[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <8355f9b3-4cda-cd2e-98df-fed020193008@redhat.com>
Date: Wed, 24 Feb 2021 17:29:07 +0800
From: Jason Wang <jasowang@...hat.com>
To: Cornelia Huck <cohuck@...hat.com>
Cc: "Michael S. Tsirkin" <mst@...hat.com>,
Si-Wei Liu <si-wei.liu@...cle.com>, elic@...dia.com,
linux-kernel@...r.kernel.org,
virtualization@...ts.linux-foundation.org, netdev@...r.kernel.org,
virtio-dev@...ts.oasis-open.org
Subject: Re: [virtio-dev] Re: [PATCH] vdpa/mlx5: set_features should allow
reset to zero
On 2021/2/23 6:58 下午, Cornelia Huck wrote:
> On Tue, 23 Feb 2021 18:31:07 +0800
> Jason Wang <jasowang@...hat.com> wrote:
>
>> On 2021/2/23 6:04 下午, Cornelia Huck wrote:
>>> On Tue, 23 Feb 2021 17:46:20 +0800
>>> Jason Wang <jasowang@...hat.com> wrote:
>>>
>>>> On 2021/2/23 下午5:25, Michael S. Tsirkin wrote:
>>>>> On Mon, Feb 22, 2021 at 09:09:28AM -0800, Si-Wei Liu wrote:
>>>>>> On 2/21/2021 8:14 PM, Jason Wang wrote:
>>>>>>> On 2021/2/19 7:54 下午, Si-Wei Liu wrote:
>>>>>>>> Commit 452639a64ad8 ("vdpa: make sure set_features is invoked
>>>>>>>> for legacy") made an exception for legacy guests to reset
>>>>>>>> features to 0, when config space is accessed before features
>>>>>>>> are set. We should relieve the verify_min_features() check
>>>>>>>> and allow features reset to 0 for this case.
>>>>>>>>
>>>>>>>> It's worth noting that not just legacy guests could access
>>>>>>>> config space before features are set. For instance, when
>>>>>>>> feature VIRTIO_NET_F_MTU is advertised some modern driver
>>>>>>>> will try to access and validate the MTU present in the config
>>>>>>>> space before virtio features are set.
>>>>>>> This looks like a spec violation:
>>>>>>>
>>>>>>> "
>>>>>>>
>>>>>>> The following driver-read-only field, mtu only exists if
>>>>>>> VIRTIO_NET_F_MTU is set. This field specifies the maximum MTU for the
>>>>>>> driver to use.
>>>>>>> "
>>>>>>>
>>>>>>> Do we really want to workaround this?
>>>>>> Isn't the commit 452639a64ad8 itself is a workaround for legacy guest?
>>>>>>
>>>>>> I think the point is, since there's legacy guest we'd have to support, this
>>>>>> host side workaround is unavoidable. Although I agree the violating driver
>>>>>> should be fixed (yes, it's in today's upstream kernel which exists for a
>>>>>> while now).
>>>>> Oh you are right:
>>>>>
>>>>>
>>>>> static int virtnet_validate(struct virtio_device *vdev)
>>>>> {
>>>>> if (!vdev->config->get) {
>>>>> dev_err(&vdev->dev, "%s failure: config access disabled\n",
>>>>> __func__);
>>>>> return -EINVAL;
>>>>> }
>>>>>
>>>>> if (!virtnet_validate_features(vdev))
>>>>> return -EINVAL;
>>>>>
>>>>> if (virtio_has_feature(vdev, VIRTIO_NET_F_MTU)) {
>>>>> int mtu = virtio_cread16(vdev,
>>>>> offsetof(struct virtio_net_config,
>>>>> mtu));
>>>>> if (mtu < MIN_MTU)
>>>>> __virtio_clear_bit(vdev, VIRTIO_NET_F_MTU);
>>>> I wonder why not simply fail here?
>>> I think both failing or not accepting the feature can be argued to make
>>> sense: "the device presented us with a mtu size that does not make
>>> sense" would point to failing, "we cannot work with the mtu size that
>>> the device presented us" would point to not negotiating the feature.
