From patchwork Mon Sep 10 16:39:11 2018 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Dennis Dalessandro X-Patchwork-Id: 10594521 X-Patchwork-Delegate: jgg@ziepe.ca Return-Path: Received: from mail.wl.linuxfoundation.org (pdx-wl-mail.web.codeaurora.org [172.30.200.125]) by pdx-korg-patchwork-2.web.codeaurora.org (Postfix) with ESMTP id 644D5921 for ; Mon, 10 Sep 2018 16:39:26 +0000 (UTC) Received: from mail.wl.linuxfoundation.org (localhost [127.0.0.1]) by mail.wl.linuxfoundation.org (Postfix) with ESMTP id 538BE293A0 for ; Mon, 10 Sep 2018 16:39:26 +0000 (UTC) Received: by mail.wl.linuxfoundation.org (Postfix, from userid 486) id 47DB8293AF; Mon, 10 Sep 2018 16:39:26 +0000 (UTC) X-Spam-Checker-Version: SpamAssassin 3.3.1 (2010-03-16) on pdx-wl-mail.web.codeaurora.org X-Spam-Level: X-Spam-Status: No, score=-7.9 required=2.0 tests=BAYES_00,MAILING_LIST_MULTI, RCVD_IN_DNSWL_HI autolearn=ham version=3.3.1 Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by mail.wl.linuxfoundation.org (Postfix) with ESMTP id C7793293A0 for ; Mon, 10 Sep 2018 16:39:25 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1728137AbeIJVeT (ORCPT ); Mon, 10 Sep 2018 17:34:19 -0400 Received: from mga14.intel.com ([192.55.52.115]:8230 "EHLO mga14.intel.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1727815AbeIJVeT (ORCPT ); Mon, 10 Sep 2018 17:34:19 -0400 X-Amp-Result: SKIPPED(no attachment in message) X-Amp-File-Uploaded: False Received: from orsmga003.jf.intel.com ([10.7.209.27]) by fmsmga103.fm.intel.com with ESMTP/TLS/DHE-RSA-AES256-GCM-SHA384; 10 Sep 2018 09:39:24 -0700 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.53,356,1531810800"; d="scan'208";a="82346580" Received: from scymds01.sc.intel.com ([10.82.194.37]) by orsmga003.jf.intel.com with ESMTP; 10 Sep 2018 09:39:12 -0700 Received: from scvm10.sc.intel.com (scvm10.sc.intel.com [10.82.195.27]) by scymds01.sc.intel.com with ESMTP id w8AGdBB9009096; Mon, 10 Sep 2018 09:39:11 -0700 Received: from scvm10.sc.intel.com (localhost [127.0.0.1]) by scvm10.sc.intel.com with ESMTP id w8AGdB9q028854; Mon, 10 Sep 2018 09:39:11 -0700 Subject: [PATCH v2 for-next 2/4] IB/hfi1: Remove race conditions in user_sdma send path From: Dennis Dalessandro To: jgg@ziepe.ca, dledford@redhat.com Cc: linux-rdma@vger.kernel.org, "Michael J. Ruhl" , Mitko Haralanov , Mike Marciniszyn Date: Mon, 10 Sep 2018 09:39:11 -0700 Message-ID: <20180910163908.23326.45371.stgit@scvm10.sc.intel.com> In-Reply-To: <20180910163707.23326.1348.stgit@scvm10.sc.intel.com> References: <20180910163707.23326.1348.stgit@scvm10.sc.intel.com> User-Agent: StGit/0.17.1-18-g2e886-dirty MIME-Version: 1.0 Sender: linux-rdma-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-rdma@vger.kernel.org X-Virus-Scanned: ClamAV using ClamSMTP From: Michael J. Ruhl Packet queue state is over used to determine SDMA descriptor availablitity and packet queue request state. cpu 0 ret = user_sdma_send_pkts(req, pcount); cpu 0 if (atomic_read(&pq->n_reqs)) cpu 1 IRQ user_sdma_txreq_cb calls pq_update() (state to _INACTIVE) cpu 0 xchg(&pq->state, SDMA_PKT_Q_ACTIVE); At this point pq->n_reqs == 0 and pq->state is incorrectly SDMA_PKT_Q_ACTIVE. The close path will hang waiting for the state to return to _INACTIVE. This can also change the state from _DEFERRED to _ACTIVE. However, this is a mostly benign race. Remove the racy code path. Use n_reqs to determine if a packet queue is active or not. Reviewed-by: Mitko Haralanov Reviewed-by: Mike Marciniszyn Signed-off-by: Michael J. Ruhl Signed-off-by: Dennis Dalessandro --- drivers/infiniband/hw/hfi1/user_sdma.c | 24 ++++++++++-------------- drivers/infiniband/hw/hfi1/user_sdma.h | 9 +++++---- 2 files changed, 15 insertions(+), 18 deletions(-) diff --git a/drivers/infiniband/hw/hfi1/user_sdma.c b/drivers/infiniband/hw/hfi1/user_sdma.c index 9f47167..fd6e9f5 100644 --- a/drivers/infiniband/hw/hfi1/user_sdma.c +++ b/drivers/infiniband/hw/hfi1/user_sdma.c @@ -187,7 +187,6 @@ int hfi1_user_sdma_alloc_queues(struct hfi1_ctxtdata *uctxt, pq->ctxt = uctxt->ctxt; pq->subctxt = fd->subctxt; pq->n_max_reqs = hfi1_sdma_comp_ring_size; - pq->state = SDMA_PKT_Q_INACTIVE; atomic_set(&pq->n_reqs, 0); init_waitqueue_head(&pq->wait); atomic_set(&pq->n_locked, 0); @@ -276,7 +275,7 @@ int hfi1_user_sdma_free_queues(struct hfi1_filedata *fd, /* Wait until all requests have been freed. */ wait_event_interruptible( pq->wait, - (READ_ONCE(pq->state) == SDMA_PKT_Q_INACTIVE)); + !atomic_read(&pq->n_reqs)); kfree(pq->reqs); kfree(pq->req_in_use); kmem_cache_destroy(pq->txreq_cache); @@ -312,6 +311,13 @@ static u8 dlid_to_selector(u16 dlid) return mapping[hash]; } +/** + * hfi1_user_sdma_process_request() - Process and start a user sdma request + * @fd: valid file descriptor + * @iovec: array of io vectors to process + * @dim: overall iovec array size + * @count: number of io vector array entries processed + */ int hfi1_user_sdma_process_request(struct hfi1_filedata *fd, struct iovec *iovec, unsigned long dim, unsigned long *count) @@ -560,21 +566,13 @@ int hfi1_user_sdma_process_request(struct hfi1_filedata *fd, req->ahg_idx = sdma_ahg_alloc(req->sde); set_comp_state(pq, cq, info.comp_idx, QUEUED, 0); + pq->state = SDMA_PKT_Q_ACTIVE; /* Send the first N packets in the request to buy us some time */ ret = user_sdma_send_pkts(req, pcount); if (unlikely(ret < 0 && ret != -EBUSY)) goto free_req; /* - * It is possible that the SDMA engine would have processed all the - * submitted packets by the time we get here. Therefore, only set - * packet queue state to ACTIVE if there are still uncompleted - * requests. - */ - if (atomic_read(&pq->n_reqs)) - xchg(&pq->state, SDMA_PKT_Q_ACTIVE); - - /* * This is a somewhat blocking send implementation. * The driver will block the caller until all packets of the * request have been submitted to the SDMA engine. However, it @@ -1409,10 +1407,8 @@ static void user_sdma_txreq_cb(struct sdma_txreq *txreq, int status) static inline void pq_update(struct hfi1_user_sdma_pkt_q *pq) { - if (atomic_dec_and_test(&pq->n_reqs)) { - xchg(&pq->state, SDMA_PKT_Q_INACTIVE); + if (atomic_dec_and_test(&pq->n_reqs)) wake_up(&pq->wait); - } } static void user_sdma_free_request(struct user_sdma_request *req, bool unpin) diff --git a/drivers/infiniband/hw/hfi1/user_sdma.h b/drivers/infiniband/hw/hfi1/user_sdma.h index 0ae0645..91c343f 100644 --- a/drivers/infiniband/hw/hfi1/user_sdma.h +++ b/drivers/infiniband/hw/hfi1/user_sdma.h @@ -105,9 +105,10 @@ static inline int ahg_header_set(u32 *arr, int idx, size_t array_size, #define TXREQ_FLAGS_REQ_ACK BIT(0) /* Set the ACK bit in the header */ #define TXREQ_FLAGS_REQ_DISABLE_SH BIT(1) /* Disable header suppression */ -#define SDMA_PKT_Q_INACTIVE BIT(0) -#define SDMA_PKT_Q_ACTIVE BIT(1) -#define SDMA_PKT_Q_DEFERRED BIT(2) +enum pkt_q_sdma_state { + SDMA_PKT_Q_ACTIVE, + SDMA_PKT_Q_DEFERRED, +}; /* * Maximum retry attempts to submit a TX request @@ -133,7 +134,7 @@ struct hfi1_user_sdma_pkt_q { struct user_sdma_request *reqs; unsigned long *req_in_use; struct iowait busy; - unsigned state; + enum pkt_q_sdma_state state; wait_queue_head_t wait; unsigned long unpinned; struct mmu_rb_handler *handler;