Commit 5bb3c1e9 authored by Parvi Kaustubhi's avatar Parvi Kaustubhi Committed by Jason Gunthorpe

IB/usnic: Fix deadlock

There is a dead lock in usnic ib_register and netdev_notify path.

	usnic_ib_discover_pf()
	| mutex_lock(&usnic_ib_ibdev_list_lock);
	 | usnic_ib_device_add();
	  | ib_register_device()
	   | usnic_ib_query_port()
	    | mutex_lock(&us_ibdev->usdev_lock);
	     | ib_get_eth_speed()
	      | rtnl_lock()

order of lock: &usnic_ib_ibdev_list_lock -> usdev_lock -> rtnl_lock

	rtnl_lock()
	 | usnic_ib_netdevice_event()
	  | mutex_lock(&usnic_ib_ibdev_list_lock);

order of lock: rtnl_lock -> &usnic_ib_ibdev_list_lock

Solution is to use the core's lock-free ib_device_get_by_netdev() scheme
to lookup ib_dev while handling netdev & inet events.
Signed-off-by: default avatarParvi Kaustubhi <pkaustub@cisco.com>
Reviewed-by: default avatarGovindarajulu Varadarajan <gvaradar@cisco.com>
Reviewed-by: default avatarTanmay Inamdar <tinamdar@cisco.com>
Signed-off-by: default avatarJason Gunthorpe <jgg@mellanox.com>
parent ca22354b
...@@ -216,18 +216,17 @@ static int usnic_ib_netdevice_event(struct notifier_block *notifier, ...@@ -216,18 +216,17 @@ static int usnic_ib_netdevice_event(struct notifier_block *notifier,
unsigned long event, void *ptr) unsigned long event, void *ptr)
{ {
struct usnic_ib_dev *us_ibdev; struct usnic_ib_dev *us_ibdev;
struct ib_device *ibdev;
struct net_device *netdev = netdev_notifier_info_to_dev(ptr); struct net_device *netdev = netdev_notifier_info_to_dev(ptr);
mutex_lock(&usnic_ib_ibdev_list_lock); ibdev = ib_device_get_by_netdev(netdev, RDMA_DRIVER_USNIC);
list_for_each_entry(us_ibdev, &usnic_ib_ibdev_list, ib_dev_link) { if (!ibdev)
if (us_ibdev->netdev == netdev) { return NOTIFY_DONE;
usnic_ib_handle_usdev_event(us_ibdev, event);
break;
}
}
mutex_unlock(&usnic_ib_ibdev_list_lock);
us_ibdev = container_of(ibdev, struct usnic_ib_dev, ib_dev);
usnic_ib_handle_usdev_event(us_ibdev, event);
ib_device_put(ibdev);
return NOTIFY_DONE; return NOTIFY_DONE;
} }
...@@ -282,16 +281,15 @@ static int usnic_ib_inetaddr_event(struct notifier_block *notifier, ...@@ -282,16 +281,15 @@ static int usnic_ib_inetaddr_event(struct notifier_block *notifier,
struct usnic_ib_dev *us_ibdev; struct usnic_ib_dev *us_ibdev;
struct in_ifaddr *ifa = ptr; struct in_ifaddr *ifa = ptr;
struct net_device *netdev = ifa->ifa_dev->dev; struct net_device *netdev = ifa->ifa_dev->dev;
struct ib_device *ibdev;
mutex_lock(&usnic_ib_ibdev_list_lock); ibdev = ib_device_get_by_netdev(netdev, RDMA_DRIVER_USNIC);
list_for_each_entry(us_ibdev, &usnic_ib_ibdev_list, ib_dev_link) { if (!ibdev)
if (us_ibdev->netdev == netdev) { return NOTIFY_DONE;
usnic_ib_handle_inet_event(us_ibdev, event, ptr);
break;
}
}
mutex_unlock(&usnic_ib_ibdev_list_lock);
us_ibdev = container_of(ibdev, struct usnic_ib_dev, ib_dev);
usnic_ib_handle_inet_event(us_ibdev, event, ptr);
ib_device_put(ibdev);
return NOTIFY_DONE; return NOTIFY_DONE;
} }
static struct notifier_block usnic_ib_inetaddr_notifier = { static struct notifier_block usnic_ib_inetaddr_notifier = {
