { /* Allocate port data and ownership shared memory. */ mz = rte_memzone_reserve(MZ_RTE_ETH_DEV_DATA, @@ -337,10 +338,13 @@ eth_dev_shared_data_prepare(void) goto out; } + eth_dev_shared_mz = mz; eth_dev_shared_data = mz->addr; if (rte_eal_process_type() == RTE_PROC_PRIMARY) { + eth_dev_shared_data->allocated_owners = 0; eth_dev_shared_data->next_owner_id = RTE_ETH_DEV_NO_OWNER + 1; + eth_dev_shared_data->allocated_ports = 0; memset(eth_dev_shared_data->data, 0, sizeof(eth_dev_shared_data->data)); } @@ -349,6 +353,21 @@ eth_dev_shared_data_prepare(void) return eth_dev_shared_data; } +void +eth_dev_shared_data_release(void) +{ + RTE_ASSERT(rte_eal_process_type() == RTE_PROC_PRIMARY); + + if (eth_dev_shared_data->allocated_ports != 0) + return; + if (eth_dev_shared_data->allocated_owners != 0) + return; + + rte_memzone_free(eth_dev_shared_mz); + eth_dev_shared_mz = NULL; + eth_dev_shared_data = NULL; +} + void eth_dev_rxq_release(struct rte_eth_dev *dev, uint16_t qid) { diff --git a/lib/ethdev/ethdev_private.h b/lib/ethdev/ethdev_private.h index 1572da7b48..0d36b9c30f 100644 --- a/lib/ethdev/ethdev_private.h +++ b/lib/ethdev/ethdev_private.h @@ -14,7 +14,9 @@ #include "rte_ethdev.h" struct eth_dev_shared { + uint64_t allocated_owners; uint64_t next_owner_id; + uint64_t allocated_ports; struct rte_eth_dev_data data[RTE_MAX_ETHPORTS]; }; @@ -69,6 +71,8 @@ void eth_dev_fp_ops_setup(struct rte_eth_fp_ops *fpo, void *eth_dev_shared_data_prepare(void) __rte_exclusive_locks_required(rte_mcfg_ethdev_get_lock()); +void eth_dev_shared_data_release(void) + __rte_exclusive_locks_required(rte_mcfg_ethdev_get_lock()); void eth_dev_rxq_release(struct rte_eth_dev *dev, uint16_t qid); void eth_dev_txq_release(struct rte_eth_dev *dev, uint16_t qid); diff --git a/lib/ethdev/rte_ethdev.c b/lib/ethdev/rte_ethdev.c index 5c9495ecfe..61572d0cd1 100644 --- a/lib/ethdev/rte_ethdev.c +++ b/lib/ethdev/rte_ethdev.c @@ -444,6 +444,7 @@ rte_eth_dev_owner_new(uint64_t *owner_id) if (eth_dev_shared_data_prepare() != NULL) { *owner_id = eth_dev_shared_data->next_owner_id++; + eth_dev_shared_data->allocated_owners++; ret = 0; } else { ret = -ENOMEM; @@ -566,6 +567,8 @@ rte_eth_dev_owner_delete(const uint64_t owner_id) RTE_ETHDEV_LOG(NOTICE, "All port owners owned by %016"PRIx64" identifier have removed\n", owner_id); + eth_dev_shared_data->allocated_owners--; + eth_dev_shared_data_release(); } else { RTE_ETHDEV_LOG(ERR, "Invalid owner ID=%016"PRIx64"\n",