[v2,1/2] misc: Add Nitro Secure Module driver

  When running Linux inside a Nitro Enclave, the hypervisor provides a
special virtio device called "NSM". This device has 2 main functions:

  1) Provide attestation reports
  2) Modify PCR state
  3) Provide entropy

This patch adds the core NSM driver that exposes a /dev/nsm device node
which user space can use to request attestation documents and influence
PCR states. A follow up patch will add a hwrng driver to feed its entropy
into the kernel.

Originally-by: Petre Eftime <petre.eftime@gmail.com>
Signed-off-by: Alexander Graf <graf@amazon.com>

---

v1 -> v2:

  - Remove boilerplate
  - Add uapi header
---
 MAINTAINERS              |  10 +
 drivers/misc/Kconfig     |  11 +
 drivers/misc/Makefile    |   1 +
 drivers/misc/nsm.c       | 448 +++++++++++++++++++++++++++++++++++++++
 include/linux/nsm.h      |  35 +++
 include/uapi/linux/nsm.h |  24 +++
 6 files changed, 529 insertions(+)
 create mode 100644 drivers/misc/nsm.c
 create mode 100644 include/linux/nsm.h
 create mode 100644 include/uapi/linux/nsm.h
  

Hi Arnd!

On 29.09.23 19:28, Arnd Bergmann wrote:
> On Fri, Sep 29, 2023, at 09:33, Alexander Graf wrote:
>> When running Linux inside a Nitro Enclave, the hypervisor provides a
>> special virtio device called "NSM". This device has 2 main functions:
>>
>>    1) Provide attestation reports
>>    2) Modify PCR state
>>    3) Provide entropy
>>
>> This patch adds the core NSM driver that exposes a /dev/nsm device node
>> which user space can use to request attestation documents and influence
>> PCR states. A follow up patch will add a hwrng driver to feed its entropy
>> into the kernel.
>>
>> Originally-by: Petre Eftime <petre.eftime@gmail.com>
>> Signed-off-by: Alexander Graf <graf@amazon.com>
> Hi Alex,
>
> I've taken a first look at this driver and have some minor comments.


Thanks a bunch!


> The main point here is that I think we need to look at possible
> alternatives for the user space interface, and (if possible) change
> to a set of higher-level ioctl commands from the simple passthrough.


I'm slightly torn on that bit. I think in hindsight the NSM device 
probably should have been a reserved vsock CID and the hwrng one should 
have just been virtio-rng.

The problem is that Nitro Enclaves were launched in 2020 and since an 
ecosystem developed in multiple languages to support building code inside:

https://github.com/aws/aws-nitro-enclaves-nsm-api/blob/main/src/driver/mod.rs#L66
https://github.com/donkersgoed/aws-nsm-interface/blob/main/aws_nsm_interface/__init__.py#L264-L274
   https://github.com/hf/nsm/blob/main/nsm.go#L99-L129


All of these use the (downstream) ioctl that this patch also implements. 
We could change it, but instead of making it easier for user space to 
adapt the device node, it would probably hurt more.

I agree that this is not a great place to be in. This driver absolutely 
should have been upstreamed 3 years ago. But I can't turn back time 
(yet) :).

>
>> +/* Virtio MMIO device definition */
>> +struct virtio_mmio_device {
>> +     struct virtio_device vdev;
>> +     struct platform_device *pdev;
>> +
>> +     void __iomem *base;
>> +     unsigned long version;
>> +
>> +     /* a list of queues so we can dispatch IRQs */
>> +     spinlock_t lock;
>> +     struct list_head virtqueues;
>> +};
>> +
>> +/* Virtqueue list entry */
>> +struct virtio_mmio_vq_info {
>> +     /* The actual virtqueue */
>> +     struct virtqueue *vq;
>> +
>> +     /* The list node for the virtqueues list */
>> +     struct list_head node;
>> +};
>
> It looks like you are duplicating these structures from the
> virtio_mmio.c file, which seems like a bad idea for a number
> of reasons. What is it that you actually need that the
> virtio subsystem does not provide? Can you add interfaces
> to the common code instead?


