[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <20140819130551.GA6850@hmsreliant.think-freely.org>
Date: Tue, 19 Aug 2014 09:05:51 -0400
From: Neil Horman <nhorman@...hat.com>
To: Anish Bhatt <anish@...lsio.com>
Cc: netdev@...r.kernel.org, davem@...emloft.net, leedom@...lsio.com,
hariprasad@...lsio.com, svemuri@...lsio.com
Subject: Re: [PATCH net] cxgb4: Fix possible race condition in cleanup
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 <anish@...lsio.com>
> Signed-off-by: Casey Leedom <leedom@...lsio.com>
> ---
> 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
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Powered by blists - more mailing lists