On 7/21/2021 5:24 PM, Oliver Hartkopp wrote: > Answering myself ... > > On 21.07.21 08:35, Oliver Hartkopp wrote: >> >> >> On 21.07.21 06:53, Greg KH wrote: >>> On Wed, Jul 21, 2021 at 09:09:37AM +0800, Ziyang Xuan wrote: >>>> We get a bug during ltp can_filter test as following. >>>> >>>> =========================================== >>>> [60919.264984] BUG: unable to handle kernel NULL pointer dereference at 0000000000000010 >>>> [60919.265223] PGD 8000003dda726067 P4D 8000003dda726067 PUD 3dda727067 PMD 0 >>>> [60919.265443] Oops: 0000 [#1] SMP PTI >>>> [60919.265550] CPU: 30 PID: 3638365 Comm: can_filter Kdump: loaded Tainted: G W 4.19.90+ #1 >> >> This kernel version 4.19.90 is definitely outdated. >> >> Can you please check your issue with the latest uptream kernel as this problem should have been fixed with this patch: >> >> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=8d0caedb759683041d9db82069937525999ada53 >> ("can: bcm/raw/isotp: use per module netdevice notifier") >> >> Thanks! > > I think my hint had a wrong assumption. The suggestion to add some locking seems correct. > >>>> [60919.266068] RIP: 0010:selinux_socket_sock_rcv_skb+0x3e/0x200 >>>> [60919.293289] RSP: 0018:ffff8d53bfc03cf8 EFLAGS: 00010246 >>>> [60919.307140] RAX: 0000000000000000 RBX: 000000000000001d RCX: 0000000000000007 >>>> [60919.320756] RDX: 0000000000000001 RSI: ffff8d5104a8ed00 RDI: ffff8d53bfc03d30 >>>> [60919.334319] RBP: ffff8d9338056800 R08: ffff8d53bfc29d80 R09: 0000000000000001 >>>> [60919.347969] R10: ffff8d53bfc03ec0 R11: ffffb8526ef47c98 R12: ffff8d53bfc03d30 >>>> [60919.350320] perf: interrupt took too long (3063 > 2500), lowering kernel.perf_event_max_sample_rate to 65000 >>>> [60919.361148] R13: 0000000000000001 R14: ffff8d53bcf90000 R15: 0000000000000000 >>>> [60919.361151] FS: 00007fb78b6b3600(0000) GS:ffff8d53bfc00000(0000) knlGS:0000000000000000 >>>> [60919.400812] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 >>>> [60919.413730] CR2: 0000000000000010 CR3: 0000003e3f784006 CR4: 00000000007606e0 >>>> [60919.426479] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 >>>> [60919.439339] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 >>>> [60919.451608] PKRU: 55555554 >>>> [60919.463622] Call Trace: >>>> [60919.475617] <IRQ> >>>> [60919.487122] ? update_load_avg+0x89/0x5d0 >>>> [60919.498478] ? update_load_avg+0x89/0x5d0 >>>> [60919.509822] ? account_entity_enqueue+0xc5/0xf0 >>>> [60919.520709] security_sock_rcv_skb+0x2a/0x40 >>>> [60919.531413] sk_filter_trim_cap+0x47/0x1b0 >>>> [60919.542178] ? kmem_cache_alloc+0x38/0x1b0 >>>> [60919.552444] sock_queue_rcv_skb+0x17/0x30 >>>> [60919.562477] raw_rcv+0x110/0x190 [can_raw] >>>> [60919.572539] can_rcv_filter+0xbc/0x1b0 [can] >>>> [60919.582173] can_receive+0x6b/0xb0 [can] >>>> [60919.591595] can_rcv+0x31/0x70 [can] >>>> [60919.600783] __netif_receive_skb_one_core+0x5a/0x80 >>>> [60919.609864] process_backlog+0x9b/0x150 >>>> [60919.618691] net_rx_action+0x156/0x400 >>>> [60919.627310] ? sched_clock_cpu+0xc/0xa0 >>>> [60919.635714] __do_softirq+0xe8/0x2e9 >>>> [60919.644161] do_softirq_own_stack+0x2a/0x40 >>>> [60919.652154] </IRQ> >>>> [60919.659899] do_softirq.part.17+0x4f/0x60 >>>> [60919.667475] __local_bh_enable_ip+0x60/0x70 >>>> [60919.675089] __dev_queue_xmit+0x539/0x920 >>>> [60919.682267] ? finish_wait+0x80/0x80 >>>> [60919.689218] ? finish_wait+0x80/0x80 >>>> [60919.695886] ? sock_alloc_send_pskb+0x211/0x230 >>>> [60919.702395] ? can_send+0xe5/0x1f0 [can] >>>> [60919.708882] can_send+0xe5/0x1f0 [can] >>>> [60919.715037] raw_sendmsg+0x16d/0x268 [can_raw] >>>> >>>> It's because raw_setsockopt() concurrently with >>>> unregister_netdevice_many(). Concurrent scenario as following. >>>> >>>> cpu0 cpu1 >>>> raw_bind >>>> raw_setsockopt unregister_netdevice_many >>>> unlist_netdevice >>>> dev_get_by_index raw_notifier >>>> raw_enable_filters ...... >>>> can_rx_register >>>> can_rcv_list_find(..., net->can.rx_alldev_list) >>>> >>>> ...... >>>> >>>> sock_close >>>> raw_release(sock_a) >>>> >>>> ...... >>>> >>>> can_receive >>>> can_rcv_filter(net->can.rx_alldev_list, ...) >>>> raw_rcv(skb, sock_a) >>>> BUG >>>> >>>> After unlist_netdevice(), dev_get_by_index() return NULL in >>>> raw_setsockopt(). Function raw_enable_filters() will add sock >>>> and can_filter to net->can.rx_alldev_list. > > Btw. this should not happen too! > > dev_get_by_index() is executed depending on ro->ifindex which means there should be a real network interface. When dev_get_by_index() returns NULL this can considered to be wrong. > > Adding a new filter to net->can.rx_alldev_list as a consequence is wrong too. > >>>> Then the sock is closed. >>>> Followed by, we sock_sendmsg() to a new vcan device use the same >>>> can_filter. Protocol stack match the old receiver whose sock has >>>> been released on net->can.rx_alldev_list in can_rcv_filter(). >>>> Function raw_rcv() uses the freed sock. UAF BUG is triggered. >>>> >>>> We can find that the key issue is that net_device has not been >>>> protected in raw_setsockopt(). Use rtnl_lock to protect net_device >>>> in raw_setsockopt(). >>>> >>>> Fixes: c18ce101f2e4 ("[CAN]: Add raw protocol") >>>> Signed-off-by: Ziyang Xuan <william.xuanziyang@xxxxxxxxxx> > > Can you please resend the below patch as suggested by Greg KH and add my > > Signed-off-by: Oliver Hartkopp <socketcan@xxxxxxxxxxxx> > > as it also adds the dev_get_by_index() return check. > > diff --git a/net/can/raw.c b/net/can/raw.c > index ed4fcb7ab0c3..d3cbc32036c7 100644 > --- a/net/can/raw.c > +++ b/net/can/raw.c > @@ -544,14 +544,18 @@ static int raw_setsockopt(struct socket *sock, int level, int optname, > } else if (count == 1) { > if (copy_from_sockptr(&sfilter, optval, sizeof(sfilter))) > return -EFAULT; > } > > + rtnl_lock(); > lock_sock(sk); > > - if (ro->bound && ro->ifindex) > + if (ro->bound && ro->ifindex) { > dev = dev_get_by_index(sock_net(sk), ro->ifindex); > + if (!dev) > + goto out_fil; > + } At first, I also use this modification. After discussion with my partner, we found that it is impossible scenario if we use rtnl_lock to protect net_device object. We can see two sequences: 1. raw_setsockopt first get rtnl_lock, unregister_netdevice_many later. It can be simplified to add the filter in raw_setsockopt, then remove the filter in raw_notify. 