Thanks for catching this. There are proper interfaces to get the virt 
queues already, let me use them instead.


>
>> +static struct virtio_device *nsm_vdev;
>> +static struct nsm_hwrng *nsm_hwrng;
>> +static struct mutex nsm_lock;
>> +static wait_queue_head_t nsm_waitqueue;
>> +static bool nsm_device_notified;
> Instead of global structures, these should ideally all be
> part of a per-device structure, even if you are sure that
> there is only ever one of these devices.


Let me give that a try :)


>
>> +/* Copy an entire message from user-space to kernel-space */
>> +static int message_memdup_from_user(struct nsm_kernel_message *dst,
>> +     struct nsm_message *src)
>> +{
>> +     struct nsm_message shallow_copy;
>> +
>> +     if (!src || !dst)
>> +             return -EINVAL;
>> +
>> +     /* The destination's request and response buffers should be NULL. */
>> +     if (dst->request.iov_base || dst->response.iov_base)
>> +             return -EINVAL;
>> +
>> +     /* First, make a shallow copy to be able to read the inner pointers */
>> +     if (copy_from_user(&shallow_copy, src, sizeof(shallow_copy)) != 0)
>> +             return -EINVAL;
>> +
>> +     /* Verify the user input size. */
>> +     if (shallow_copy.request.iov_len > NSM_REQUEST_MAX_SIZE)
>> +             return -EMSGSIZE;
>> +
>> +     /* Allocate kernel memory for the user request */
>> +     dst->request.iov_len = shallow_copy.request.iov_len;
>> +     dst->request.iov_base = kmalloc(dst->request.iov_len, GFP_KERNEL);
>> +     if (!dst->request.iov_base)
>> +             return -ENOMEM;
>> +
>> +     /* Copy the request content */
>> +     if (copy_from_user(dst->request.iov_base,
>> +             shallow_copy.request.iov_base, dst->request.iov_len) != 0) {
>> +             kfree(dst->request.iov_base);
>> +             return -EFAULT;
>> +     }
> It looks like the ioctl interface just provides an interface
> for passing through raw messages, which is often not the best
> idea. Are you able to enumerate the possible request types and
> provide a separate ioctl for each one?


See above. I could, but I think it would not improve the situation. It 
would also require a significant amount of CBOR parsing in the kernel 
which I'd rather avoid :).


>
>> +/* Supported driver operations */
>> +static const struct file_operations nsm_dev_fops = {
>> +     .open = nsm_dev_file_open,
>> +     .release = nsm_dev_file_close,
>> +     .unlocked_ioctl = nsm_dev_ioctl,
>> +};
> This breaks on 32-bit userspace, which would need a separate .compat_ioctl
> handler with the current command definition. It's often better to define
> the ioctl interface to be the same on 32-bit and 64-bit userspace
> and then use the trivial compat_ptr_ioctl wrapper.
>
>> +/* Driver configuration */
>> +static struct miscdevice nsm_driver_miscdevice = {
>> +     .minor  = MISC_DYNAMIC_MINOR,
>> +     .name   = NSM_DEV_NAME,
>> +     .fops   = &nsm_dev_fops,
>> +     .mode   = 0666
>> +};
> I would suggest expanding NSM_DEV_NAME here, it's much easier to
> grep for the actual string if a user wants to know which driver
> is responsible. Probably even less code.
>
>> +     if (nsm_hwrng)
>> +             nsm_hwrng->probe(vdev);
>> +
>> +     pr_debug("NSM device has been probed.\n");
>> +     return 0;
>> +}
> The debug statements can probably get removed, especially
> the whitespace damaged ones.
>
>> +int nsm_register_hwrng(struct nsm_hwrng *_nsm_hwrng)
>> +{
>> +     if (nsm_hwrng)
>> +             return -EEXIST;
>> +
>> +     nsm_hwrng = _nsm_hwrng;
>> +     if (nsm_vdev)
>> +             nsm_hwrng->probe(nsm_vdev);
>> +
>> +     return 0;
>> +}
>> +EXPORT_SYMBOL_GPL(nsm_register_hwrng);
> This should get easier of you reverse the dependency between
> the two drivers and just call into the nsm_hwrng_probe()
> function from the main driver's probe.


