Re: [syzbot] BUG: unable to handle kernel NULL pointer dereference in htb_select_queue

[Maxim Mikityanskiy <maximmi@nvidia.com>]

On 2021-03-10 19:03, Eric Dumazet wrote:
> 
> 
> On 3/10/21 3:54 PM, Maxim Mikityanskiy wrote:
>> On 2021-03-09 17:20, Eric Dumazet wrote:
>>>
>>>
>>> On 3/9/21 4:13 PM, syzbot wrote:
>>>> Hello,
>>>>
>>>> syzbot found the following issue on:
>>>>
>>>> HEAD commit:    38b5133a octeontx2-pf: Fix otx2_get_fecparam()
>>>> git tree:       net-next
>>>> console output: https://syzkaller.appspot.com/x/log.txt?x=166288a8d00000
>>>> kernel config:  https://syzkaller.appspot.com/x/.config?x=dbc1ca9e55dc1f9f
>>>> dashboard link: https://syzkaller.appspot.com/bug?extid=b53a709f04722ca12a3c
>>>> syz repro:      https://syzkaller.appspot.com/x/repro.syz?x=119454ccd00000
>>>>
>>>> The issue was bisected to:
>>>>
>>>> commit d03b195b5aa015f6c11988b86a3625f8d5dbac52
>>>> Author: Maxim Mikityanskiy <maximmi@mellanox.com>
>>>> Date:   Tue Jan 19 12:08:13 2021 +0000
>>>>
>>>>       sch_htb: Hierarchical QoS hardware offload
>>>>
>>>> bisection log:  https://syzkaller.appspot.com/x/bisect.txt?x=13ab12ecd00000
>>>> final oops:     https://syzkaller.appspot.com/x/report.txt?x=106b12ecd00000
>>>> console output: https://syzkaller.appspot.com/x/log.txt?x=17ab12ecd00000
>>>>
>>>> IMPORTANT: if you fix the issue, please add the following tag to the commit:
>>>> Reported-by: syzbot+b53a709f04722ca12a3c@syzkaller.appspotmail.com
>>>> Fixes: d03b195b5aa0 ("sch_htb: Hierarchical QoS hardware offload")
>>>>
>>>> BUG: kernel NULL pointer dereference, address: 0000000000000000
>>>> #PF: supervisor instruction fetch in kernel mode
>>>> #PF: error_code(0x0010) - not-present page
>>>> PGD 183fe067 P4D 183fe067 PUD 21aef067 PMD 0
>>>> Oops: 0010 [#1] PREEMPT SMP KASAN
>>>> CPU: 0 PID: 10125 Comm: syz-executor.0 Not tainted 5.11.0-rc7-syzkaller #0
>>>> Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
>>>> RIP: 0010:0x0
>>>> Code: Unable to access opcode bytes at RIP 0xffffffffffffffd6.
>>>> RSP: 0018:ffffc9000a9c74e8 EFLAGS: 00010246
>>>> RAX: dffffc0000000000 RBX: 1ffff92001538e9e RCX: 0000000000000000
>>>> RDX: ffffc9000a9c7520 RSI: 0000000000000012 RDI: ffff88802d158000
>>>> RBP: ffff88802d158000 R08: 00000000fffffff1 R09: 0000000000000400
>>>> R10: ffffffff871631c4 R11: 0000000000000000 R12: ffffffff89ea6b40
>>>> R13: dffffc0000000000 R14: ffff888012b79c00 R15: 00000000ffff0000
>>>> FS:  00007f73f9698700(0000) GS:ffff8880b9c00000(0000) knlGS:0000000000000000
>>>> CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
>>>> CR2: ffffffffffffffd6 CR3: 00000000173b5000 CR4: 00000000001506f0
>>>> DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
>>>> DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
>>>> Call Trace:
>>>>    htb_offload net/sched/sch_htb.c:1011 [inline]
>>>>    htb_select_queue+0x17f/0x2c0 net/sched/sch_htb.c:1349
>>>>    tc_modify_qdisc+0x44a/0x1a50 net/sched/sch_api.c:1657
>>>>    rtnetlink_rcv_msg+0x44e/0xad0 net/core/rtnetlink.c:5553
>>>>    netlink_rcv_skb+0x153/0x420 net/netlink/af_netlink.c:2502
>>>>    netlink_unicast_kernel net/netlink/af_netlink.c:1312 [inline]
>>>>    netlink_unicast+0x533/0x7d0 net/netlink/af_netlink.c:1338
>>>>    netlink_sendmsg+0x856/0xd90 net/netlink/af_netlink.c:1927
>>>>    sock_sendmsg_nosec net/socket.c:652 [inline]
>>>>    sock_sendmsg+0xcf/0x120 net/socket.c:672
>>>>    ____sys_sendmsg+0x6e8/0x810 net/socket.c:2348
>>>>    ___sys_sendmsg+0xf3/0x170 net/socket.c:2402
>>>>    __sys_sendmsg+0xe5/0x1b0 net/socket.c:2435
>>>>    do_syscall_64+0x2d/0x70 arch/x86/entry/common.c:46
>>>>    entry_SYSCALL_64_after_hwframe+0x44/0xa9
>>>> RIP: 0033:0x466019
>>>> Code: ff ff c3 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 40 00 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 c7 c1 bc ff ff ff f7 d8 64 89 01 48
>>>> RSP: 002b:00007f73f9698188 EFLAGS: 00000246 ORIG_RAX: 000000000000002e
>>>> RAX: ffffffffffffffda RBX: 000000000056bf60 RCX: 0000000000466019
>>>> RDX: 0000000000000000 RSI: 00000000200007c0 RDI: 0000000000000004
>>>> RBP: 00000000004bd067 R08: 0000000000000000 R09: 0000000000000000
>>>> R10: 0000000000000000 R11: 0000000000000246 R12: 000000000056bf60
>>>> R13: 00007fffefccc11f R14: 00007f73f9698300 R15: 0000000000022000
>>>> Modules linked in:
>>>> CR2: 0000000000000000
>>>> ---[ end trace e1544e8206616773 ]---
>>>> RIP: 0010:0x0
>>>> Code: Unable to access opcode bytes at RIP 0xffffffffffffffd6.
>>>> RSP: 0018:ffffc9000a9c74e8 EFLAGS: 00010246
>>>> RAX: dffffc0000000000 RBX: 1ffff92001538e9e RCX: 0000000000000000
