[PATCH v2] net/smc: hold a reference on net_device returned by pnet_find_base_ndev()

[PATCH v2] net/smc: hold a reference on net_device returned by pnet_find_base_ndev()

[Atharva Vartak]

pnet_find_base_ndev() resolves the base net_device for stacked devices
under RTNL, then drops the lock and returns the raw pointer without
taking a reference.  All three callers dereference the pointer after
RTNL has been released, creating a use-after-free window if the device
is concurrently unregistered.

Take a reference with netdev_hold() before dropping RTNL and add the
matching netdev_put() in every return path of the three callers:

  - smc_pnet_add_eth()
  - smc_pnet_find_roce_by_pnetid()
  - smc_pnet_find_ism_by_pnetid()

Fixes: 0afff91c6f5e ("net/smc: add pnetid support")
Fixes: 1619f770589a ("net/smc: add pnetid support for SMC-D and ISM")
Cc: "D. Wythe" <alibuda@linux.alibaba.com>
Cc: Dust Li <dust.li@linux.alibaba.com>
Cc: Sidraya Jayagond <sidraya@linux.ibm.com>
Cc: Mahanta Jambigi <mjambigi@linux.ibm.com>
Cc: Tony Lu <tonylu@linux.alibaba.com>
Cc: Wen Gu <guwen@linux.alibaba.com>
Cc: Willy Tarreau <w@1wt.eu>
Cc: linux-rdma@vger.kernel.org
Cc: linux-s390@vger.kernel.org
Cc: netdev@vger.kernel.org
Signed-off-by: Atharva Vartak <atharva.a.vartak@gmail.com>
---
v2: use netdev_hold()/netdev_put() with a netdevice_tracker instead of
    the deprecated dev_hold()/dev_put() (Jakub Kicinski).
v1: https://lore.kernel.org/netdev/6aba080c.b933486d.297b72.5042@mx.google.com/

 net/smc/smc_pnet.c | 29 ++++++++++++++++++++++-------
 1 file changed, 22 insertions(+), 7 deletions(-)

diff --git a/net/smc/smc_pnet.c b/net/smc/smc_pnet.c
index ff9c9c35cc2f..61ef1d697eee 100644
--- a/net/smc/smc_pnet.c
+++ b/net/smc/smc_pnet.c
@@ -30,7 +30,8 @@
 #include "smc_core.h"
 
 static struct net_device *__pnet_find_base_ndev(struct net_device *ndev);
-static struct net_device *pnet_find_base_ndev(struct net_device *ndev);
+static struct net_device *pnet_find_base_ndev(struct net_device *ndev,
+					      netdevice_tracker *tracker);
 
 static const struct nla_policy smc_pnet_policy[SMC_PNETID_MAX + 1] = {
 	[SMC_PNETID_NAME] = {
@@ -356,6 +357,7 @@ static int smc_pnet_add_eth(struct smc_pnettable *pnettable, struct net *net,
 	struct smc_pnetentry *tmp_pe, *new_pe;
 	struct net_device *ndev, *base_ndev;
 	u8 ndev_pnetid[SMC_MAX_PNETID_LEN];
+	netdevice_tracker base_tracker;
 	bool new_netdev;
 	int rc;
 
@@ -365,10 +367,13 @@ static int smc_pnet_add_eth(struct smc_pnettable *pnettable, struct net *net,
 	rc = -EEXIST;
 	ndev = dev_get_by_name(net, eth_name);	/* dev_hold() */
 	if (ndev) {
-		base_ndev = pnet_find_base_ndev(ndev);
+		base_ndev = pnet_find_base_ndev(ndev, &base_tracker);
 		if (!smc_pnetid_by_dev_port(base_ndev->dev.parent,
-					    base_ndev->dev_port, ndev_pnetid))
+					    base_ndev->dev_port, ndev_pnetid)) {
+			netdev_put(base_ndev, &base_tracker);
 			goto out_put;
+		}
+		netdev_put(base_ndev, &base_tracker);
 	}
 
