[PATCH net-next] net: set initial device refcount to 1

From: Eric Dumazet <edumazet@google.com>

When adding CONFIG_PCPU_DEV_REFCNT, I forgot that the
initial net device refcount was 0.

When CONFIG_PCPU_DEV_REFCNT is not set, this means
the first dev_hold() triggers an illegal refcount
operation (addition on 0)

refcount_t: addition on 0; use-after-free.
WARNING: CPU: 0 PID: 1 at lib/refcount.c:25 refcount_warn_saturate+0x128/0x1a4

Fix is to change initial (and final) refcount to be 1.

Also add a missing kerneldoc piece, as reported by
Stephen Rothwell.

Fixes: 919067cc845f ("net: add CONFIG_PCPU_DEV_REFCNT")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Guenter Roeck <redacted>
---
 include/linux/netdevice.h | 1 +
 net/core/dev.c            | 9 ++++++---
 2 files changed, 7 insertions(+), 3 deletions(-)

diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
index 8f003955c485b81210ed56f7e1c24080b4bb46eb..b11c2c1890b2a28ba2d02fc4466380703a12efaf 100644
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h

@@ -1792,6 +1792,7 @@ enum netdev_ml_priv_type {
  *
  *	@proto_down_reason:	reason a netdev interface is held down
  *	@pcpu_refcnt:		Number of references to this device
+ *	@dev_refcnt:		Number of references to this device
  *	@todo_list:		Delayed register/unregister
  *	@link_watch_list:	XXX: need comments on this one
  *

diff --git a/net/core/dev.c b/net/core/dev.c
index be941ed754ac71d0839604bcfdd8ab67c339d27f..95c78279d900796c8a3ed0df59b168d5c5e0e309 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c

@@ -10348,7 +10348,7 @@ static void netdev_wait_allrefs(struct net_device *dev)
 	rebroadcast_time = warning_time = jiffies;
 	refcnt = netdev_refcnt_read(dev);
 
-	while (refcnt != 0) {
+	while (refcnt != 1) {
 		if (time_after(jiffies, rebroadcast_time + 1 * HZ)) {
 			rtnl_lock();
 

@@ -10385,7 +10385,7 @@ static void netdev_wait_allrefs(struct net_device *dev)
 
 		refcnt = netdev_refcnt_read(dev);
 
-		if (refcnt && time_after(jiffies, warning_time + 10 * HZ)) {
+		if (refcnt != 1 && time_after(jiffies, warning_time + 10 * HZ)) {
 			pr_emerg("unregister_netdevice: waiting for %s to become free. Usage count = %d\n",
 				 dev->name, refcnt);
 			warning_time = jiffies;

@@ -10461,7 +10461,7 @@ void netdev_run_todo(void)
 		netdev_wait_allrefs(dev);
 
 		/* paranoia */
-		BUG_ON(netdev_refcnt_read(dev));
+		BUG_ON(netdev_refcnt_read(dev) != 1);
 		BUG_ON(!list_empty(&dev->ptype_all));
 		BUG_ON(!list_empty(&dev->ptype_specific));
 		WARN_ON(rcu_access_pointer(dev->ip_ptr));

@@ -10682,6 +10682,9 @@ struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
 	dev->pcpu_refcnt = alloc_percpu(int);
 	if (!dev->pcpu_refcnt)
 		goto free_dev;
+	dev_hold(dev);
+#else
+	refcount_set(&dev->dev_refcnt, 1);
 #endif
 
 	if (dev_addr_init(dev))

-- 
2.31.0.291.g576ba9dcdaf-goog

On Mon, Mar 22, 2021 at 11:21 AM Eric Dumazet [off-list ref] wrote:

From: Eric Dumazet <edumazet@google.com>

When adding CONFIG_PCPU_DEV_REFCNT, I forgot that the
initial net device refcount was 0.

When CONFIG_PCPU_DEV_REFCNT is not set, this means
the first dev_hold() triggers an illegal refcount
operation (addition on 0)

refcount_t: addition on 0; use-after-free.
WARNING: CPU: 0 PID: 1 at lib/refcount.c:25 refcount_warn_saturate+0x128/0x1a4

Fix is to change initial (and final) refcount to be 1.

Also add a missing kerneldoc piece, as reported by
Stephen Rothwell.

Fixes: 919067cc845f ("net: add CONFIG_PCPU_DEV_REFCNT")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Guenter Roeck <redacted>

The problems observed with next-20210322 are gone with this patch applied.

Tested-by: Guenter Roeck <redacted>

Thanks,
Guenter

From: Eric Dumazet <redacted>
Date: Mon, 22 Mar 2021 11:21:45 -0700

From: Eric Dumazet <edumazet@google.com>

When adding CONFIG_PCPU_DEV_REFCNT, I forgot that the
initial net device refcount was 0.

When CONFIG_PCPU_DEV_REFCNT is not set, this means
the first dev_hold() triggers an illegal refcount
operation (addition on 0)

refcount_t: addition on 0; use-after-free.
WARNING: CPU: 0 PID: 1 at lib/refcount.c:25 refcount_warn_saturate+0x128/0x1a4

Fix is to change initial (and final) refcount to be 1.

Also add a missing kerneldoc piece, as reported by
Stephen Rothwell.

Fixes: 919067cc845f ("net: add CONFIG_PCPU_DEV_REFCNT")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Guenter Roeck <redacted>

This hunk:

 gets rejects in net-next.  Please respin.
 

From: David Miller <davem@davemloft.net>
Date: Mon, 22 Mar 2021 16:55:26 -0700 (PDT)

From: Eric Dumazet <redacted>
Date: Mon, 22 Mar 2021 11:21:45 -0700

This hunk:

quoted

@@ -10682,6 +10682,9 @@ struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
 	dev->pcpu_refcnt = alloc_percpu(int);
 	if (!dev->pcpu_refcnt)
 		goto free_dev;
+	dev_hold(dev);
+#else
+	refcount_set(&dev->dev_refcnt, 1);
 #endif
 
 	if (dev_addr_init(dev))

 gets rejects in net-next.  Please respin.
 

My bad, I was trying to apply it to 'net' mistakedly.

All good now, thanks.

Hello:

This patch was applied to netdev/net-next.git (refs/heads/master):

On Mon, 22 Mar 2021 11:21:45 -0700 you wrote:

From: Eric Dumazet <edumazet@google.com>

When adding CONFIG_PCPU_DEV_REFCNT, I forgot that the
initial net device refcount was 0.

When CONFIG_PCPU_DEV_REFCNT is not set, this means
the first dev_hold() triggers an illegal refcount
operation (addition on 0)

[...]

Here is the summary with links:
  - [net-next] net: set initial device refcount to 1
    https://git.kernel.org/netdev/net-next/c/add2d7363107

You are awesome, thank you!
--
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html