I don't understand what you mean by reversing the dependency. Nsm_rng is 
a downstream of Nsm, because Nsm is the virtio device that owns the channel.


>
>> +     mutex_init(&nsm_lock);
>> +     init_waitqueue_head(&nsm_waitqueue);
> You can simply use DEFINE_MUTEX() and DECLARE_WAIT_QUEUE_HEAD()
> if you still need the global objects (rather than making them
> per device).
>
>> +
>> +     rc = register_virtio_driver(&virtio_nsm_driver);
>> +     if (rc)
>> +             pr_err("NSM driver initialization error: %d.\n", rc);
>> +
>> +     return rc;
>> +}
>> +
>> +static void __exit nsm_driver_exit(void)
>> +{
>> +     unregister_virtio_driver(&virtio_nsm_driver);
>> +     mutex_destroy(&nsm_lock);
>> +     pr_debug("NSM driver exited.\n");
>> +}
>> +
>> +module_init(nsm_driver_init);
>> +module_exit(nsm_driver_exit);
> Then this can use module_virtio_driver()


Definitely!

Thanks :). I'll have some typing to do.


Alex


>
>        Arnd



Amazon Development Center Germany GmbH
Krausenstr. 38
10117 Berlin
Geschaeftsfuehrung: Christian Schlaeger, Jonathan Weiss
Eingetragen am Amtsgericht Charlottenburg unter HRB 149173 B
Sitz: Berlin
Ust-ID: DE 289 237 879
  

On Fri, Sep 29, 2023 at 09:26:16PM +0200, Alexander Graf wrote:
> Hi Arnd!
> 
> On 29.09.23 19:28, Arnd Bergmann wrote:
> > On Fri, Sep 29, 2023, at 09:33, Alexander Graf wrote:
> > > When running Linux inside a Nitro Enclave, the hypervisor provides a
> > > special virtio device called "NSM". This device has 2 main functions:
> > > 
> > >    1) Provide attestation reports
> > >    2) Modify PCR state
> > >    3) Provide entropy
> > > 
> > > This patch adds the core NSM driver that exposes a /dev/nsm device node
> > > which user space can use to request attestation documents and influence
> > > PCR states. A follow up patch will add a hwrng driver to feed its entropy
> > > into the kernel.
> > > 
> > > Originally-by: Petre Eftime <petre.eftime@gmail.com>
> > > Signed-off-by: Alexander Graf <graf@amazon.com>
> > Hi Alex,
> > 
> > I've taken a first look at this driver and have some minor comments.
> 
> 
> Thanks a bunch!
> 
> 
> > The main point here is that I think we need to look at possible
> > alternatives for the user space interface, and (if possible) change
> > to a set of higher-level ioctl commands from the simple passthrough.
> 
> 
> I'm slightly torn on that bit. I think in hindsight the NSM device probably
> should have been a reserved vsock CID and the hwrng one should have just
> been virtio-rng.
> 
> The problem is that Nitro Enclaves were launched in 2020 and since an
> ecosystem developed in multiple languages to support building code inside:
> 
> https://github.com/aws/aws-nitro-enclaves-nsm-api/blob/main/src/driver/mod.rs#L66
> https://github.com/donkersgoed/aws-nsm-interface/blob/main/aws_nsm_interface/__init__.py#L264-L274
>   https://github.com/hf/nsm/blob/main/nsm.go#L99-L129
> 
> 
> All of these use the (downstream) ioctl that this patch also implements. We
> could change it, but instead of making it easier for user space to adapt the
> device node, it would probably hurt more.
> 
> I agree that this is not a great place to be in. This driver absolutely
> should have been upstreamed 3 years ago. But I can't turn back time (yet)
> :).

As you know, this is no excuse to put an api in the kernel that isn't
correct or good for the long-term.  Just because people do foolish
things outside of the kernel tree never means we have to accept them in
our tree.  Instead we can ask them to fix them properly as part of us
taking the code.

