[PATCH net-next] virtio-net: fix use-after-free in page_to_skb()

[Eric Dumazet]

From: Eric Dumazet <edum...@google.com>

KASAN/syzbot had 4 reports, one of them being:

BUG: KASAN: slab-out-of-bounds in memcpy include/linux/fortify-string.h:191 [inline]
BUG: KASAN: slab-out-of-bounds in page_to_skb+0x5cf/0xb70 drivers/net/virtio_net.c:480
Read of size 12 at addr ffff888014a5f800 by task systemd-udevd/8445

CPU: 0 PID: 8445 Comm: systemd-udevd Not tainted 5.12.0-rc8-next-20210419-syzkaller #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
Call Trace:
<IRQ>
__dump_stack lib/dump_stack.c:79 [inline]
dump_stack+0x141/0x1d7 lib/dump_stack.c:120
print_address_description.constprop.0.cold+0x5b/0x2f8 mm/kasan/report.c:233
__kasan_report mm/kasan/report.c:419 [inline]
kasan_report.cold+0x7c/0xd8 mm/kasan/report.c:436
check_region_inline mm/kasan/generic.c:180 [inline]
kasan_check_range+0x13d/0x180 mm/kasan/generic.c:186
memcpy+0x20/0x60 mm/kasan/shadow.c:65
memcpy include/linux/fortify-string.h:191 [inline]
page_to_skb+0x5cf/0xb70 drivers/net/virtio_net.c:480
receive_mergeable drivers/net/virtio_net.c:1009 [inline]
receive_buf+0x2bc0/0x6250 drivers/net/virtio_net.c:1119
virtnet_receive drivers/net/virtio_net.c:1411 [inline]
virtnet_poll+0x568/0x10b0 drivers/net/virtio_net.c:1516
__napi_poll+0xaf/0x440 net/core/dev.c:6962
napi_poll net/core/dev.c:7029 [inline]
net_rx_action+0x801/0xb40 net/core/dev.c:7116
__do_softirq+0x29b/0x9fe kernel/softirq.c:559
invoke_softirq kernel/softirq.c:433 [inline]
__irq_exit_rcu+0x136/0x200 kernel/softirq.c:637
irq_exit_rcu+0x5/0x20 kernel/softirq.c:649
common_interrupt+0xa4/0xd0 arch/x86/kernel/irq.c:240

Fixes: fb32856b16ad ("virtio-net: page_to_skb() use build_skb when there's sufficient tailroom")
Signed-off-by: Eric Dumazet <edum...@google.com>
Reported-by: syzbot <syzk...@googlegroups.com>
Reported-by: Guenter Roeck <li...@roeck-us.net>
Reported-by: Mat Martineau <mathew.j....@linux.intel.com>
Cc: Xuan Zhuo <xuan...@linux.alibaba.com>
Cc: Jason Wang <jaso...@redhat.com>
Cc: "Michael S. Tsirkin" <m...@redhat.com>
Cc: virtual...@lists.linux-foundation.org
---
drivers/net/virtio_net.c | 5 ++++-
1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
index 8cd76037c72481200ea3e8429e9fdfec005dad85..2e28c04aa6351d2b4016f7d277ce104c4970069d 100644
--- a/drivers/net/virtio_net.c
+++ b/drivers/net/virtio_net.c
@@ -385,6 +385,7 @@ static struct sk_buff *page_to_skb(struct virtnet_info *vi,
struct sk_buff *skb;
struct virtio_net_hdr_mrg_rxbuf *hdr;
unsigned int copy, hdr_len, hdr_padded_len;
+ struct page *page_to_free = NULL;
int tailroom, shinfo_size;
char *p, *hdr_p;

@@ -445,7 +446,7 @@ static struct sk_buff *page_to_skb(struct virtnet_info *vi,
if (len)
skb_add_rx_frag(skb, 0, page, offset, len, truesize);
else
- put_page(page);
+ page_to_free = page;
goto ok;
}

@@ -479,6 +480,8 @@ static struct sk_buff *page_to_skb(struct virtnet_info *vi,
hdr = skb_vnet_hdr(skb);
memcpy(hdr, hdr_p, hdr_len);
}
+ if (page_to_free)
+ put_page(page_to_free);

if (metasize) {
__skb_pull(skb, metasize);
--
2.31.1.368.gbe11c130af-goog

[Michael S. Tsirkin]

