Re: [PATCH v4 4/4] seccomp: add support for passing fds via USER_NOTIF

[Andy Lutomirski <luto@amacapital.net>]

> On Jun 22, 2018, at 9:23 AM, Jann Horn <jannh@google.com> wrote:
> 
>> On Fri, Jun 22, 2018 at 12:05 AM Tycho Andersen <tycho@tycho.ws> wrote:
>> 
>> The idea here is that the userspace handler should be able to pass an fd
>> back to the trapped task, for example so it can be returned from socket().
>> 
>> I've proposed one API here, but I'm open to other options. In particular,
>> this only lets you return an fd from a syscall, which may not be enough in
>> all cases. For example, if an fd is written to an output parameter instead
>> of returned, the current API can't handle this. Another case is that
>> netlink takes as input fds sometimes (IFLA_NET_NS_FD, e.g.). If netlink
>> ever decides to install an fd and output it, we wouldn't be able to handle
>> this either.
>> 
>> Still, the vast majority of interesting cases are covered by this API, so
>> perhaps it is Enough.
>> 
>> I've left it as a separate commit for two reasons:
>>  * It illustrates the way in which we would grow struct seccomp_notif and
>>    struct seccomp_notif_resp without using netlink
>>  * It shows just how little code is needed to accomplish this :)
>> 
> [...]
>> @@ -1669,10 +1706,20 @@ static ssize_t seccomp_notify_write(struct file *file, const char __user *buf,
>>                goto out;
>>        }
>> 
>> +       if (resp.return_fd) {
>> +               knotif->flags = resp.fd_flags;
>> +               knotif->file = fget(resp.fd);
>> +               if (!knotif->file) {
>> +                       ret = -EBADF;
>> +                       goto out;
>> +               }
>> +       }
>> +
> 
> I think this is a security bug. Imagine the following scenario:
> 
> - attacker creates processes A and B
> - process A installs a seccomp filter and sends the notification fd
> to process B
> - process A starts a syscall for which the filter returns
> SECCOMP_RET_USER_NOTIF
> - process B reads the notification from the notification fd
> - process B uses dup2() to copy the notification fd to file
> descriptor 1 (stdout)
> - process B executes a setuid root binary
> - the setuid root binary opens some privileged file descriptor
> (something like open("/etc/shadow", O_RDWR))
> - the setuid root binary tries to write some attacker-controlled data to stdout
> - seccomp_notify_write() interprets the start of the written data as
> a struct seccomp_notif_resp
> - seccomp_notify_write() grabs the privileged file descriptor and
> installs a copy in process A
> - process A now has access to the privileged file (e.g. /etc/shadow)
> 
> It isn't clear whether it would actually be exploitable - you'd need a
> setuid binary that performs the right actions - but it's still bad.

Jann is right. ->read and ->write must not reference any of the calling task’s state except the literal memory passed in.

> 
> Unless I'm missing something, can you please turn the ->read and
> ->write handlers into an ->unlocked_ioctl handler? Something like
> this:
> 
> struct seccomp_user_notif_args {
>        u64 buf;
>        u64 size;
> };
> 
> static long unlocked_ioctl(struct file *file, unsigned int cmd,
> unsigned long arg)
> {
>        struct seccomp_user_notif_args args;
>        struct seccomp_user_notif_args __user *uargs;
> 
>        if (cmd != SECCOMP_USER_NOTIF_READ && cmd != SECCOMP_USER_NOTIF_WRITE)
>                return -EINVAL;
> 
>        if (copy_from_user(&args, uargs, sizeof(args)))
>                return -EFAULT;
> 
>        switch (cmd) {
>        case SECCOMP_USER_NOTIF_READ:
>                return seccomp_notify_read(file, (char __user
> *)args.buf, (size_t)args.size);
>        case SECCOMP_USER_NOTIF_WRITE:
>                return seccomp_notify_write(file, (char __user
> *)args.buf, (size_t)args.size);
>        default:
>                return -EINVAL;
>        }
> }