net: ipvlan: fix potential UAF problem for phy_dev
Subsystems (2)
v2 cover net: ipvlan: fix potential UAF problem for phy_dev
Add the reference operation to phy_dev of ipvlan to avoid the potential UAF problem under the following known scenario:
Someone module puts the NETDEV_UNREGISTER event handler to a work, and phy_dev is accessed in the work handler. But when the work is excuted, phy_dev has been destroyed because upper ipvlan did not get reference to phy_dev correctly.
In addition, add net device refcount tracker to ipvlan and fix some error comments for ipvtap module.
------v1->v2:
- Add "Fixes: tag" for fix patches.
Ziyang Xuan (3):
net: ipvlan: fix potential UAF problem for phy_dev
net: ipvlan: add net device refcount tracker
net: ipvtap: fix error comments
drivers/net/ipvlan/ipvlan.h | 1 + drivers/net/ipvlan/ipvlan_main.c | 13 +++++++++++++ drivers/net/ipvlan/ipvtap.c | 4 ++-- 3 files changed, 16 insertions(+), 2 deletions(-)
Signature · 1 line
2.25.1
PATCH v2 1/3 net: ipvlan: fix potential UAF problem for phy_dev
Add the reference operation to phy_dev of ipvlan to avoid the potential UAF problem under the following known scenario:
Someone module puts the NETDEV_UNREGISTER event handler to a work, and phy_dev is accessed in the work handler. But when the work is excuted, phy_dev has been destroyed because upper ipvlan did not get reference to phy_dev correctly.
That likes as the scenario occurred by
commit 563bcbae3ba2 ("net: vlan: fix a UAF in vlan_dev_real_dev()").
--- drivers/net/ipvlan/ipvlan_main.c | 13 +++++++++++++ 1 file changed, 13 insertions(+)
drivers/net/ipvlan/ipvlan_main.c +13 -0
@@ -158,6 +158,10 @@ static int ipvlan_init(struct net_device *dev) } port =ipvlan_port_get_rtnl (phy_dev); port->count +=1 ; + +/* Get ipvlan's reference to phy_dev */ +dev_hold (phy_dev); +return 0 ; }
@@ -665,6 +669,14 @@ void ipvlan_link_delete(struct net_device *dev, struct list_head *head) }EXPORT_SYMBOL_GPL (ipvlan_link_delete); +static void ipvlan_dev_free (struct net_device *dev) +{ +struct ipvl_dev *ipvlan =netdev_priv (dev); + +/* Get rid of the ipvlan's reference to phy_dev */ +dev_put (ipvlan->phy_dev); +} +void ipvlan_link_setup (struct net_device *dev) {ether_setup (dev);
@@ -674,6 +686,7 @@ void ipvlan_link_setup(struct net_device *dev) dev->priv_flags |=IFF_UNICAST_FLT |IFF_NO_QUEUE ; dev->netdev_ops = &ipvlan_netdev_ops; dev->needs_free_netdev = true; + dev->priv_destructor = ipvlan_dev_free; dev->header_ops = &ipvlan_header_ops; dev->ethtool_ops = &ipvlan_ethtool_ops; }
Signature · 1 line
2.25.1
Jakub Kicinski
maintainerQuoted 10 lines · On Fri, 18 Mar 2022 09:57:47 +0800 Ziyang Xuan wrote:
Add the reference operation to
phy_devof ipvlan to avoid the potential UAF problem under the following known scenario:Someone module puts the
NETDEV_UNREGISTERevent handler to a work, andphy_devis accessed in the work handler. But when the work is excuted,phy_devhas been destroyed because upper ipvlan did not get reference tophy_devcorrectly.That likes as the scenario occurred by
commit563bcbae3ba2("net: vlan: fix a UAF invlan_dev_real_dev()").
There is no equivalent of vlan_dev_real_dev() for ipvlan, AFAICT. The definition of struct ipvl_dev is private to the driver. I don't see how a UAF can happen here.
You should either clearly explain how the bug could happen or clearly state that there is no possibility of the bug for this driver, and the patch is just future proofing.
If the latter is the case we should drop the Fixes tag and prevent this patch from getting backported into stable.