...@@ -342,7 +340,6 @@ static const struct ib_device_ops usnic_dev_ops = { ...@@ -342,7 +340,6 @@ static const struct ib_device_ops usnic_dev_ops = {
.destroy_qp = usnic_ib_destroy_qp, .destroy_qp = usnic_ib_destroy_qp,
.get_dev_fw_str = usnic_get_dev_fw_str, .get_dev_fw_str = usnic_get_dev_fw_str,
.get_link_layer = usnic_ib_port_link_layer, .get_link_layer = usnic_ib_port_link_layer,
.get_netdev = usnic_get_netdev,
.get_port_immutable = usnic_port_immutable, .get_port_immutable = usnic_port_immutable,
.mmap = usnic_ib_mmap, .mmap = usnic_ib_mmap,
.modify_qp = usnic_ib_modify_qp, .modify_qp = usnic_ib_modify_qp,
...@@ -362,6 +359,7 @@ static void *usnic_ib_device_add(struct pci_dev *dev) ...@@ -362,6 +359,7 @@ static void *usnic_ib_device_add(struct pci_dev *dev)
union ib_gid gid; union ib_gid gid;
struct in_device *ind; struct in_device *ind;
struct net_device *netdev; struct net_device *netdev;
int ret;
usnic_dbg("\n"); usnic_dbg("\n");
netdev = pci_get_drvdata(dev); netdev = pci_get_drvdata(dev);
...@@ -416,6 +414,10 @@ static void *usnic_ib_device_add(struct pci_dev *dev) ...@@ -416,6 +414,10 @@ static void *usnic_ib_device_add(struct pci_dev *dev)
us_ibdev->ib_dev.driver_id = RDMA_DRIVER_USNIC; us_ibdev->ib_dev.driver_id = RDMA_DRIVER_USNIC;
rdma_set_device_sysfs_group(&us_ibdev->ib_dev, &usnic_attr_group); rdma_set_device_sysfs_group(&us_ibdev->ib_dev, &usnic_attr_group);
ret = ib_device_set_netdev(&us_ibdev->ib_dev, us_ibdev->netdev, 1);
if (ret)
goto err_fwd_dealloc;
if (ib_register_device(&us_ibdev->ib_dev, "usnic_%d")) if (ib_register_device(&us_ibdev->ib_dev, "usnic_%d"))
goto err_fwd_dealloc; goto err_fwd_dealloc;
......
...@@ -437,16 +437,6 @@ int usnic_ib_query_gid(struct ib_device *ibdev, u8 port, int index, ...@@ -437,16 +437,6 @@ int usnic_ib_query_gid(struct ib_device *ibdev, u8 port, int index,
return 0; return 0;
} }
struct net_device *usnic_get_netdev(struct ib_device *device, u8 port_num)
{
struct usnic_ib_dev *us_ibdev = to_usdev(device);
if (us_ibdev->netdev)
dev_hold(us_ibdev->netdev);
return us_ibdev->netdev;
}
int usnic_ib_query_pkey(struct ib_device *ibdev, u8 port, u16 index, int usnic_ib_query_pkey(struct ib_device *ibdev, u8 port, u16 index,
u16 *pkey) u16 *pkey)
{ {
......
...@@ -48,7 +48,6 @@ int usnic_ib_query_qp(struct ib_qp *qp, struct ib_qp_attr *qp_attr, ...@@ -48,7 +48,6 @@ int usnic_ib_query_qp(struct ib_qp *qp, struct ib_qp_attr *qp_attr,
struct ib_qp_init_attr *qp_init_attr); struct ib_qp_init_attr *qp_init_attr);
int usnic_ib_query_gid(struct ib_device *ibdev, u8 port, int index, int usnic_ib_query_gid(struct ib_device *ibdev, u8 port, int index,
union ib_gid *gid); union ib_gid *gid);
struct net_device *usnic_get_netdev(struct ib_device *device, u8 port_num);
int usnic_ib_query_pkey(struct ib_device *ibdev, u8 port, u16 index, int usnic_ib_query_pkey(struct ib_device *ibdev, u8 port, u16 index,
u16 *pkey); u16 *pkey);
int usnic_ib_alloc_pd(struct ib_pd *ibpd, struct ib_ucontext *context, int usnic_ib_alloc_pd(struct ib_pd *ibpd, struct ib_ucontext *context,
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment