diff mbox

[for-next] iw_cxgb4: Fix possible circular dependency locking warning

Message ID 1510229853-25875-1-git-send-email-bharat@chelsio.com (mailing list archive)
State Accepted
Headers show

Commit Message

Potnuri Bharat Teja Nov. 9, 2017, 12:17 p.m. UTC
Locking sequence of iw_cxgb4 and RoCE drivers in ib_register_device() is
slightly different and this leads to possible circular dependency locking
warning when both the devices are brought up.

Here is the locking sequence upto ib_register_device():
iw_cxgb4: rtnl_mutex(net stack) --> uld_mutex --> device_mutex
RoCE drivers: device_mutex --> rtnl_mutex

Here is the possibility of cross locking:

	CPU #0 (iw_cxgb4) 		     CPU #1 (RoCE drivers)

-> on interface up cxgb4_up()
executed with rtnl_mutex held
-> hold uld_mutex and try
registering ib device
					-> In ib_register_device() hold
					   device_mutex
-> hold device mutex in
ib_register_device
					-> try acquiring rtnl_mutex in
					   ib_enum_roce_netdev()

Current patch schedules the ib_register_device() functionality of
iw_cxgb4 to a workqueue to prevent the possible cross-locking.
Also rename the labels in c4iw_reister_device().

Signed-off-by: Potnuri Bharat Teja <bharat@chelsio.com>
---
 drivers/infiniband/hw/cxgb4/device.c   | 28 +++++++++++++---------------
 drivers/infiniband/hw/cxgb4/iw_cxgb4.h | 10 +++++++++-
 drivers/infiniband/hw/cxgb4/provider.c | 26 +++++++++++++++++---------
 3 files changed, 39 insertions(+), 25 deletions(-)

Comments

Steve Wise Nov. 9, 2017, 2:44 p.m. UTC | #1
> 
> Locking sequence of iw_cxgb4 and RoCE drivers in ib_register_device() is
> slightly different and this leads to possible circular dependency locking
> warning when both the devices are brought up.
> 
> Here is the locking sequence upto ib_register_device():
> iw_cxgb4: rtnl_mutex(net stack) --> uld_mutex --> device_mutex
> RoCE drivers: device_mutex --> rtnl_mutex
> 
> Here is the possibility of cross locking:
> 
> 	CPU #0 (iw_cxgb4) 		     CPU #1 (RoCE drivers)
> 
> -> on interface up cxgb4_up()
> executed with rtnl_mutex held
> -> hold uld_mutex and try
> registering ib device
> 					-> In ib_register_device() hold
> 					   device_mutex
> -> hold device mutex in
> ib_register_device
> 					-> try acquiring rtnl_mutex in
> 					   ib_enum_roce_netdev()
> 
> Current patch schedules the ib_register_device() functionality of
> iw_cxgb4 to a workqueue to prevent the possible cross-locking.
> Also rename the labels in c4iw_reister_device().
> 
> Signed-off-by: Potnuri Bharat Teja <bharat@chelsio.com>

Reviewed-by: Steve Wise <swise@opengridcomputing.com>


--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Doug Ledford Nov. 13, 2017, 9:59 p.m. UTC | #2
On Thu, 2017-11-09 at 08:44 -0600, Steve Wise wrote:
> > 
> > Locking sequence of iw_cxgb4 and RoCE drivers in ib_register_device() is
> > slightly different and this leads to possible circular dependency locking
> > warning when both the devices are brought up.
> > 
> > Here is the locking sequence upto ib_register_device():
> > iw_cxgb4: rtnl_mutex(net stack) --> uld_mutex --> device_mutex
> > RoCE drivers: device_mutex --> rtnl_mutex
> > 
> > Here is the possibility of cross locking:
> > 
> > 	CPU #0 (iw_cxgb4) 		     CPU #1 (RoCE drivers)
> > 
> > -> on interface up cxgb4_up()
> > executed with rtnl_mutex held
> > -> hold uld_mutex and try
> > registering ib device
> > 					-> In ib_register_device() hold
> > 					   device_mutex
> > -> hold device mutex in
> > ib_register_device
> > 					-> try acquiring rtnl_mutex in
> > 					   ib_enum_roce_netdev()
> > 
> > Current patch schedules the ib_register_device() functionality of
> > iw_cxgb4 to a workqueue to prevent the possible cross-locking.
> > Also rename the labels in c4iw_reister_device().
> > 
> > Signed-off-by: Potnuri Bharat Teja <bharat@chelsio.com>
> 
> Reviewed-by: Steve Wise <swise@opengridcomputing.com>
> 
> 

