Running debug kernel on a node with infiniband card, got a KASan complain:
==================================================================
BUG: KASAN: slab-out-of-bounds in i40iw_copy_ip_ntohl+0x1c0/0x220
Read of size 4 at addr ffff88852d477380 by task swapper/6/0
CPU: 6 PID: 0 Comm: swapper/6 Not tainted 4.20.0-rc3-00087-gc8ce94b8fe53-dirty #15
Hardware name: DEPO Computers Super Server/X10DRL-i, BIOS 2.0b 05/05/2017
Call Trace:
<IRQ>
dump_stack+0x92/0xeb
print_address_description+0x6a/0x280
kasan_report+0x260/0x380
i40iw_copy_ip_ntohl+0x1c0/0x220
i40iw_net_event+0x150/0x200
notifier_call_chain+0x90/0x160
atomic_notifier_call_chain+0x6c/0x100
neigh_update+0x82f/0x15c0
neigh_event_ns+0x4c/0xe0
arp_process+0x1733/0x1f60
__netif_receive_skb_one_core+0xe6/0x150
netif_receive_skb_internal+0xe5/0x4c0
napi_gro_receive+0x2d1/0x3b0
i40e_clean_rx_irq+0x9a5/0x2eb0
i40e_napi_poll+0x11fd/0x2410
net_rx_action+0x62f/0xbf0
__do_softirq+0x256/0x9de
irq_exit+0x29b/0x2d0
do_IRQ+0x87/0x1a0
common_interrupt+0xf/0xf
Allocated by task 0:
kasan_kmalloc+0xa0/0xd0
__kmalloc+0x177/0x390
__neigh_create+0x1e3/0x1820
neigh_event_ns+0x6b/0xe0
arp_process+0x1733/0x1f60
__netif_receive_skb_one_core+0xe6/0x150
netif_receive_skb_internal+0xe5/0x4c0
napi_gro_receive+0x2d1/0x3b0
i40e_clean_rx_irq+0x9a5/0x2eb0
i40e_napi_poll+0x11fd/0x2410
net_rx_action+0x62f/0xbf0
__do_softirq+0x256/0x9de
Freed by task 0:
(stack is not available)
The buggy address belongs to the object at ffff88852d477080
to the cache kmalloc-1k of size 1024
The buggy address is located 768 bytes inside of 1024-byte region [ffff88852d477080, ffff88852d477480)
The buggy address belongs to the page:
page:ffffea0014b51c00 count:1 mapcount:0 mapping:ffff888107c0ea00 index:0x0 compound_mapcount: 0
flags: 0x17ffffc0010200(slab|head)
raw: 0017ffffc0010200 dead000000000100 dead000000000200 ffff888107c0ea00
raw: 0000000000000000 00000000801c001c 00000001ffffffff 0000000000000000
page dumped because: kasan: bad access detected
Memory state around the buggy address:
ffff88852d477280: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
ffff88852d477300: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
>ffff88852d477380: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
^
ffff88852d477400: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
ffff88852d477480: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
==================================================================
The complain is valid:
i40iw_net_event() reads unconditionally 16 bytes from neigh->primary_key
while the memory allocated for "neighbour" struct is evaluated in neigh_alloc() as
tbl->entry_size + dev->neigh_priv_len
where "dev" is a net_device.
But the driver does not setup dev->neigh_priv_len and we read beyond the neigh
entry allocated memory, so the patch in the next mail fixes this.
