[PATCH net] tun: free page on short-frame rejection in tun_xdp_one()

[PATCH net] tun: free page on short-frame rejection in tun_xdp_one()

[Weiming Shi]

tun_xdp_one() returns -EINVAL on a frame shorter than ETH_HLEN without
freeing the page that vhost_net_build_xdp() allocated for it.
tun_sendmsg() discards that -EINVAL and still returns total_len, so
vhost_tx_batch() takes the success path and never frees the page; each
short frame in a batch leaks one page-frag chunk.

A local process that can open /dev/net/tun and /dev/vhost-net can hit
this path: it attaches a tun/tap device as the vhost-net backend and
feeds TX descriptors whose length minus the virtio-net header is below
ETH_HLEN. Each kick leaks the page-frag chunks for that batch, and a
tight submission loop exhausts host memory and triggers an OOM panic.
Free the page before returning -EINVAL, matching the XDP-program error
path in the same function.

Fixes: 049584807f1d ("tun: add missing verification for short frame")
Reported-by: Xiang Mei <xmei5@asu.edu>
Assisted-by: Claude:claude-opus-4-7
Signed-off-by: Weiming Shi <bestswngs@gmail.com>
---
 drivers/net/tun.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/drivers/net/tun.c b/drivers/net/tun.c
index b183189f1853..f594360d66d6 100644
--- a/drivers/net/tun.c
+++ b/drivers/net/tun.c
@@ -2394,8 +2394,10 @@ static int tun_xdp_one(struct tun_struct *tun,
 	bool skb_xdp = false;
 	struct page *page;
 
-	if (unlikely(datasize < ETH_HLEN))
+	if (unlikely(datasize < ETH_HLEN)) {
+		put_page(virt_to_head_page(xdp->data));
 		return -EINVAL;
+	}
 
 	xdp_prog = rcu_dereference(tun->xdp_prog);
 	if (xdp_prog) {
-- 
2.43.0

Re: [PATCH net] tun: free page on short-frame rejection in tun_xdp_one()

[Willem de Bruijn]

Weiming Shi wrote:
> tun_xdp_one() returns -EINVAL on a frame shorter than ETH_HLEN without
> freeing the page that vhost_net_build_xdp() allocated for it.
> tun_sendmsg() discards that -EINVAL and still returns total_len, so
> vhost_tx_batch() takes the success path and never frees the page; each
> short frame in a batch leaks one page-frag chunk.
> 
> A local process that can open /dev/net/tun and /dev/vhost-net can hit
> this path: it attaches a tun/tap device as the vhost-net backend and
> feeds TX descriptors whose length minus the virtio-net header is below
> ETH_HLEN. Each kick leaks the page-frag chunks for that batch, and a
> tight submission loop exhausts host memory and triggers an OOM panic.
> Free the page before returning -EINVAL, matching the XDP-program error
> path in the same function.
> 
> Fixes: 049584807f1d ("tun: add missing verification for short frame")
> Reported-by: Xiang Mei <xmei5@asu.edu>
> Assisted-by: Claude:claude-opus-4-7
> Signed-off-by: Weiming Shi <bestswngs@gmail.com>
> ---
>  drivers/net/tun.c | 4 +++-
>  1 file changed, 3 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/net/tun.c b/drivers/net/tun.c
> index b183189f1853..f594360d66d6 100644
> --- a/drivers/net/tun.c
> +++ b/drivers/net/tun.c
> @@ -2394,8 +2394,10 @@ static int tun_xdp_one(struct tun_struct *tun,
>  	bool skb_xdp = false;
>  	struct page *page;
>  
> -	if (unlikely(datasize < ETH_HLEN))
> +	if (unlikely(datasize < ETH_HLEN)) {
> +		put_page(virt_to_head_page(xdp->data));
>  		return -EINVAL;
> +	}

Make sense, thanks.

The error path from tun_xdp_act does the same. And the default: label
used to too, before a batching optimization was introduced.

Is the same then also missing if build_skb fails?

