From mboxrd@z Thu Jan 1 00:00:00 1970 From: Neil Horman Subject: Re: [PATCH net] cxgb4: Fix possible race condition in cleanup Date: Tue, 19 Aug 2014 09:05:51 -0400 Message-ID: <20140819130551.GA6850@hmsreliant.think-freely.org> References: <1408394819-31073-1-git-send-email-anish@chelsio.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: netdev@vger.kernel.org, davem@davemloft.net, leedom@chelsio.com, hariprasad@chelsio.com, svemuri@chelsio.com To: Anish Bhatt Return-path: Received: from mx1.redhat.com ([209.132.183.28]:6071 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753385AbaHSNF7 (ORCPT ); Tue, 19 Aug 2014 09:05:59 -0400 Content-Disposition: inline In-Reply-To: <1408394819-31073-1-git-send-email-anish@chelsio.com> Sender: netdev-owner@vger.kernel.org List-ID: On Mon, Aug 18, 2014 at 01:46:59PM -0700, Anish Bhatt wrote: > There is a possible race condition when we unregister the PCI Driver and then > flush/destroy the global workq. This could lead to situations where there > are tasks on the work queue with references to now deleted adapter data > structures. Instead, have per-adapter work queues which were instantiated and > torn down in init_one() and remove_one(), respectively. > > Signed-off-by: Anish Bhatt > Signed-off-by: Casey Leedom > --- > drivers/net/ethernet/chelsio/cxgb4/cxgb4.h | 1 + > drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c | 33 ++++++++++++++----------- > 2 files changed, 20 insertions(+), 14 deletions(-) > > diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h b/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h > index d572821..c067b78 100644 > --- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h > +++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h > @@ -652,6 +652,7 @@ struct adapter { > struct tid_info tids; > void **tid_release_head; > spinlock_t tid_release_lock; > + struct workqueue_struct *workq; > struct work_struct tid_release_task; > struct work_struct db_full_task; > struct work_struct db_drop_task; > diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c > index 1afee70..e92aba8 100644 > --- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c > +++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c > @@ -643,8 +643,6 @@ static int set_rxmode(struct net_device *dev, int mtu, bool sleep_ok) > return ret; > } > > -static struct workqueue_struct *workq; > - > /** > * link_start - enable a port > * @dev: the port to enable > @@ -3340,7 +3338,7 @@ static void cxgb4_queue_tid_release(struct tid_info *t, unsigned int chan, > adap->tid_release_head = (void **)((uintptr_t)p | chan); > if (!adap->tid_release_task_busy) { > adap->tid_release_task_busy = true; > - queue_work(workq, &adap->tid_release_task); > + queue_work(adap->workq, &adap->tid_release_task); > } > spin_unlock_bh(&adap->tid_release_lock); > } > @@ -4140,7 +4138,7 @@ void t4_db_full(struct adapter *adap) > notify_rdma_uld(adap, CXGB4_CONTROL_DB_FULL); > t4_set_reg_field(adap, SGE_INT_ENABLE3, > DBFIFO_HP_INT | DBFIFO_LP_INT, 0); > - queue_work(workq, &adap->db_full_task); > + queue_work(adap->workq, &adap->db_full_task); > } > } > > @@ -4150,7 +4148,7 @@ void t4_db_dropped(struct adapter *adap) > disable_dbs(adap); > notify_rdma_uld(adap, CXGB4_CONTROL_DB_FULL); > } > - queue_work(workq, &adap->db_drop_task); > + queue_work(adap->workq, &adap->db_drop_task); > } > > static void uld_attach(struct adapter *adap, unsigned int uld) > @@ -6517,6 +6515,12 @@ static int init_one(struct pci_dev *pdev, const struct pci_device_id *ent) > goto out_disable_device; > } > > + adapter->workq = create_singlethread_workqueue("cxgb4"); > + if (!adapter->workq) { > + err = -ENOMEM; > + goto out_free_adapter; > + } > + > /* PCI device has been enabled */ > adapter->flags |= DEV_ENABLED; > > @@ -6715,6 +6719,9 @@ sriov: > out_unmap_bar0: > iounmap(adapter->regs); > out_free_adapter: > + if (adapter->workq) > + destroy_workqueue(adapter->workq); > + > kfree(adapter); > out_disable_device: > pci_disable_pcie_error_reporting(pdev); > @@ -6736,6 +6743,12 @@ static void remove_one(struct pci_dev *pdev) > if (adapter) { > int i; > > + /* Tear down per-adapter Work Queue first since it can contain > + * references to our adapter data structure. > + */ > + flush_workqueue(adapter->workq); > + destroy_workqueue(adapter->workq); > + nit: flush_workqueue is implied by destroy_workqueue. It doesn't change any functionality but you don't need to do both here. destroy_workqueue is sufficient. Neil