More debug details:
crash> list net_device.dev_list -H 0xffffffffa908ec88 -s net_device.name -s net_device.neigh_priv_len
ffff88065a92a200
name = "lo\000\000\000\000\000\000\000\000\000\000\000\000\000"
neigh_priv_len = 0
ffff880642340000
name = "eno1\000\000\071:00.0\000\000\000"
neigh_priv_len = 0
ffff88064aa6a200
name = "enp6s0f0\000\000\000\000\000\000\000"
neigh_priv_len = 0
ffff880641180000
name = "eno2\000\000a:00.0\000\000\000"
neigh_priv_len = 0
ffff88063e8fd500
name = "enp6s0f1\000\000\000\000\000\000\000"
neigh_priv_len = 0
ffff880031400000
name = "ens11f0\000\000\000\000\000\000\000\000"
neigh_priv_len = 0
ffff88063c800000
name = "ens11f1\000\000\000\000\000\000\000\000"
neigh_priv_len = 0
ffff8808ff4ea100
name = "bond0\000\000\000\000\000\000\000\000\000\000"
neigh_priv_len = 0
ffff88101e334400
name = "ib0\000\000\000\000\000\000\000\000\000\000\000\000"
neigh_priv_len = 200
=========================================
crash> list i40iw_handler.list -H i40iw_handlers
ffff88004bbc0000 ldev.netdev == 0xffff88063e8fd500
struct net_device {
name = "enp6s0f1\000\000\000\000\000\000\000",
ffff881049120000 ldev.netdev == 0xffff88064aa6a200
struct net_device {
name = "enp6s0f0\000\000\000\000\000\000\000",
=========================================
net_device allocation stack:
alloc_netdev_mqs
alloc_etherdev_mq
i40e_config_netdev
i40e_vsi_setup
i40e_setup_pf_switch
i40e_probe
=========================================
After the patch:
crash> list net_device.dev_list -H 0xffffffff92a19b48 -s net_device.name -s net_device.neigh_priv_len
ffff88065a2dc400
name = "lo\000\000\000\000\000\000\000\000\000\000\000\000\000"
neigh_priv_len = 0
ffff8808fb6dc200
name = "bond0\000\000\000\000\000\000\000\000\000\000"
neigh_priv_len = 0
ffff880652600000
name = "ens11f0\000\000\000\000\000\000\000\000"
neigh_priv_len = 0
ffff880651a00000
name = "ens11f1\000\000\000\000\000\000\000\000"
neigh_priv_len = 0
ffff880651454000
name = "eno1\000\000\071:00.0\000\000\000"
neigh_priv_len = 0
ffff880651550000
name = "eno2\000\000a:00.0\000\000\000"
neigh_priv_len = 0
ffff8806515cc400
name = "enp6s0f0\000\000\000\000\000\000\000"
neigh_priv_len = 16
ffff880650932200
name = "enp6s0f1\000\000\000\000\000\000\000"
neigh_priv_len = 16
ffff880642903300
name = "ib0\000\000\000\000\000\000\000\000\000\000\000\000"
neigh_priv_len = 200
=========================================
Konstantin Khorenko (1):
drivers/net/i40e: define proper net_device::neigh_priv_len
drivers/net/ethernet/intel/i40e/i40e_main.c | 3 +++
1 file changed, 3 insertions(+)
--
2.15.1
Out of bound read reported by KASan.
i40iw_net_event() reads unconditionally 16 bytes from
neigh->primary_key while the memory allocated for
"neighbour" struct is evaluated in neigh_alloc() as
tbl->entry_size + dev->neigh_priv_len
where "dev" is a net_device.
But the driver does not setup dev->neigh_priv_len and
we read beyond the neigh entry allocated memory,
so the patch in the next mail fixes this.
Signed-off-by: Konstantin Khorenko <[email protected]>
---
drivers/net/ethernet/intel/i40e/i40e_main.c | 3 +++
1 file changed, 3 insertions(+)
diff --git a/drivers/net/ethernet/intel/i40e/i40e_main.c b/drivers/net/ethernet/intel/i40e/i40e_main.c
index 21c2688d6308..539b8f3738e6 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_main.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_main.c
@@ -12334,6 +12334,9 @@ static int i40e_config_netdev(struct i40e_vsi *vsi)
ether_addr_copy(netdev->dev_addr, mac_addr);
ether_addr_copy(netdev->perm_addr, mac_addr);
+ /* i40iw_net_event() reads 16 bytes from neigh->primary_key */
+ netdev->neigh_priv_len = sizeof(u32) * 4;
+
netdev->priv_flags |= IFF_UNICAST_FLT;
netdev->priv_flags |= IFF_SUPP_NOFCS;
/* Setup netdev TC information */
--
2.15.1
> -----Original Message-----
> From: Intel-wired-lan [mailto:[email protected]] On
> Behalf Of Konstantin Khorenko
> Sent: Friday, November 23, 2018 8:10 AM
> To: Kirsher, Jeffrey T <[email protected]>
> Cc: [email protected]; [email protected]; linux-
> [email protected]; Konstantin Khorenko <[email protected]>;
> David S . Miller <[email protected]>
> Subject: [Intel-wired-lan] [PATCH 1/1] drivers/net/i40e: define proper
> net_device::neigh_priv_len
>
> Out of bound read reported by KASan.
>
> i40iw_net_event() reads unconditionally 16 bytes from
> neigh->primary_key while the memory allocated for
> "neighbour" struct is evaluated in neigh_alloc() as
>
> tbl->entry_size + dev->neigh_priv_len
>
> where "dev" is a net_device.
>
> But the driver does not setup dev->neigh_priv_len and we read beyond the
> neigh entry allocated memory, so the patch in the next mail fixes this.
>
> Signed-off-by: Konstantin Khorenko <[email protected]>
> ---
> drivers/net/ethernet/intel/i40e/i40e_main.c | 3 +++
> 1 file changed, 3 insertions(+)
Tested-by: Andrew Bowers <[email protected]>