Thanks, applied.
diff mbox

Patch

diff --git a/drivers/infiniband/hw/cxgb4/device.c b/drivers/infiniband/hw/cxgb4/device.c
index 33bfddf3b064..af77d128d242 100644
--- a/drivers/infiniband/hw/cxgb4/device.c
+++ b/drivers/infiniband/hw/cxgb4/device.c
@@ -64,14 +64,9 @@  module_param(c4iw_wr_log_size_order, int, 0444);
 MODULE_PARM_DESC(c4iw_wr_log_size_order,
 		 "Number of entries (log2) in the work request timing log.");
 
-struct uld_ctx {
-	struct list_head entry;
-	struct cxgb4_lld_info lldi;
-	struct c4iw_dev *dev;
-};
-
 static LIST_HEAD(uld_ctx_list);
 static DEFINE_MUTEX(dev_mutex);
+struct workqueue_struct *reg_workq;
 
 #define DB_FC_RESUME_SIZE 64
 #define DB_FC_RESUME_DELAY 1
@@ -912,7 +907,7 @@  static void c4iw_rdev_close(struct c4iw_rdev *rdev)
 	c4iw_destroy_resource(&rdev->resource);
 }
 
-static void c4iw_dealloc(struct uld_ctx *ctx)
+void c4iw_dealloc(struct uld_ctx *ctx)
 {
 	c4iw_rdev_close(&ctx->dev->rdev);
 	WARN_ON_ONCE(!idr_is_empty(&ctx->dev->cqidr));
@@ -1208,8 +1203,6 @@  static int c4iw_uld_state_change(void *handle, enum cxgb4_state new_state)
 	case CXGB4_STATE_UP:
 		pr_info("%s: Up\n", pci_name(ctx->lldi.pdev));
 		if (!ctx->dev) {
-			int ret;
-
 			ctx->dev = c4iw_alloc(&ctx->lldi);
 			if (IS_ERR(ctx->dev)) {
 				pr_err("%s: initialization failed: %ld\n",
@@ -1218,12 +1211,9 @@  static int c4iw_uld_state_change(void *handle, enum cxgb4_state new_state)
 				ctx->dev = NULL;
 				break;
 			}
-			ret = c4iw_register_device(ctx->dev);
-			if (ret) {
-				pr_err("%s: RDMA registration failed: %d\n",
-				       pci_name(ctx->lldi.pdev), ret);
-				c4iw_dealloc(ctx);
-			}
+
+			INIT_WORK(&ctx->reg_work, c4iw_register_device);
+			queue_work(reg_workq, &ctx->reg_work);
 		}
 		break;
 	case CXGB4_STATE_DOWN:
@@ -1551,6 +1541,12 @@  static int __init c4iw_init_module(void)
 	if (!c4iw_debugfs_root)
 		pr_warn("could not create debugfs entry, continuing\n");
 
+	reg_workq = create_singlethread_workqueue("Register_iWARP_device");
+	if (!reg_workq) {
+		pr_err("Failed creating workqueue to register iwarp device\n");
+		return -ENOMEM;
+	}
+
 	cxgb4_register_uld(CXGB4_ULD_RDMA, &c4iw_uld_info);
 
 	return 0;
@@ -1567,6 +1563,8 @@  static void __exit c4iw_exit_module(void)
 		kfree(ctx);
 	}
 	mutex_unlock(&dev_mutex);
+	flush_workqueue(reg_workq);
+	destroy_workqueue(reg_workq);
 	cxgb4_unregister_uld(CXGB4_ULD_RDMA);
 	c4iw_cm_term();
 	debugfs_remove_recursive(c4iw_debugfs_root);
diff --git a/drivers/infiniband/hw/cxgb4/iw_cxgb4.h b/drivers/infiniband/hw/cxgb4/iw_cxgb4.h
index 788ddb6aa57b..42500cbb9669 100644
--- a/drivers/infiniband/hw/cxgb4/iw_cxgb4.h
+++ b/drivers/infiniband/hw/cxgb4/iw_cxgb4.h
@@ -322,6 +322,13 @@  struct c4iw_dev {
 	wait_queue_head_t wait;
 };
 
