BUG: vnet0 selects TX queue 11, but real number of TX queues is 11

BUG: vnet0 selects TX queue 11, but real number of TX queues is 11

[George Amanakis]

Since 4.20.4 when running a KVM with vhost_net I am seeing in dmesg:
vnet0 selects TX queue 11, but real number of TX queues is 11

The corresponding part in the xml definition of the virtual machine is:
-------8<-------
<interface type='bridge'>
  <mac address='xx:xx:xx:xx:xx:xx'/>
  <source bridge='br0'/>
  <model type='virtio'/>
  <driver name='vhost' queues='12'>
  </driver>
  <address type='pci' domain='0x0000' bus='0x01' slot='0x00' function='0x0'/>
</interface>
-------8<-------

Doing a git-bisect with 4.20.3 last known good, and 4.20.4 as bad, this
commit turned up:
-------8<-------
commit 9ff0436e2c3575ffe64d359fb3b67aee237dc519
Author: Stanislav Fomichev <sdf@google.com>
Date:   Mon Jan 7 13:38:38 2019 -0800

    tun: publish tfile after it's fully initialized

    [ Upstream commit 0b7959b6257322f7693b08a459c505d4938646f2 ]

    BUG: unable to handle kernel NULL pointer dereference at
00000000000000d1
-------8<-------


Applying the following patch corrects it in 4.20.5. Would this be the
correct thing to do?
---
 drivers/net/tun.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/drivers/net/tun.c b/drivers/net/tun.c
index 6658658246d2..e0dc004c6483 100644
--- a/drivers/net/tun.c
+++ b/drivers/net/tun.c
@@ -862,8 +862,6 @@ static int tun_attach(struct tun_struct *tun, struct file *file,
 	if (rtnl_dereference(tun->xdp_prog))
 		sock_set_flag(&tfile->sk, SOCK_XDP);
 
-	tun_set_real_num_queues(tun);
-
 	/* device is allowed to go away first, so no need to hold extra
 	 * refcnt.
 	 */
@@ -875,6 +873,9 @@ static int tun_attach(struct tun_struct *tun, struct file *file,
 	rcu_assign_pointer(tfile->tun, tun);
 	rcu_assign_pointer(tun->tfiles[tun->numqueues], tfile);
 	tun->numqueues++;
+
+	tun_set_real_num_queues(tun);
+
 out:
 	return err;
 }
-- 
2.20.1

Re: BUG: vnet0 selects TX queue 11, but real number of TX queues is 11

[Stanislav Fomichev]

On Tue, Jan 29, 2019 at 6:16 PM George Amanakis <gamanakis@gmail.com> wrote:
>
> Since 4.20.4 when running a KVM with vhost_net I am seeing in dmesg:
> vnet0 selects TX queue 11, but real number of TX queues is 11
>
> The corresponding part in the xml definition of the virtual machine is:
> -------8<-------
> <interface type='bridge'>
>   <mac address='xx:xx:xx:xx:xx:xx'/>
>   <source bridge='br0'/>
>   <model type='virtio'/>
>   <driver name='vhost' queues='12'>
>   </driver>
>   <address type='pci' domain='0x0000' bus='0x01' slot='0x00' function='0x0'/>
> </interface>
> -------8<-------
>
> Doing a git-bisect with 4.20.3 last known good, and 4.20.4 as bad, this
> commit turned up:
> -------8<-------
> commit 9ff0436e2c3575ffe64d359fb3b67aee237dc519
> Author: Stanislav Fomichev <sdf@google.com>
> Date:   Mon Jan 7 13:38:38 2019 -0800
>
>     tun: publish tfile after it's fully initialized
>
>     [ Upstream commit 0b7959b6257322f7693b08a459c505d4938646f2 ]
>
>     BUG: unable to handle kernel NULL pointer dereference at
> 00000000000000d1
> -------8<-------
>
>
> Applying the following patch corrects it in 4.20.5. Would this be the
> correct thing to do?
Ouch, tun_set_real_num_queues uses tun->numqueues internally :-(
Your patch looks good to me, care to do a proper submission (with a Fixes tag)?
I wonder whether we should use it as an opportunity to also do
something like the following (to make it more explicit):

diff --git a/drivers/net/tun.c b/drivers/net/tun.c
index 18656c4094b3..ea9928b3b930 100644
--- a/drivers/net/tun.c
+++ b/drivers/net/tun.c
@@ -632,10 +632,10 @@ static inline bool tun_not_capable(struct tun_struct *tun)
  !ns_capable(net->user_ns, CAP_NET_ADMIN);
 }

-static void tun_set_real_num_queues(struct tun_struct *tun)
+static void tun_set_real_num_queues(struct tun_struct *tun, unsigned int nr)
 {
- netif_set_real_num_tx_queues(tun->dev, tun->numqueues);
- netif_set_real_num_rx_queues(tun->dev, tun->numqueues);
+ netif_set_real_num_tx_queues(tun->dev, nr);
+ netif_set_real_num_rx_queues(tun->dev, nr);
 }

 static void tun_disable_queue(struct tun_struct *tun, struct tun_file *tfile)
@@ -712,7 +712,7 @@ static void __tun_detach(struct tun_file *tfile, bool clean)
  tun_flow_delete_by_queue(tun, tun->numqueues + 1);
  /* Drop read queue */
  tun_queue_purge(tfile);
- tun_set_real_num_queues(tun);
+ tun_set_real_num_queues(tun, tun->numqueues);
  } else if (tfile->detached && clean) {
  tun = tun_enable_queue(tfile);
  sock_put(&tfile->sk);
@@ -866,8 +866,6 @@ static int tun_attach(struct tun_struct *tun,
struct file *file,
  if (rtnl_dereference(tun->xdp_prog))
  sock_set_flag(&tfile->sk, SOCK_XDP);