Quoted 2 lines
Fixes 2ad7bf363841 ("ipvlan: Initial check-in of the IPVLAN driver.")Signed-off-by Ziyang Xuan <[email protected]>
Quoted 16 lines · On Fri, 18 Mar 2022 09:57:47 +0800 Ziyang Xuan wrote:
Quoted 10 lines
Add the reference operation to
phy_devof ipvlan to avoid the potential UAF problem under the following known scenario:Someone module puts the
NETDEV_UNREGISTERevent handler to a work, andphy_devis accessed in the work handler. But when the work is excuted,phy_devhas been destroyed because upper ipvlan did not get reference tophy_devcorrectly.That likes as the scenario occurred by
commit563bcbae3ba2("net: vlan: fix a UAF invlan_dev_real_dev()").There is no equivalent of
vlan_dev_real_dev()for ipvlan, AFAICT. The definition of structipvl_devis private to the driver. I don't see how a UAF can happen here.
Quoted 7 lines
You should either clearly explain how the bug could happen or clearly state that there is no possibility of the bug for this driver, and the patch is just future proofing.
If the latter is the case we should drop the Fixes tag and prevent this patch from getting backported into stable.
It is to prevent ipvlan from occurring the similar problem in the future. For now, there is not way to access phy_dev outside ipvlan module as vlan real_dev using vlan_dev_real_dev(). I will make this clear.
I think it is necessary to protect lower netdevice using the feature of netdevice refcnt. So I do it.
Thank you for your valuable opinions! I believe I can do more and more better with your help.
Quoted 4 lines
Quoted 2 lines
Fixes 2ad7bf363841 ("ipvlan: Initial check-in of the IPVLAN driver.")Signed-off-by Ziyang Xuan <[email protected]>.
PATCH v2 2/3 net: ipvlan: add net device refcount tracker
In reply to Ziyang XuanAdd net device refcount tracker to ipvlan.
--- drivers/net/ipvlan/ipvlan.h | 1 + drivers/net/ipvlan/ipvlan_main.c | 4 ++-- 2 files changed, 3 insertions(+), 2 deletions(-)
drivers/net/ipvlan/ipvlan.h +1 -0
@@ -64,6 +64,7 @@ struct ipvl_dev {struct list_head pnode;struct ipvl_port *port;struct net_device *phy_dev; +netdevice_tracker dev_tracker;struct list_head addrs;struct ipvl_pcpu_stats __percpu *pcpu_stats;DECLARE_BITMAP (mac_filters,IPVLAN_MAC_FILTER_SIZE );
drivers/net/ipvlan/ipvlan_main.c +2 -2
@@ -160,7 +160,7 @@ static int ipvlan_init(struct net_device *dev) port->count +=1 ;/* Get ipvlan's reference to phy_dev */ -dev_hold (phy_dev); +dev_hold_track (phy_dev, &ipvlan->dev_tracker,GFP_KERNEL );return 0 ; }
@@ -674,7 +674,7 @@ static void ipvlan_dev_free(struct net_device *dev)struct ipvl_dev *ipvlan =netdev_priv (dev);/* Get rid of the ipvlan's reference to phy_dev */ -dev_put (ipvlan->phy_dev); +dev_put_track (ipvlan->phy_dev, &ipvlan->dev_tracker); }void ipvlan_link_setup (struct net_device *dev)
Signature · 1 line
2.25.1
PATCH v2 3/3 net: ipvtap: fix error comments
In reply to Ziyang XuanUse "macvlan" comment inappropriately in ipvtap module. Fix them with "ipvlan" comment.
--- drivers/net/ipvlan/ipvtap.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/net/ipvlan/ipvtap.c +2 -2
@@ -83,7 +83,7 @@ static int ipvtap_newlink(struct net *src_net, struct net_device *dev,INIT_LIST_HEAD (&vlantap->tap.queue_list); -/* Since macvlan supports all offloads by default, make +/* Since ipvlan supports all offloads by default, make * tap support all offloads also. */ vlantap->tap.tap_features =TUN_OFFLOADS ;
@@ -95,7 +95,7 @@ static int ipvtap_newlink(struct net *src_net, struct net_device *dev,if (err)return err; -/* Don't put anything that may fail after macvlan_common_newlink +/* Don't put anything that may fail after ipvlan_link_new * because we can't undo what it does. */ err =ipvlan_link_new (src_net, dev, tb, data, extack);
Signature · 1 line
2.25.1