>>>> RDX: ffffc9000a9c7520 RSI: 0000000000000012 RDI: ffff88802d158000
>>>> RBP: ffff88802d158000 R08: 00000000fffffff1 R09: 0000000000000400
>>>> R10: ffffffff871631c4 R11: 0000000000000000 R12: ffffffff89ea6b40
>>>> R13: dffffc0000000000 R14: ffff888012b79c00 R15: 00000000ffff0000
>>>> FS:  00007f73f9698700(0000) GS:ffff8880b9d00000(0000) knlGS:0000000000000000
>>>> CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
>>>> CR2: ffffffffffffffd6 CR3: 00000000173b5000 CR4: 00000000001506e0
>>>> DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
>>>> DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
>>>>
>>>>
>>>> ---
>>>> This report is generated by a bot. It may contain errors.
>>>> See https://goo.gl/tpsmEJ for more information about syzbot.
>>>> syzbot engineers can be reached at syzkaller@googlegroups.com.
>>>>
>>>> syzbot will keep track of this issue. See:
>>>> https://goo.gl/tpsmEJ#status for how to communicate with syzbot.
>>>> For information about bisection process see: https://goo.gl/tpsmEJ#bisection
>>>> syzbot can test patches for this issue, for details see:
>>>> https://goo.gl/tpsmEJ#testing-patches
>>>>
>>>
>>>
>>> Hmm... what about this :
>>>
>>> diff --git a/net/sched/sch_api.c b/net/sched/sch_api.c
>>> index f87d07736a1404edcfd17a792321758cd4bdd173..680afb5bfe2294a5531c7aaeed698b95ea3ab20c 100644
>>> --- a/net/sched/sch_api.c
>>> +++ b/net/sched/sch_api.c
>>> @@ -1651,15 +1651,16 @@ static int tc_modify_qdisc(struct sk_buff *skb, struct nlmsghdr *n,
>>>                           err = -ENOENT;
>>>                   }
>>>           } else {
>>> -               struct netdev_queue *dev_queue;
>>> +               struct netdev_queue *dev_queue = NULL;
>>>                     if (p && p->ops->cl_ops && p->ops->cl_ops->select_queue)
>>>                           dev_queue = p->ops->cl_ops->select_queue(p, tcm);
>>> -               else if (p)
>>> -                       dev_queue = p->dev_queue;
>>> -               else
>>> -                       dev_queue = netdev_get_tx_queue(dev, 0);
>>> -
>>> +               if (!dev_queue) {
>>> +                       if (p)
>>> +                               dev_queue = p->dev_queue;
>>> +                       else
>>> +                               dev_queue = netdev_get_tx_queue(dev, 0);
>>> +               }
>>>                   q = qdisc_create(dev, dev_queue, p,
>>>                                    tcm->tcm_parent, tcm->tcm_handle,
>>>                                    tca, &err, extack);
>>> diff --git a/net/sched/sch_htb.c b/net/sched/sch_htb.c
>>> index dff3adf5a9156c2412c64a10ad1b2ce9e1367433..cc6eccd688701ae00255f07e32fb4b0efbaf45ce 100644
>>> --- a/net/sched/sch_htb.c
>>> +++ b/net/sched/sch_htb.c
>>> @@ -1008,6 +1008,8 @@ static void htb_set_lockdep_class_child(struct Qdisc *q)
>>>      static int htb_offload(struct net_device *dev, struct tc_htb_qopt_offload *opt)
>>>    {
>>> +       if (!tc_can_offload(dev) || !dev->netdev_ops->ndo_setup_tc)
>>> +               return -EOPNOTSUPP;
>>
>> My fault, all calls to htb_offload must be protected by if (q->offload). Rather than checking tc_can_offload and ndo_setup_tc in htb_offload every time, I suggest to fix htb_select_queue:
>>
>> diff --git a/net/sched/sch_htb.c b/net/sched/sch_htb.c
>> index dff3adf5a915..b23203159996 100644
>> --- a/net/sched/sch_htb.c
>> +++ b/net/sched/sch_htb.c
>> @@ -1340,8 +1340,12 @@ htb_select_queue(struct Qdisc *sch, struct tcmsg *tcm)
>>   {
>>       struct net_device *dev = qdisc_dev(sch);
>>       struct tc_htb_qopt_offload offload_opt;
>> +    struct htb_sched *q = qdisc_priv(sch);
>>       int err;
>>
>> +    if (!q->offload)
>> +        return sch->dev_queue;
>> +
>>       offload_opt = (struct tc_htb_qopt_offload) {
>>           .command = TC_HTB_LEAF_QUERY_QUEUE,
>>           .classid = TC_H_MIN(tcm->tcm_parent),
>>
>> htb_init ensures that tc_can_offload and ndo_setup_tc are checked if q->offload is true. Also, we can avoid changing tc_modify_qdisc if htb_select_queue mimics its behavior in non-offload mode, as shown above.
>>
>> There is also a case where htb_select_queue returns NULL on errors, and that is handled in qdisc_create (the error message will be "No device queue given"), which I think is a sane behavior.
>>
>> What do you think of this fix? If it fits, I'll send it as a patch.
> 
> 
> I think that it is not enough, since you overwrite q->offload in htb_init()
> even if an error will be provided.
> 
> So a malicious user will find its way.