2. unregister_netdevice_many first get rtnl_lock, raw_setsockopt later. raw_notify will set ro->ifindex, ro->bound and ro->count to zero firstly. The filter will not be added to any filter_list in raw_notify. So I selected the current modification. Do you think so? My first modification as following: diff --git a/net/can/raw.c b/net/can/raw.c index ed4fcb7ab0c3..a0ce4908317f 100644 --- a/net/can/raw.c +++ b/net/can/raw.c @@ -546,10 +546,16 @@ static int raw_setsockopt(struct socket *sock, int level, int optname, return -EFAULT; } + rtnl_lock(); lock_sock(sk); - if (ro->bound && ro->ifindex) + if (ro->bound && ro->ifindex) { dev = dev_get_by_index(sock_net(sk), ro->ifindex); + if (!dev) { + err = -ENODEV; + goto out_fil; + } + } if (ro->bound) { /* (try to) register the new filters */ @@ -559,11 +565,8 @@ static int raw_setsockopt(struct socket *sock, int level, int optname, else err = raw_enable_filters(sock_net(sk), dev, sk, filter, count); - if (err) { - if (count > 1) - kfree(filter); + if (err) goto out_fil; - } /* remove old filter registrations */ raw_disable_filters(sock_net(sk), dev, sk, ro->filter, @@ -584,10 +587,14 @@ static int raw_setsockopt(struct socket *sock, int level, int optname, ro->count = count; out_fil: + if (err && count > 1) + kfree(filter); + if (dev) dev_put(dev); release_sock(sk); + rtnl_unlock(); break; @@ -600,10 +607,16 @@ static int raw_setsockopt(struct socket *sock, int level, int optname, err_mask &= CAN_ERR_MASK; + rtnl_lock(); lock_sock(sk); - if (ro->bound && ro->ifindex) + if (ro->bound && ro->ifindex) { dev = dev_get_by_index(sock_net(sk), ro->ifindex); + if (!dev) { + err = -ENODEV; + goto out_err; + } + } /* remove current error mask */ if (ro->bound) { @@ -627,6 +640,7 @@ static int raw_setsockopt(struct socket *sock, int level, int optname, dev_put(dev); release_sock(sk); + rtnl_unlock(); break; > > if (ro->bound) { > /* (try to) register the new filters */ > if (count == 1) > err = raw_enable_filters(sock_net(sk), dev, sk, > @@ -586,10 +590,11 @@ static int raw_setsockopt(struct socket *sock, int level, int optname, > out_fil: > if (dev) > dev_put(dev); > > release_sock(sk); > + rtnl_unlock(); > > break; > > case CAN_RAW_ERR_FILTER: > if (optlen != sizeof(err_mask)) > @@ -598,14 +603,18 @@ static int raw_setsockopt(struct socket *sock, int level, int optname, > if (copy_from_sockptr(&err_mask, optval, optlen)) > return -EFAULT; > > err_mask &= CAN_ERR_MASK; > > + rtnl_lock(); > lock_sock(sk); > > - if (ro->bound && ro->ifindex) > + if (ro->bound && ro->ifindex) { > dev = dev_get_by_index(sock_net(sk), ro->ifindex); > + if (!dev) > + goto out_err; > + } > > /* remove current error mask */ > if (ro->bound) { > /* (try to) register the new err_mask */ > err = raw_enable_errfilter(sock_net(sk), dev, sk, > @@ -625,10 +634,11 @@ static int raw_setsockopt(struct socket *sock, int level, int optname, > out_err: > if (dev) > dev_put(dev); > > release_sock(sk); > + rtnl_unlock(); > > break; > > case CAN_RAW_LOOPBACK: > if (optlen != sizeof(ro->loopback)) > > > > > Thanks for the finding! > > Best regards, > Oliver > > (..) >>> >>> >>> <formletter> >>> >>> This is not the correct way to submit patches for inclusion in the >>> stable kernel tree. Please read: >>> https://www.kernel.org/doc/html/latest/process/stable-kernel-rules.html >>> for how to do this properly. >>> >>> </formletter> >>> > .