Re: [PATCH net] tun: free page on short-frame rejection in tun_xdp_one()

[Dongli Zhang]

On 2026-05-20 5:05 PM, Willem de Bruijn wrote:
> Weiming Shi wrote:
>> tun_xdp_one() returns -EINVAL on a frame shorter than ETH_HLEN without
>> freeing the page that vhost_net_build_xdp() allocated for it.
>> tun_sendmsg() discards that -EINVAL and still returns total_len, so
>> vhost_tx_batch() takes the success path and never frees the page; each
>> short frame in a batch leaks one page-frag chunk.
>>
>> A local process that can open /dev/net/tun and /dev/vhost-net can hit
>> this path: it attaches a tun/tap device as the vhost-net backend and
>> feeds TX descriptors whose length minus the virtio-net header is below
>> ETH_HLEN. Each kick leaks the page-frag chunks for that batch, and a
>> tight submission loop exhausts host memory and triggers an OOM panic.
>> Free the page before returning -EINVAL, matching the XDP-program error
>> path in the same function.
>>
>> Fixes: 049584807f1d ("tun: add missing verification for short frame")
>> Reported-by: Xiang Mei <xmei5@asu.edu>
>> Assisted-by: Claude:claude-opus-4-7
>> Signed-off-by: Weiming Shi <bestswngs@gmail.com>
>> ---
>>  drivers/net/tun.c | 4 +++-
>>  1 file changed, 3 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/net/tun.c b/drivers/net/tun.c
>> index b183189f1853..f594360d66d6 100644
>> --- a/drivers/net/tun.c
>> +++ b/drivers/net/tun.c
>> @@ -2394,8 +2394,10 @@ static int tun_xdp_one(struct tun_struct *tun,
>>  	bool skb_xdp = false;
>>  	struct page *page;
>>  
>> -	if (unlikely(datasize < ETH_HLEN))
>> +	if (unlikely(datasize < ETH_HLEN)) {
>> +		put_page(virt_to_head_page(xdp->data));
>>  		return -EINVAL;
>> +	}
> 
> Make sense, thanks.
> 
> The error path from tun_xdp_act does the same. And the default: label
> used to too, before a batching optimization was introduced.
> 
> Is the same then also missing if build_skb fails?
> 

I also agree that we may need to handle build_skb() failure.

In addition, I think we may need this fix for tap_get_user_xdp() as well.

Thank you very much!

Dongli Zhang

Re: [PATCH net] tun: free page on short-frame rejection in tun_xdp_one()

[Willem de Bruijn]

Dongli Zhang wrote:
> 
> 
> On 2026-05-20 5:05 PM, Willem de Bruijn wrote:
> > Weiming Shi wrote:
> >> tun_xdp_one() returns -EINVAL on a frame shorter than ETH_HLEN without
> >> freeing the page that vhost_net_build_xdp() allocated for it.
> >> tun_sendmsg() discards that -EINVAL and still returns total_len, so
> >> vhost_tx_batch() takes the success path and never frees the page; each
> >> short frame in a batch leaks one page-frag chunk.
> >>
> >> A local process that can open /dev/net/tun and /dev/vhost-net can hit
> >> this path: it attaches a tun/tap device as the vhost-net backend and
> >> feeds TX descriptors whose length minus the virtio-net header is below
> >> ETH_HLEN. Each kick leaks the page-frag chunks for that batch, and a
> >> tight submission loop exhausts host memory and triggers an OOM panic.
> >> Free the page before returning -EINVAL, matching the XDP-program error
> >> path in the same function.
> >>
> >> Fixes: 049584807f1d ("tun: add missing verification for short frame")
> >> Reported-by: Xiang Mei <xmei5@asu.edu>
> >> Assisted-by: Claude:claude-opus-4-7
> >> Signed-off-by: Weiming Shi <bestswngs@gmail.com>
> >> ---
> >>  drivers/net/tun.c | 4 +++-
> >>  1 file changed, 3 insertions(+), 1 deletion(-)
> >>
> >> diff --git a/drivers/net/tun.c b/drivers/net/tun.c
> >> index b183189f1853..f594360d66d6 100644
> >> --- a/drivers/net/tun.c
> >> +++ b/drivers/net/tun.c
> >> @@ -2394,8 +2394,10 @@ static int tun_xdp_one(struct tun_struct *tun,
> >>  	bool skb_xdp = false;
> >>  	struct page *page;
> >>  
> >> -	if (unlikely(datasize < ETH_HLEN))
> >> +	if (unlikely(datasize < ETH_HLEN)) {
> >> +		put_page(virt_to_head_page(xdp->data));
> >>  		return -EINVAL;
> >> +	}
> > 
> > Make sense, thanks.
> > 
> > The error path from tun_xdp_act does the same. And the default: label
> > used to too, before a batching optimization was introduced.
> > 
> > Is the same then also missing if build_skb fails?
> > 
> 
> I also agree that we may need to handle build_skb() failure.

Thanks. Fine to defer to a separate patch btw. Either way.
 
> In addition, I think we may need this fix for tap_get_user_xdp() as well.
>
> Thank you very much!
> 
> Dongli Zhang