I doubt that, because if htb_init returns an error, the qdisc gets 
destroyed immediately (well, after a call to htb_destroy), and I believe 
all these operations are protected by RTNL, so a malicious user has no 
way to insert a call to another callback.

> You probably also need this :

However, I'll likely need something like this anyway, because HTB must 
not call ndo_setup_tc on destroy if it didn't call it on init. It may 
crash in a similar way if ndo_setup_tc is not implemented. Thanks for 
helping me spot that - if you don't mind, I'll base my second patch on 
your code below.

> 
> 
> diff --git a/net/sched/sch_htb.c b/net/sched/sch_htb.c
> index dff3adf5a9156c2412c64a10ad1b2ce9e1367433..d15ee7cf33b34221d09dfc81105dcb6c2b2fd489 100644
> --- a/net/sched/sch_htb.c
> +++ b/net/sched/sch_htb.c
> @@ -1020,6 +1020,7 @@ static int htb_init(struct Qdisc *sch, struct nlattr *opt,
>          struct nlattr *tb[TCA_HTB_MAX + 1];
>          struct tc_htb_glob *gopt;
>          unsigned int ntx;
> +       bool offload;
>          int err;
>   
>          qdisc_watchdog_init(&q->watchdog, sch);
> @@ -1044,9 +1045,9 @@ static int htb_init(struct Qdisc *sch, struct nlattr *opt,
>          if (gopt->version != HTB_VER >> 16)
>                  return -EINVAL;
>   
> -       q->offload = nla_get_flag(tb[TCA_HTB_OFFLOAD]);
> +       offload = nla_get_flag(tb[TCA_HTB_OFFLOAD]);
>   
> -       if (q->offload) {
> +       if (offload) {
>                  if (sch->parent != TC_H_ROOT)
>                          return -EOPNOTSUPP;
>   
> @@ -1060,6 +1061,7 @@ static int htb_init(struct Qdisc *sch, struct nlattr *opt,
>                  if (!q->direct_qdiscs)
>                          return -ENOMEM;
>          }
> +       q->offload = offload;
>   
>          err = qdisc_class_hash_init(&q->clhash);
>          if (err < 0)
>