So please, work on doing this right.

thanks,

greg k-h
  

Hey Greg,

On 30.09.23 08:20, Greg Kroah-Hartman wrote:
> On Fri, Sep 29, 2023 at 09:26:16PM +0200, Alexander Graf wrote:
>> Hi Arnd!
>>
>> On 29.09.23 19:28, Arnd Bergmann wrote:
>>> On Fri, Sep 29, 2023, at 09:33, Alexander Graf wrote:
>>>> When running Linux inside a Nitro Enclave, the hypervisor provides a
>>>> special virtio device called "NSM". This device has 2 main functions:
>>>>
>>>>     1) Provide attestation reports
>>>>     2) Modify PCR state
>>>>     3) Provide entropy
>>>>
>>>> This patch adds the core NSM driver that exposes a /dev/nsm device node
>>>> which user space can use to request attestation documents and influence
>>>> PCR states. A follow up patch will add a hwrng driver to feed its entropy
>>>> into the kernel.
>>>>
>>>> Originally-by: Petre Eftime <petre.eftime@gmail.com>
>>>> Signed-off-by: Alexander Graf <graf@amazon.com>
>>> Hi Alex,
>>>
>>> I've taken a first look at this driver and have some minor comments.
>>
>> Thanks a bunch!
>>
>>
>>> The main point here is that I think we need to look at possible
>>> alternatives for the user space interface, and (if possible) change
>>> to a set of higher-level ioctl commands from the simple passthrough.
>>
>> I'm slightly torn on that bit. I think in hindsight the NSM device probably
>> should have been a reserved vsock CID and the hwrng one should have just
>> been virtio-rng.
>>
>> The problem is that Nitro Enclaves were launched in 2020 and since an
>> ecosystem developed in multiple languages to support building code inside:
>>
>> https://github.com/aws/aws-nitro-enclaves-nsm-api/blob/main/src/driver/mod.rs#L66
>> https://github.com/donkersgoed/aws-nsm-interface/blob/main/aws_nsm_interface/__init__.py#L264-L274
>>    https://github.com/hf/nsm/blob/main/nsm.go#L99-L129
>>
>>
>> All of these use the (downstream) ioctl that this patch also implements. We
>> could change it, but instead of making it easier for user space to adapt the
>> device node, it would probably hurt more.
>>
>> I agree that this is not a great place to be in. This driver absolutely
>> should have been upstreamed 3 years ago. But I can't turn back time (yet)
>> :).
> As you know, this is no excuse to put an api in the kernel that isn't
> correct or good for the long-term.  Just because people do foolish
> things outside of the kernel tree never means we have to accept them in
> our tree.  Instead we can ask them to fix them properly as part of us
> taking the code.
>
> So please, work on doing this right.


Sorry if my message above came over as a push to put an "incorrect api" 
into the kernel.

In situations like this where you can either give user space full access 
to the device's command space through a generic API or you can create 
command awareness in the kernel and make it the kernel's task to learn 
about each command, IMHO it's never a clear cut on which one is better. 
Especially in virtual environments where the set of commands can change 
quickly over time.

So what I was trying to say above is that *if* we consider both paths 
equally viable, I'd err on the one that enables the existing ecosystem. 
However if there are good reasons to not do command pass-through, I'm 
all for abstracting it away :)

Looking at prior art, the most similar implementations to this are TPMs 
and virtio-vsock. With virtio-vsock, kernel space has no idea what it 
talks to on the other hand and makes it 100% user space's problem. With 
TPMs, you typically use /dev/tpm0 to gain raw command access to the 
target device. So while we could engineer something smarter here, I'm 
not convinced yet it's a net win.


Alex




Amazon Development Center Germany GmbH
Krausenstr. 38
10117 Berlin
Geschaeftsfuehrung: Christian Schlaeger, Jonathan Weiss
Eingetragen am Amtsgericht Charlottenburg unter HRB 149173 B
Sitz: Berlin
Ust-ID: DE 289 237 879
  

On Mon, Oct 2, 2023, at 14:28, Alexander Graf wrote:
> On 30.09.23 08:20, Greg Kroah-Hartman wrote:
>> On Fri, Sep 29, 2023 at 09:26:16PM +0200, Alexander Graf wrote:
>>> On 29.09.23 19:28, Arnd Bergmann wrote:

>>> All of these use the (downstream) ioctl that this patch also implements. We
>>> could change it, but instead of making it easier for user space to adapt the
>>> device node, it would probably hurt more.
>>>
>>> I agree that this is not a great place to be in. This driver absolutely
>>> should have been upstreamed 3 years ago. But I can't turn back time (yet)
>>> :).
>> As you know, this is no excuse to put an api in the kernel that isn't
>> correct or good for the long-term.  Just because people do foolish
>> things outside of the kernel tree never means we have to accept them in
>> our tree.  Instead we can ask them to fix them properly as part of us
>> taking the code.
>>
>> So please, work on doing this right.
>
>
> Sorry if my message above came over as a push to put an "incorrect api" 
> into the kernel.
>
> In situations like this where you can either give user space full access 
> to the device's command space through a generic API or you can create 
> command awareness in the kernel and make it the kernel's task to learn 
> about each command, IMHO it's never a clear cut on which one is better. 
> Especially in virtual environments where the set of commands can change 
> quickly over time.
>
> So what I was trying to say above is that *if* we consider both paths 
> equally viable, I'd err on the one that enables the existing ecosystem. 
> However if there are good reasons to not do command pass-through, I'm 
> all for abstracting it away :)
>
> Looking at prior art, the most similar implementations to this are TPMs 
> and virtio-vsock. With virtio-vsock, kernel space has no idea what it 
> talks to on the other hand and makes it 100% user space's problem. With 
> TPMs, you typically use /dev/tpm0 to gain raw command access to the 
> target device. So while we could engineer something smarter here, I'm 
> not convinced yet it's a net win.

Generally speaking, I can see a number of advantages to using an
in-kernel abstraction:

- if there are both in-kernel and userspace API users, or multiple
  concurrent userspace clients, an abstraction layer helps to serialize
  between any stateful commands.

- in an abstract interface, the kernel can enforce command specific
  permission checks, rather than allowing access either to all or none
  of the commands.

- having the actual commands created by the kernel means that a bug
  in the virtio device implementation parsing the commands is less
  likely to be exploitable from user space.

- An explicit set of defined ioctl commands is easier to review and
  audit for kernel developers as we try to ensure that this is a
  sensible kernel interface

I don't know enough about your use cases or the specific
command set to tell if any of those points actually matter
here. In the python implementation you linked to, there are
only a handful of commands that actually get passed through.
It should be fairly easy to prototype a kernel driver that
implements them as individual ioctl commands, to give us a
better idea of how this compares to the current code.

     Arnd
  

On Fri, Sep 29, 2023, at 21:26, Alexander Graf wrote:
> On 29.09.23 19:28, Arnd Bergmann wrote:
>> On Fri, Sep 29, 2023, at 09:33, Alexander Graf wrote:

>>
>>> +int nsm_register_hwrng(struct nsm_hwrng *_nsm_hwrng)
>>> +{
>>> +     if (nsm_hwrng)
>>> +             return -EEXIST;
>>> +
>>> +     nsm_hwrng = _nsm_hwrng;
>>> +     if (nsm_vdev)
>>> +             nsm_hwrng->probe(nsm_vdev);
>>> +
>>> +     return 0;
>>> +}
>>> +EXPORT_SYMBOL_GPL(nsm_register_hwrng);
>> This should get easier of you reverse the dependency between
>> the two drivers and just call into the nsm_hwrng_probe()
>> function from the main driver's probe.
>
>
> I don't understand what you mean by reversing the dependency. Nsm_rng is 
> a downstream of Nsm, because Nsm is the virtio device that owns the channel.

To clarify: I don't really consider Nsm_rng as a downstream driver,
since the main driver does not have a bus model or device detection
below it that would allow probing any function specific drivers.

I would remove the dynamic registration entirely and just merge both
parts into a single file that happens to expose an rng interface.

     Arnd