- tun_set_real_num_queues(tun);
-
  /* device is allowed to go away first, so no need to hold extra
  * refcnt.
  */
@@ -879,6 +877,7 @@ static int tun_attach(struct tun_struct *tun,
struct file *file,
  rcu_assign_pointer(tfile->tun, tun);
  rcu_assign_pointer(tun->tfiles[tun->numqueues], tfile);
  tun->numqueues++;
+ tun_set_real_num_queues(tun, tun->numqueues);
 out:
  return err;
 }


> ---
>  drivers/net/tun.c | 5 +++--
>  1 file changed, 3 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/net/tun.c b/drivers/net/tun.c
> index 6658658246d2..e0dc004c6483 100644
> --- a/drivers/net/tun.c
> +++ b/drivers/net/tun.c
> @@ -862,8 +862,6 @@ static int tun_attach(struct tun_struct *tun, struct file *file,
>         if (rtnl_dereference(tun->xdp_prog))
>                 sock_set_flag(&tfile->sk, SOCK_XDP);
>
> -       tun_set_real_num_queues(tun);
> -
>         /* device is allowed to go away first, so no need to hold extra
>          * refcnt.
>          */
> @@ -875,6 +873,9 @@ static int tun_attach(struct tun_struct *tun, struct file *file,
>         rcu_assign_pointer(tfile->tun, tun);
>         rcu_assign_pointer(tun->tfiles[tun->numqueues], tfile);
>         tun->numqueues++;
> +
> +       tun_set_real_num_queues(tun);
> +
>  out:
>         return err;
>  }
> --
> 2.20.1
>

[PATCH] tun: move the call to tun_set_real_num_queues

[George Amanakis]

Call tun_set_real_num_queues() after the increment of tun->numqueues
since the former depends on it. Otherwise, the number of queues is not
correctly accounted for, which results to warnings similar to:
"vnet0 selects TX queue 11, but real number of TX queues is 11".

Fixes: 0b7959b62573 ("tun: publish tfile after it's fully initialized")
Reported-and-tested-by: George Amanakis <gamanakis@gmail.com>
Signed-off-by: George Amanakis <gamanakis@gmail.com>
Signed-off-by: Stanislav Fomichev <sdf@google.com>
---
 drivers/net/tun.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/drivers/net/tun.c b/drivers/net/tun.c
index 18656c4094b3..fed298c0cb39 100644
--- a/drivers/net/tun.c
+++ b/drivers/net/tun.c
@@ -866,8 +866,6 @@ static int tun_attach(struct tun_struct *tun, struct file *file,
 	if (rtnl_dereference(tun->xdp_prog))
 		sock_set_flag(&tfile->sk, SOCK_XDP);
 
-	tun_set_real_num_queues(tun);
-
 	/* device is allowed to go away first, so no need to hold extra
 	 * refcnt.
 	 */
@@ -879,6 +877,7 @@ static int tun_attach(struct tun_struct *tun, struct file *file,
 	rcu_assign_pointer(tfile->tun, tun);
 	rcu_assign_pointer(tun->tfiles[tun->numqueues], tfile);
 	tun->numqueues++;
+	tun_set_real_num_queues(tun);
 out:
 	return err;
 }
-- 
2.20.1

[PATCH] tun: move the call to tun_set_real_num_queues

[George Amanakis]

Call tun_set_real_num_queues() after the increment of tun->numqueues
since the former depends on it. Otherwise, the number of queues is not
correctly accounted for, which results to warnings similar to:
"vnet0 selects TX queue 11, but real number of TX queues is 11".

Fixes: 0b7959b62573 ("tun: publish tfile after it's fully initialized")
Reported-and-tested-by: George Amanakis <gamanakis@gmail.com>
Signed-off-by: George Amanakis <gamanakis@gmail.com>
Signed-off-by: Stanislav Fomichev <sdf@google.com>
---
 drivers/net/tun.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/drivers/net/tun.c b/drivers/net/tun.c
index 18656c4094b3..fed298c0cb39 100644
--- a/drivers/net/tun.c
+++ b/drivers/net/tun.c
@@ -866,8 +866,6 @@ static int tun_attach(struct tun_struct *tun, struct file *file,
 	if (rtnl_dereference(tun->xdp_prog))
 		sock_set_flag(&tfile->sk, SOCK_XDP);
 
-	tun_set_real_num_queues(tun);
-
 	/* device is allowed to go away first, so no need to hold extra
 	 * refcnt.
 	 */
@@ -879,6 +877,7 @@ static int tun_attach(struct tun_struct *tun, struct file *file,
 	rcu_assign_pointer(tfile->tun, tun);
 	rcu_assign_pointer(tun->tfiles[tun->numqueues], tfile);
 	tun->numqueues++;
+	tun_set_real_num_queues(tun);
 out:
 	return err;
 }
-- 
2.20.1