Acked-by: Michael S. Tsirkin <m...@redhat.com>

[Guenter Roeck]

Unfortunately that doesn't fix the problem for me. With this patch applied
on top of next-20210419, I still get the same crash as before:

udhcpc: sending discover^M
Unable to handle kernel paging request at virtual address 0000000000000004^M
udhcpc(169): Oops -1^M
pc = [<0000000000000004>] ra = [<fffffc0000b8c5b8>] ps = 0000 Not tainted^M
pc is at 0x4^M
ra is at napi_gro_receive+0x68/0x150^M
v0 = 0000000000000000 t0 = 0000000000000008 t1 = 0000000000000000^M
t2 = 0000000000000000 t3 = 000000000000000e t4 = 0000000000000038^M
t5 = 000000000000ffff t6 = fffffc00002f298a t7 = fffffc0002c78000^M
s0 = fffffc00010b3ca0 s1 = 0000000000000000 s2 = fffffc00011267e0^M
s3 = 0000000000000000 s4 = fffffc00025f2008 s5 = fffffc00002f2940^M
s6 = fffffc00025f2040^M
a0 = fffffc00025f2008 a1 = fffffc00002f2940 a2 = fffffc0002ca000c^M
a3 = fffffc00000250d0 a4 = 0000000effff0008 a5 = 0000000000000000^M
t8 = fffffc00010b3c80 t9 = fffffc0002ca04cc t10= 0000000000000000^M
t11= 00000000000004c0 pv = fffffc0000b8bc40 at = 0000000000000000^M
gp = fffffc00010f9fb8 sp = 00000000df74db09^M
Disabling lock debugging due to kernel taint^M
Trace:^M
[<fffffc0000b8c5b8>] napi_gro_receive+0x68/0x150^M
[<fffffc00009b409c>] receive_buf+0x50c/0x1b80^M
[<fffffc00009b58b8>] virtnet_poll+0x1a8/0x5b0^M
[<fffffc00009b58ec>] virtnet_poll+0x1dc/0x5b0^M
[<fffffc0000b8d17c>] __napi_poll+0x4c/0x270^M
[<fffffc0000b8d670>] net_rx_action+0x130/0x2c0^M
[<fffffc0000bd6cb0>] sch_direct_xmit+0x170/0x360^M
[<fffffc0000bd7000>] __qdisc_run+0x160/0x6c0^M
[<fffffc0000337b64>] do_softirq+0xa4/0xd0^M
[<fffffc0000337ca4>] __local_bh_enable_ip+0x114/0x120^M
[<fffffc0000b89554>] __dev_queue_xmit+0x484/0xa60^M
[<fffffc0000cd072c>] packet_sendmsg+0xe7c/0x1ba0^M
[<fffffc0000b53338>] __sys_sendto+0xf8/0x170^M
[<fffffc0000cfec18>] _raw_spin_unlock+0x18/0x30^M
[<fffffc0000a9bf7c>] ehci_irq+0x2cc/0x5c0^M
[<fffffc0000a71334>] usb_hcd_irq+0x34/0x50^M
[<fffffc0000b521bc>] move_addr_to_kernel+0x3c/0x60^M
[<fffffc0000b532e4>] __sys_sendto+0xa4/0x170^M
[<fffffc0000b533d4>] sys_sendto+0x24/0x40^M
[<fffffc0000cfea38>] _raw_spin_lock+0x18/0x30^M
[<fffffc0000cfec18>] _raw_spin_unlock+0x18/0x30^M
[<fffffc0000325298>] clipper_enable_irq+0x98/0x100^M
[<fffffc0000cfec18>] _raw_spin_unlock+0x18/0x30^M
[<fffffc0000311514>] entSys+0xa4/0xc0^M
^M
Code:^M
00000000 ^M
00063301 ^M
00000897 ^M
00001111 ^M
00001672 ^M
^M
Kernel panic - not syncing: Aiee, killing interrupt handler!^M

Guenter

[Eric Dumazet]

OK, it would be nice if you could get line number from this stack trace.

