* [RFC PATCH] tuntap: Fix for a race in accessing numqueue
@ 2014-01-18 0:26 Dominic Curran
2014-01-18 1:30 ` Eric Dumazet
2014-01-18 17:43 ` Sergei Shtylyov
0 siblings, 2 replies; 3+ messages in thread
From: Dominic Curran @ 2014-01-18 0:26 UTC (permalink / raw)
To: netdev; +Cc: Dominic Curran, Jason Wang, Maxim Krasnyansky
A patch for fixing a race between queue selection and changing queues
was introduced in commit 92bb73ea2c434618a68a5.
The fix was to prevent the driver from re-reading the tun->numqueues
more than once within tun_select_queue().
We have been experiancing 'Divide-by-zero' errors in
tun_net_xmit() since we moved from 3.6 to 3.10, and believe that they
come from a simular source where the value of tun->numqueues changes
to zero between the first and second read of tun->numqueues.
Signed-off-by: Dominic Curran <dominic.curran@citrix.com>
Cc: Jason Wang <jasowang@redhat.com>
Cc: Maxim Krasnyansky <maxk@qualcomm.com>
---
drivers/net/tun.c | 6 ++++--
1 file changed, 4 insertions(+), 2 deletions(-)
Index: net-next/drivers/net/tun.c
===================================================================
--- net-next.orig/drivers/net/tun.c 2014-01-17 23:41:54.000000000 +0000
+++ net-next/drivers/net/tun.c 2014-01-17 23:51:55.000000000 +0000
@@ -738,12 +738,14 @@ static netdev_tx_t tun_net_xmit(struct s
struct tun_struct *tun = netdev_priv(dev);
int txq = skb->queue_mapping;
struct tun_file *tfile;
+ u32 numqueues = 0;
rcu_read_lock();
tfile = rcu_dereference(tun->tfiles[txq]);
+ numqueues = ACCESS_ONCE(tun->numqueues);
/* Drop packet if interface is not attached */
- if (txq >= tun->numqueues)
+ if (txq >= numqueues)
goto drop;
if (tun->numqueues == 1) {
@@ -780,7 +782,7 @@ static netdev_tx_t tun_net_xmit(struct s
* number of queues.
*/
if (skb_queue_len(&tfile->socket.sk->sk_receive_queue)
- >= dev->tx_queue_len / tun->numqueues)
+ >= dev->tx_queue_len / numqueues)
goto drop;
if (unlikely(skb_orphan_frags(skb, GFP_ATOMIC)))
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [RFC PATCH] tuntap: Fix for a race in accessing numqueue
2014-01-18 0:26 [RFC PATCH] tuntap: Fix for a race in accessing numqueue Dominic Curran
@ 2014-01-18 1:30 ` Eric Dumazet
2014-01-18 17:43 ` Sergei Shtylyov
1 sibling, 0 replies; 3+ messages in thread
From: Eric Dumazet @ 2014-01-18 1:30 UTC (permalink / raw)
To: Dominic Curran; +Cc: netdev, Jason Wang, Maxim Krasnyansky
On Sat, 2014-01-18 at 00:26 +0000, Dominic Curran wrote:
> A patch for fixing a race between queue selection and changing queues
> was introduced in commit 92bb73ea2c434618a68a5.
>
> The fix was to prevent the driver from re-reading the tun->numqueues
> more than once within tun_select_queue().
>
> We have been experiancing 'Divide-by-zero' errors in
> tun_net_xmit() since we moved from 3.6 to 3.10, and believe that they
> come from a simular source where the value of tun->numqueues changes
> to zero between the first and second read of tun->numqueues.
>
> Signed-off-by: Dominic Curran <dominic.curran@citrix.com>
> Cc: Jason Wang <jasowang@redhat.com>
> Cc: Maxim Krasnyansky <maxk@qualcomm.com>
>
> ---
> drivers/net/tun.c | 6 ++++--
> 1 file changed, 4 insertions(+), 2 deletions(-)
>
> Index: net-next/drivers/net/tun.c
> ===================================================================
> --- net-next.orig/drivers/net/tun.c 2014-01-17 23:41:54.000000000 +0000
> +++ net-next/drivers/net/tun.c 2014-01-17 23:51:55.000000000 +0000
> @@ -738,12 +738,14 @@ static netdev_tx_t tun_net_xmit(struct s
> struct tun_struct *tun = netdev_priv(dev);
> int txq = skb->queue_mapping;
> struct tun_file *tfile;
> + u32 numqueues = 0;
>
> rcu_read_lock();
> tfile = rcu_dereference(tun->tfiles[txq]);
> + numqueues = ACCESS_ONCE(tun->numqueues);
>
> /* Drop packet if interface is not attached */
> - if (txq >= tun->numqueues)
> + if (txq >= numqueues)
> goto drop;
>
> if (tun->numqueues == 1) {
This is net-next tree, not net tree
(This part was added in commit 9bc8893937c83)
You might change this "tun->numqueues" as well.
> @@ -780,7 +782,7 @@ static netdev_tx_t tun_net_xmit(struct s
> * number of queues.
> */
> if (skb_queue_len(&tfile->socket.sk->sk_receive_queue)
> - >= dev->tx_queue_len / tun->numqueues)
> + >= dev->tx_queue_len / numqueues)
> goto drop;
Could use a multiply instead....
if (skb_queue_len(&tfile->socket.sk->sk_receive_queue) * numqueues >=
dev->tx_queue_len )
Please submit this patch on net tree, since its a bug fix.
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [RFC PATCH] tuntap: Fix for a race in accessing numqueue
2014-01-18 0:26 [RFC PATCH] tuntap: Fix for a race in accessing numqueue Dominic Curran
2014-01-18 1:30 ` Eric Dumazet
@ 2014-01-18 17:43 ` Sergei Shtylyov
1 sibling, 0 replies; 3+ messages in thread
From: Sergei Shtylyov @ 2014-01-18 17:43 UTC (permalink / raw)
To: Dominic Curran, netdev; +Cc: Jason Wang, Maxim Krasnyansky
Hello.
On 18-01-2014 4:26, Dominic Curran wrote:
> A patch for fixing a race between queue selection and changing queues
> was introduced in commit 92bb73ea2c434618a68a5.
Please also specify that commit's summary line in parens.
> The fix was to prevent the driver from re-reading the tun->numqueues
> more than once within tun_select_queue().
> We have been experiancing 'Divide-by-zero' errors in
> tun_net_xmit() since we moved from 3.6 to 3.10, and believe that they
> come from a simular source where the value of tun->numqueues changes
> to zero between the first and second read of tun->numqueues.
> Signed-off-by: Dominic Curran <dominic.curran@citrix.com>
> Cc: Jason Wang <jasowang@redhat.com>
> Cc: Maxim Krasnyansky <maxk@qualcomm.com>
WBR, Sergei
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2014-01-18 17:43 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2014-01-18 0:26 [RFC PATCH] tuntap: Fix for a race in accessing numqueue Dominic Curran
2014-01-18 1:30 ` Eric Dumazet
2014-01-18 17:43 ` Sergei Shtylyov
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).