[net] qmi_wwan: Add missing skb_reset_mac_header-call

Message ID 20171107124756.3029-1-kristian.evensen@gmail.com
State Accepted
Delegated to: David Miller
Headers show
Series
  • [net] qmi_wwan: Add missing skb_reset_mac_header-call
Related show

Commit Message

Kristian Evensen Nov. 7, 2017, 12:47 p.m.
When we receive a packet on a QMI device in raw IP mode, we should call
skb_reset_mac_header() to ensure that skb->mac_header contains a valid
offset in the packet. While it shouldn't really matter, the packets have
no MAC header and the interface is configured as-such, it seems certain
parts of the network stack expects a "good" value in skb->mac_header.

Without the skb_reset_mac_header() call added in this patch, for example
shaping traffic (using tc) triggers the following oops on the first
received packet:

[  303.642957] skbuff: skb_under_panic: text:8f137918 len:177 put:67 head:8e4b0f00 data:8e4b0eff tail:0x8e4b0fb0 end:0x8e4b1520 dev:wwan0
[  303.655045] Kernel bug detected[#1]:
[  303.658622] CPU: 1 PID: 1002 Comm: logd Not tainted 4.9.58 #0
[  303.664339] task: 8fdf05e0 task.stack: 8f15c000
[  303.668844] $ 0   : 00000000 00000001 0000007a 00000000
[  303.674062] $ 4   : 8149a2fc 8149a2fc 8149ce20 00000000
[  303.679284] $ 8   : 00000030 3878303a 31623465 20303235
[  303.684510] $12   : ded731e3 2626a277 00000000 03bd0000
[  303.689747] $16   : 8ef62b40 00000043 8f137918 804db5fc
[  303.694978] $20   : 00000001 00000004 8fc13800 00000003
[  303.700215] $24   : 00000001 8024ab10
[  303.705442] $28   : 8f15c000 8fc19cf0 00000043 802cc920
[  303.710664] Hi    : 00000000
[  303.713533] Lo    : 74e58000
[  303.716436] epc   : 802cc920 skb_panic+0x58/0x5c
[  303.721046] ra    : 802cc920 skb_panic+0x58/0x5c
[  303.725639] Status: 11007c03 KERNEL EXL IE
[  303.729823] Cause : 50800024 (ExcCode 09)
[  303.733817] PrId  : 0001992f (MIPS 1004Kc)
[  303.737892] Modules linked in: rt2800pci rt2800mmio rt2800lib qcserial ppp_async option usb_wwan rt2x00pci rt2x00mmio rt2x00lib rndis_host qmi_wwan ppp_generic nf_nat_pptp nf_conntrack_pptp nf_conntrack_ipv6 mt76x2i
Process logd (pid: 1002, threadinfo=8f15c000, task=8fdf05e0, tls=77b3eee4)
[  303.962509] Stack : 00000000 80408990 8f137918 000000b1 00000043 8e4b0f00 8e4b0eff 8e4b0fb0
[  303.970871]         8e4b1520 8fec1800 00000043 802cd2a4 6e000045 00000043 00000000 8ef62000
[  303.979219]         8eef5d00 8ef62b40 8fea7300 8f137918 00000000 00000000 0002bb01 793e5664
[  303.987568]         8ef08884 00000001 8fea7300 00000002 8fc19e80 8eef5d00 00000006 00000003
[  303.995934]         00000000 8030ba90 00000003 77ab3fd0 8149dc80 8004d1bc 8f15c000 8f383700
[  304.004324]         ...
[  304.006767] Call Trace:
[  304.009241] [<802cc920>] skb_panic+0x58/0x5c
[  304.013504] [<802cd2a4>] skb_push+0x78/0x90
[  304.017783] [<8f137918>] 0x8f137918
[  304.021269] Code: 00602825  0c02a3b4  24842888 <000c000d> 8c870060  8c8200a0  0007382b  00070336  8c88005c
[  304.031034]
[  304.032805] ---[ end trace b778c482b3f0bda9 ]---
[  304.041384] Kernel panic - not syncing: Fatal exception in interrupt
[  304.051975] Rebooting in 3 seconds..

While the oops is for a 4.9-kernel, I was able to trigger the same oops with
net-next as of yesterday.

Signed-off-by: Kristian Evensen <kristian.evensen@gmail.com>
---
 drivers/net/usb/qmi_wwan.c | 1 +
 1 file changed, 1 insertion(+)

Comments

Bjørn Mork Nov. 7, 2017, 1:02 p.m. | #1
Kristian Evensen <kristian.evensen@gmail.com> writes:

> When we receive a packet on a QMI device in raw IP mode, we should call
> skb_reset_mac_header() to ensure that skb->mac_header contains a valid
> offset in the packet. While it shouldn't really matter, the packets have
> no MAC header and the interface is configured as-such, it seems certain
> parts of the network stack expects a "good" value in skb->mac_header.

Thanks.  Yes, I see that the tun driver also does this.

Acked-by: Bjørn Mork <bjorn@mork.no>

And his should probably go to stable as well? with a

Fixes: 32f7adf633b9 ("net: qmi_wwan: support "raw IP" mode")
Kristian Evensen Nov. 7, 2017, 2:32 p.m. | #2
Hei,

On Tue, Nov 7, 2017 at 2:02 PM, Bjørn Mork <bjorn@mork.no> wrote:
> And his should probably go to stable as well? with a
>
> Fixes: 32f7adf633b9 ("net: qmi_wwan: support "raw IP" mode")

Yes, I forgot to add the fixes-tag + comment about stable. Should I submit a v2?

-Kristian
David Miller Nov. 8, 2017, 7:12 a.m. | #3
From: Kristian Evensen <kristian.evensen@gmail.com>
Date: Tue,  7 Nov 2017 13:47:56 +0100

> When we receive a packet on a QMI device in raw IP mode, we should call
> skb_reset_mac_header() to ensure that skb->mac_header contains a valid
> offset in the packet. While it shouldn't really matter, the packets have
> no MAC header and the interface is configured as-such, it seems certain
> parts of the network stack expects a "good" value in skb->mac_header.
> 
> Without the skb_reset_mac_header() call added in this patch, for example
> shaping traffic (using tc) triggers the following oops on the first
> received packet:
 ...
> While the oops is for a 4.9-kernel, I was able to trigger the same oops with
> net-next as of yesterday.
> 
> Signed-off-by: Kristian Evensen <kristian.evensen@gmail.com>

Applied, with Fixes: tag added, and queued up for -stable.

Thanks!

Patch

diff --git a/drivers/net/usb/qmi_wwan.c b/drivers/net/usb/qmi_wwan.c
index db7279d5b250..5b22645c7c8c 100644
--- a/drivers/net/usb/qmi_wwan.c
+++ b/drivers/net/usb/qmi_wwan.c
@@ -499,6 +499,7 @@  static int qmi_wwan_rx_fixup(struct usbnet *dev, struct sk_buff *skb)
 		return 1;
 	}
 	if (rawip) {
+		skb_reset_mac_header(skb);
 		skb->dev = dev->net; /* normally set by eth_type_trans */
 		skb->protocol = proto;
 		return 1;