(scripts/decode_stacktrace.sh is your friend)

[Guenter Roeck]

Here you are:

napi_gro_receive (net/core/dev.c:6196)
receive_buf (drivers/net/virtio_net.c:1150)
virtnet_poll (drivers/net/virtio_net.c:1414 drivers/net/virtio_net.c:1519)
clipper_srm_device_interrupt (arch/alpha/kernel/sys_dp264.c:256)
virtnet_poll (drivers/net/virtio_net.c:1413 drivers/net/virtio_net.c:1519)
__napi_poll (net/core/dev.c:6962)
net_rx_action (net/core/dev.c:7029 net/core/dev.c:7116)
__qdisc_run (net/sched/sch_generic.c:376 net/sched/sch_generic.c:384)
do_softirq (./include/asm-generic/softirq_stack.h:10 kernel/softirq.c:460 kernel/softirq.c:447)
__local_bh_enable_ip (kernel/softirq.c:384)
__dev_queue_xmit (./include/linux/bottom_half.h:32 ./include/linux/rcupdate.h:746 net/core/dev.c:4272)
packet_sendmsg (net/packet/af_packet.c:3009 net/packet/af_packet.c:3034)
__sys_sendto (net/socket.c:654 net/socket.c:674 net/socket.c:1977)
__d_alloc (fs/dcache.c:1744)
packet_create (net/packet/af_packet.c:1192 net/packet/af_packet.c:3296)
move_addr_to_kernel (./include/linux/uaccess.h:192 net/socket.c:198 net/socket.c:192)
__sys_sendto (net/socket.c:1968)
sys_sendto (net/socket.c:1989 net/socket.c:1985)
sys_bind (net/socket.c:1648 net/socket.c:1646)
entSys (arch/alpha/kernel/entry.S:477)

Guenter

[Eric Dumazet]

OK, I guess we are back to unaligned access, right ?
I guess sh arch should have failed as well ?

diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
index 8cd76037c72481200ea3e8429e9fdfec005dad85..0579914d3dd84c24982c1ff85314cc7b8d0f8d2d

100644
--- a/drivers/net/virtio_net.c
+++ b/drivers/net/virtio_net.c

@@ -415,7 +415,8 @@ static struct sk_buff *page_to_skb(struct virtnet_info *vi,

shinfo_size = SKB_DATA_ALIGN(sizeof(struct skb_shared_info));

- if (len > GOOD_COPY_LEN && tailroom >= shinfo_size) {
+ if (len > GOOD_COPY_LEN && tailroom >= shinfo_size &&
+ (!NET_IP_ALIGN || ((unsigned long)p & 3) == 2)) {
skb = build_skb(p, truesize);
if (unlikely(!skb))
return NULL;

[Guenter Roeck]

sh does indeed fail, with the same symptoms as before, but so far I was not
able to track it down to a specific commit. The alpha failure is different,
though. It is a NULL pointer access.

Anyway, testing ...

The patch below does indeed fix the problem I am seeing on sh.

... and it does fix the alpha problem as well. Neat, though I don't really understand
what a NULL pointer access and an unaligned access have to do with each other.

Great catch, thanks!

Guenter

[Eric Dumazet]

Not sure alpha linux kernel deals nicely with unaligned trap ?

[Eric Dumazet]

On 4/20/21 7:51 PM, Guenter Roeck wrote:

>
> sh does indeed fail, with the same symptoms as before, but so far I was not
> able to track it down to a specific commit. The alpha failure is different,
> though. It is a NULL pointer access.
>
> Anyway, testing ...
>
> The patch below does indeed fix the problem I am seeing on sh.
>
> ... and it does fix the alpha problem as well. Neat, though I don't really understand
> what a NULL pointer access and an unaligned access have to do with each other.
>
> Great catch, thanks!
>
> Guenter
>

Note that build_skb(), without an additional skb_reserve(skb, NET_IP_ALIGN)
can not possibly work on arches that care about alignments.

That is because we can not both align skb->data and skb_shinfo(skb)

So unless we change build_skb() to make sure to align skb_shinfo(),
a fix could be simply :

diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
index 8cd76037c72481200ea3e8429e9fdfec005dad85..9cbe9c1737649450e451e3c65f59f794d1bf34b0 100644
--- a/drivers/net/virtio_net.c
+++ b/drivers/net/virtio_net.c
@@ -415,7 +415,7 @@ static struct sk_buff *page_to_skb(struct virtnet_info *vi,

shinfo_size = SKB_DATA_ALIGN(sizeof(struct skb_shared_info));

- if (len > GOOD_COPY_LEN && tailroom >= shinfo_size) {

+ if (!_NET_IP_ALIGN && len > GOOD_COPY_LEN && tailroom >= shinfo_size) {

[Jason Wang]

Acked-by: Jason Wang <jaso...@redhat.com>