linux-kernel - Kernel config for fuzz testing.

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Date:   Tue, 8 Oct 2019 18:57:53 +0900
From:   Tetsuo Handa <penguin-kernel@...ove.sakura.ne.jp>
To:     Dmitry Vyukov <dvyukov@...gle.com>,
        Linus Torvalds <torvalds@...ux-foundation.org>
Cc:     Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        Arnd Bergmann <arnd@...db.de>,
        LKML <linux-kernel@...r.kernel.org>,
        syzbot <syzbot+8ab2d0f39fb79fe6ca40@...kaller.appspotmail.com>,
        Eric Biggers <ebiggers@...nel.org>
Subject: Kernel config for fuzz testing.

On 2019/08/23 17:17, Tetsuo Handa wrote:
> On 2019/08/23 8:59, Dmitry Vyukov wrote:
>>> Can't we introduce a kernel config which selectively blocks specific actions?
>>> If we don't need to worry about bypassing blacklist checks, we will be able to
>>> enable syz_execute_func() again.
>>
>>
>> We can consider this, but we need some set of good use cases first.
>> For /dev/{mem,kmem} we disable them with config, right?
> 
> /dev/{mem,kmem} can be disabled by kernel config options. But
> 
>>                                                         That looks
>> like the right thing to do because we don't want fuzzer to do anything
>> with these files anyway.
> 
> I don't think so. To examine as corner as possible (e.g. lock dependency),
> I consider that even doing
> 
> ----------
> +#ifdef CONFIG_KERNEL_BUILT_FOR_FUZZ_TESTING
> +static char dummybuf[PAGE_SIZE];
> +#endif
> ----------
> 
> ----------
>                         ptr = xlate_dev_mem_ptr(p);
>                         if (!ptr) {
>                                 if (written)
>                                         break;
>                                 return -EFAULT;
>                         }
> +#ifndef CONFIG_KERNEL_BUILT_FOR_FUZZ_TESTING
>                         copied = copy_from_user(ptr, buf, sz);
> +#else
> +                       copied = copy_from_user(dummybuf, buf, min(sizeof(dummybuf), sz));
> +#endif
>                         unxlate_dev_mem_ptr(p, ptr);
> ----------
> 
> makes sense, for copy_from_user() might find new lock dependency
> which would otherwise be unnoticed.
> 
>>                          So this won't be a good use case for
>> CONFIG_KERNEL_BUILT_FOR_FUZZ_TESTING.
>> Fuzzer can also reliably filter out based on syscall numbers of
>> top-level argument values. The potential problem is with (1)
>> pointers/indirect memory and (2) where blacklisting some top-level
>> argument values would backlist too much (e.g. prohibiting 3rd ioctl
>> argument 0 entirely).
> 
> I consider that functions that freezes processes/filesystems, 
> reboots/shutdowns a system, changes console loglevels can be blocked
> as well. Trying to examine up to last-second conditional branches will
> catch more bugs (e.g. bugs in error recovery paths).
> 

A USB fuzz test is triggering SysRq-t until RCU stall happens. ;-)
Can we introduce a kernel config (for linux.git) which selectively
blocks specific actions?

https://syzkaller.appspot.com/text?tag=CrashLog&x=12eb1e67600000

[  563.439044][    C1] ksoftirqd/1     R  running task    28264    16      2 0x80004008
[  563.447037][    C1] Call Trace:
[  563.450321][    C1]  sched_show_task.cold+0x2e0/0x359
[  563.455502][    C1]  show_state_filter+0x164/0x209
[  563.460421][    C1]  ? fn_caps_on+0x90/0x90
[  563.464730][    C1]  k_spec+0xdc/0x120
[  563.468605][    C1]  kbd_event+0x927/0x3790
[  563.472914][    C1]  ? k_pad+0x720/0x720
[  563.476964][    C1]  ? mark_held_locks+0xe0/0xe0
[  563.481791][    C1]  ? sysrq_filter+0xdf/0xeb0
[  563.486359][    C1]  ? k_pad+0x720/0x720
[  563.490419][    C1]  input_to_handler+0x3b6/0x4c0
[  563.495247][    C1]  input_pass_values.part.0+0x2e3/0x720
[  563.500770][    C1]  input_repeat_key+0x1ee/0x2c0
[  563.505622][    C1]  ? input_dev_suspend+0x80/0x80
[  563.510535][    C1]  ? rcu_read_lock_bh_held+0xb0/0xb0
[  563.515805][    C1]  call_timer_fn+0x179/0x650
[  563.520385][    C1]  ? input_dev_suspend+0x80/0x80
[  563.525389][    C1]  ? msleep_interruptible+0x130/0x130
[  563.531528][    C1]  ? rcu_read_lock_sched_held+0x9c/0xd0
[  563.537058][    C1]  ? rcu_read_lock_bh_held+0xb0/0xb0
[  563.542334][    C1]  ? _raw_spin_unlock_irq+0x24/0x30
[  563.547519][    C1]  ? input_dev_suspend+0x80/0x80
[  563.552443][    C1]  run_timer_softirq+0x5e3/0x1490
[  563.557454][    C1]  ? add_timer+0x7a0/0x7a0
[  563.561853][    C1]  ? rcu_read_lock_sched_held+0x9c/0xd0
[  563.567377][    C1]  ? rcu_read_lock_bh_held+0xb0/0xb0
[  563.572644][    C1]  __do_softirq+0x221/0x912
[  563.577147][    C1]  ? takeover_tasklets+0x720/0x720
[  563.582246][    C1]  run_ksoftirqd+0x1f/0x40
[  563.586638][    C1]  smpboot_thread_fn+0x3e8/0x850
[  563.591640][    C1]  ? smpboot_unregister_percpu_thread+0x190/0x190
[  563.598031][    C1]  ? __kthread_parkme+0x10a/0x1c0
[  563.603031][    C1]  ? smpboot_unregister_percpu_thread+0x190/0x190
[  563.609440][    C1]  kthread+0x318/0x420
[  563.613746][    C1]  ? kthread_create_on_node+0xf0/0xf0
[  563.619092][    C1]  ret_from_fork+0x24/0x30