 	/* add a new netdev entry to the pnet table if there isn't one */
@@ -945,10 +950,13 @@ static struct net_device *__pnet_find_base_ndev(struct net_device *ndev)
  * (for instance with bonding slaves), just the first device
  * is used to reach a base device.
  */
-static struct net_device *pnet_find_base_ndev(struct net_device *ndev)
+static struct net_device *pnet_find_base_ndev(struct net_device *ndev,
+					      netdevice_tracker *tracker)
 {
 	rtnl_lock();
 	ndev = __pnet_find_base_ndev(ndev);
+	/* keep ndev alive after dropping RTNL, callers must netdev_put() */
+	netdev_hold(ndev, tracker, GFP_KERNEL);
 	rtnl_unlock();
 	return ndev;
 }
@@ -1085,17 +1093,20 @@ static void smc_pnet_find_roce_by_pnetid(struct net_device *ndev,
 {
 	u8 ndev_pnetid[SMC_MAX_PNETID_LEN];
 	struct net_device *base_ndev;
+	netdevice_tracker base_tracker;
 	struct net *net;
 
-	base_ndev = pnet_find_base_ndev(ndev);
+	base_ndev = pnet_find_base_ndev(ndev, &base_tracker);
 	net = dev_net(ndev);
 	if (smc_pnetid_by_dev_port(base_ndev->dev.parent, base_ndev->dev_port,
 				   ndev_pnetid) &&
 	    smc_pnet_find_ndev_pnetid_by_table(base_ndev, ndev_pnetid) &&
 	    smc_pnet_find_ndev_pnetid_by_table(ndev, ndev_pnetid)) {
 		smc_pnet_find_rdma_dev(base_ndev, ini);
+		netdev_put(base_ndev, &base_tracker);
 		return; /* pnetid could not be determined */
 	}
+	netdev_put(base_ndev, &base_tracker);
 	_smc_pnet_find_roce_by_pnetid(ndev_pnetid, ini, NULL, net);
 }
 
@@ -1103,13 +1114,16 @@ static void smc_pnet_find_ism_by_pnetid(struct net_device *ndev,
 					struct smc_init_info *ini)
 {
 	u8 ndev_pnetid[SMC_MAX_PNETID_LEN];
+	netdevice_tracker base_tracker;
 	struct smcd_dev *ismdev;
 
-	ndev = pnet_find_base_ndev(ndev);
+	ndev = pnet_find_base_ndev(ndev, &base_tracker);
 	if (smc_pnetid_by_dev_port(ndev->dev.parent, ndev->dev_port,
 				   ndev_pnetid) &&
-	    smc_pnet_find_ndev_pnetid_by_table(ndev, ndev_pnetid))
+	    smc_pnet_find_ndev_pnetid_by_table(ndev, ndev_pnetid)) {
+		netdev_put(ndev, &base_tracker);
 		return; /* pnetid could not be determined */
+	}
 
 	mutex_lock(&smcd_dev_list.mutex);
 	list_for_each_entry(ismdev, &smcd_dev_list.list, list) {
@@ -1123,6 +1137,7 @@ static void smc_pnet_find_ism_by_pnetid(struct net_device *ndev,
 		}
 	}
 	mutex_unlock(&smcd_dev_list.mutex);
+	netdev_put(ndev, &base_tracker);
 }
 
 /* PNET table analysis for a given sock:
-- 
2.56.0

Re: [PATCH v2] net/smc: hold a reference on net_device returned by pnet_find_base_ndev()

[netdev-bot+sinfo]

Hi!

This is an automated message. This series looks like a fix, but its
commit messages seem to be missing some information:

 - How the issue was discovered, e.g. hit in production, hit during
   development, syzbot report, manual code inspection, LLM or static
   analysis tool scan.

 - Whether the issue was actually triggered, or is only theoretical
   (e.g. found by code inspection). If it was triggered please include
   the symptoms, like the stack trace or error messages.

Please do not repost the series just to address the above. Instead,
reply to this email with the missing information, so that reviewers
can take it into account. If the series needs another revision for
other reasons, please include the information in the commit messages
then.

The evaluation is done by an LLM so it may be wrong, if you think
that is the case please reply and explain.