+struct uld_ctx {
+	struct list_head entry;
+	struct cxgb4_lld_info lldi;
+	struct c4iw_dev *dev;
+	struct work_struct reg_work;
+};
+
 static inline struct c4iw_dev *to_c4iw_dev(struct ib_device *ibdev)
 {
 	return container_of(ibdev, struct c4iw_dev, ibdev);
@@ -989,7 +996,7 @@  void c4iw_rqtpool_destroy(struct c4iw_rdev *rdev);
 void c4iw_ocqp_pool_destroy(struct c4iw_rdev *rdev);
 void c4iw_destroy_resource(struct c4iw_resource *rscp);
 int c4iw_destroy_ctrl_qp(struct c4iw_rdev *rdev);
-int c4iw_register_device(struct c4iw_dev *dev);
+void c4iw_register_device(struct work_struct *work);
 void c4iw_unregister_device(struct c4iw_dev *dev);
 int __init c4iw_cm_init(void);
 void c4iw_cm_term(void);
@@ -1015,6 +1022,7 @@  struct ib_mr *c4iw_alloc_mr(struct ib_pd *pd,
 int c4iw_map_mr_sg(struct ib_mr *ibmr, struct scatterlist *sg, int sg_nents,
 		   unsigned int *sg_offset);
 int c4iw_dealloc_mw(struct ib_mw *mw);
+void c4iw_dealloc(struct uld_ctx *ctx);
 struct ib_mw *c4iw_alloc_mw(struct ib_pd *pd, enum ib_mw_type type,
 			    struct ib_udata *udata);
 struct ib_mr *c4iw_reg_user_mr(struct ib_pd *pd, u64 start,
diff --git a/drivers/infiniband/hw/cxgb4/provider.c b/drivers/infiniband/hw/cxgb4/provider.c
index fb99a05562a3..3387776915a2 100644
--- a/drivers/infiniband/hw/cxgb4/provider.c
+++ b/drivers/infiniband/hw/cxgb4/provider.c
@@ -530,10 +530,12 @@  static void get_dev_fw_str(struct ib_device *dev, char *str)
 		 FW_HDR_FW_VER_BUILD_G(c4iw_dev->rdev.lldi.fw_vers));
 }
 
-int c4iw_register_device(struct c4iw_dev *dev)
+void c4iw_register_device(struct work_struct *work)
 {
 	int ret;
 	int i;
+	struct uld_ctx *ctx = container_of(work, struct uld_ctx, reg_work);
+	struct c4iw_dev *dev = ctx->dev;
 
 	pr_debug("c4iw_dev %p\n", dev);
 	BUG_ON(!dev->rdev.lldi.ports[0]);
@@ -609,8 +611,10 @@  int c4iw_register_device(struct c4iw_dev *dev)
 	dev->ibdev.get_dev_fw_str = get_dev_fw_str;
 
 	dev->ibdev.iwcm = kmalloc(sizeof(struct iw_cm_verbs), GFP_KERNEL);
-	if (!dev->ibdev.iwcm)
-		return -ENOMEM;
+	if (!dev->ibdev.iwcm) {
+		ret = -ENOMEM;
+		goto err_dealloc_ctx;
+	}
 
 	dev->ibdev.iwcm->connect = c4iw_connect;
 	dev->ibdev.iwcm->accept = c4iw_accept_cr;
@@ -625,20 +629,24 @@  int c4iw_register_device(struct c4iw_dev *dev)
 
 	ret = ib_register_device(&dev->ibdev, NULL);
 	if (ret)
-		goto bail1;
+		goto err_kfree_iwcm;
 
 	for (i = 0; i < ARRAY_SIZE(c4iw_class_attributes); ++i) {
 		ret = device_create_file(&dev->ibdev.dev,
 					 c4iw_class_attributes[i]);
 		if (ret)
-			goto bail2;
+			goto err_unregister_device;
 	}
-	return 0;
-bail2:
+	return;
+err_unregister_device:
 	ib_unregister_device(&dev->ibdev);
-bail1:
+err_kfree_iwcm:
 	kfree(dev->ibdev.iwcm);
-	return ret;
+err_dealloc_ctx:
+	pr_err("%s - Failed registering iwarp device: %d\n",
+	       pci_name(ctx->lldi.pdev), ret);
+	c4iw_dealloc(ctx);
+	return;
 }
 
 void c4iw_unregister_device(struct c4iw_dev *dev)