Skip to contentSkip to conversation

tun: free page on short-frame rejection in tun_xdp_one()

PATCH Applied Weiming Shi 4 participants 8 messages May 20 – May 22

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 tun: free page on short-frame rejection in tun_xdp_one()

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")
Assisted-by Claude:claude-opus-4-7
---
 drivers/net/tun.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
drivers/net/tun.c +3 -1
index b183189f1853..f594360d66d6 100644
@@ -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) {
Signature · 1 line
2.43.0
maintainer
#1 lore raw
Quoted 27 lines · 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")
Assisted-by Claude:claude-opus-4-7
---
 drivers/net/tun.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
drivers/net/tun.c +0 -0
index b183189f1853..f594360d66d6 100644

drivers/net/tun.c static int tun_xdp_one(struct tun_struct *tun, · line 2400

Quoted 8 lines
drivers/net/tun.c +3 -1
@@ -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?

#2 lore raw
Quoted 39 lines · Willem de Bruijn
Quoted 27 lines · 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")
Assisted-by Claude:claude-opus-4-7
---
 drivers/net/tun.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
drivers/net/tun.c +0 -0
index b183189f1853..f594360d66d6 100644

drivers/net/tun.c static int tun_xdp_one(struct tun_struct *tun, · line 2400

Quoted 8 lines
drivers/net/tun.c +3 -1
@@ -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.

Quoted 5 lines

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

maintainer
#3 lore raw
Quoted 42 lines · Dongli Zhang
Quoted 39 lines · Willem de Bruijn
Quoted 27 lines · 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")
Assisted-by Claude:claude-opus-4-7
---
 drivers/net/tun.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
drivers/net/tun.c +0 -0
index b183189f1853..f594360d66d6 100644

drivers/net/tun.c static int tun_xdp_one(struct tun_struct *tun, · line 2400

Quoted 8 lines
drivers/net/tun.c +3 -1
@@ -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.

Quoted 7 lines
Quoted 5 lines

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.

Quoted 5 lines

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

Thank you very much!

Dongli Zhang

#4 lore raw
Quoted 46 lines · Willem de Bruijn
Quoted 45 lines · Dongli Zhang
Quoted 42 lines · Willem de Bruijn
Quoted 27 lines · 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")
Assisted-by Claude:claude-opus-4-7
---
 drivers/net/tun.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
drivers/net/tun.c +0 -0
index b183189f1853..f594360d66d6 100644

drivers/net/tun.c static int tun_xdp_one(struct tun_struct *tun, · line 2400

Quoted 8 lines
drivers/net/tun.c +3 -1
@@ -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.

Quoted 14 lines
Quoted 4 lines
Quoted 2 lines

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.

Quoted 5 lines

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

Thank you very much!

Dongli Zhang

Hi,

I have sent a patch to fix the same issue and the patch is under review now. Please take a look if you have time, thanks!

https://lore.kernel.org/all/[email protected]/
https://lore.kernel.org/all/[email protected]/

Reviewed-by
#5 lore raw
Quoted 61 lines · Weiming Shi
Quoted 46 lines · Willem de Bruijn
Quoted 45 lines · Dongli Zhang
Quoted 42 lines · Willem de Bruijn
Quoted 27 lines · 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")
Assisted-by Claude:claude-opus-4-7
---
 drivers/net/tun.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
drivers/net/tun.c +0 -0
index b183189f1853..f594360d66d6 100644

drivers/net/tun.c static int tun_xdp_one(struct tun_struct *tun, · line 2400

Quoted 8 lines
drivers/net/tun.c +3 -1
@@ -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.

Quoted 14 lines
Quoted 4 lines
Quoted 2 lines

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.

Quoted 5 lines

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

Thank you very much!

Dongli Zhang

Quoted 8 lines

Hi,

I have sent a patch to fix the same issue and the patch is under review now. Please take a look if you have time, thanks!

https://urldefense.com/v3/__https://lore.kernel.org/all/[email protected]/__;!!ACWV5N9M2RV99hQ!IDJKNE6zGJJ7hjjCFU8OviGWfh4A_mNZAhrTowR0Kb4p5bm4JmoCmEI6_GfOXFjcxd2ctJ3pbcJypErFgC4-$
https://urldefense.com/v3/__https://lore.kernel.org/all/[email protected]/__;!!ACWV5N9M2RV99hQ!IDJKNE6zGJJ7hjjCFU8OviGWfh4A_mNZAhrTowR0Kb4p5bm4JmoCmEI6_GfOXFjcxd2ctJ3pbcJypMyOq4Op$

The bug fix applies to both TAP and TUN.

TAP has a single patch, while TUN has two separate patches. I'm not sure which approach is better.

[PATCH net] tap: free page on error paths in tap_get_user_xdp()
https://lore.kernel.org/netdev/[email protected]

[PATCH net] tun: free page on short-frame rejection in tun_xdp_one()
https://lore.kernel.org/all/[email protected]

[PATCH net] tun: free page on build_skb failure in tun_xdp_one()
https://lore.kernel.org/netdev/[email protected]/

The bug fix looks good to me.

+ Michael for FYI regarding vhost as well.

Thank you very much!

Dongli Zhang

maintainer Reviewed-by on v1 1/1
#6 lore raw
In reply to Weiming Shi
Quoted 14 lines · 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.

Quoted 4 lines
Fixes 049584807f1d ("tun: add missing verification for short frame")
Assisted-by Claude:claude-opus-4-7
#7 lore raw
In reply to Weiming Shi
This patch was applied to netdev/net.git (main) by Jakub Kicinski <[email protected]>:

Hello:

This patch was applied to netdev/net.git (main)
by Jakub Kicinski <[email protected]>:

Quoted 14 lines · On Wed, 20 May 2026 09:00:21 -0700 you 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.

Quoted 1 line

[...]

Here is the summary with links:
- [net] tun: free page on short-frame rejection in tun_xdp_one()

    https://git.kernel.org/netdev/net/c/f4feb1e20058

You are awesome, thank you!

Signature · 2 lines
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html