>>>
>>>>
>>>>> }
>>>>>
>>>>> return 0;
>>>>> }
>>>>>
>>>>> And the spec says:
>>>>>
>>>>>
>>>>> The driver MUST follow this sequence to initialize a device:
>>>>> 1. Reset the device.
>>>>> 2. Set the ACKNOWLEDGE status bit: the guest OS has noticed the device.
>>>>> 3. Set the DRIVER status bit: the guest OS knows how to drive the device.
>>>>> 4. Read device feature bits, and write the subset of feature bits understood by the OS and driver to the
>>>>> device. During this step the driver MAY read (but MUST NOT write) the device-specific configuration
>>>>> fields to check that it can support the device before accepting it.
>>>>> 5. Set the FEATURES_OK status bit. The driver MUST NOT accept new feature bits after this step.
>>>>> 6. Re-read device status to ensure the FEATURES_OK bit is still set: otherwise, the device does not
>>>>> support our subset of features and the device is unusable.
>>>>> 7. Perform device-specific setup, including discovery of virtqueues for the device, optional per-bus setup,
>>>>> reading and possibly writing the device’s virtio configuration space, and population of virtqueues.
>>>>> 8. Set the DRIVER_OK status bit. At this point the device is “live”.
>>>>>
>>>>>
>>>>> Item 4 on the list explicitly allows reading config space before
>>>>> FEATURES_OK.
>>>>>
>>>>> I conclude that VIRTIO_NET_F_MTU is set means "set in device features".
>>>> So this probably need some clarification. "is set" is used many times in
>>>> the spec that has different implications.
>>> Before FEATURES_OK is set by the driver, I guess it means "the device
>>> has offered the feature";
>>
>> For me this part is ok since it clarify that it's the driver that set
>> the bit.
>>
>>
>>
>>> during normal usage, it means "the feature
>>> has been negotiated".
>> /?
>>
>> It looks to me the feature negotiation is done only after device set
>> FEATURES_OK, or FEATURES_OK could be read from device status?
> I'd consider feature negotiation done when the driver reads FEATURES_OK
> back from the status.
I agree.
>
>>
>>> (This is a bit fuzzy for legacy mode.)
> ...because legacy does not have FEATURES_OK.
>
>>
>> The problem is the MTU description for example:
>>
>> "The following driver-read-only field, mtu only exists if
>> VIRTIO_NET_F_MTU is set."
>>
>> It looks to me need to use "if VIRTIO_NET_F_MTU is set by device".
> "offered by the device"? I don't think it should 'disappear' from the
> config space if the driver won't use it. (Same for other config space
> fields that are tied to feature bits.)
But what happens if e.g device doesn't offer VIRTIO_NET_F_MTU? It looks
to according to the spec there will be no mtu field.
And a more interesting case is VIRTIO_NET_F_MQ is not offered but
VIRTIO_NET_F_MTU offered. To me, it means we don't have
max_virtqueue_pairs but it's not how the driver is wrote today.
>
>> Otherwise readers (at least for me), may think the MTU is only valid
>> if driver set the bit.
> I think it would still be 'valid' in the sense that it exists and has
> some value in there filled in by the device, but a driver reading it
> without negotiating the feature would be buggy. (Like in the kernel
> code above; the kernel not liking the value does not make the field
> invalid.)
See Michael's reply, the spec allows read the config before setting
features.
>
> Maybe a statement covering everything would be:
>
> "The following driver-read-only field mtu only exists if the device
> offers VIRTIO_NET_F_MTU and may be read by the driver during feature
> negotiation and after VIRTIO_NET_F_MTU has been successfully
> negotiated."
>
>>
>>> Should we add a wording clarification to the spec?
>>
>> I think so.
> Some clarification would be needed for each field that depends on a
> feature; that would be quite verbose. Maybe we can get away with a
> clarifying statement?
>
> "Some config space fields may depend on a certain feature. In that
> case, the field exits if the device has offered the corresponding
> feature,
So this implies for !VIRTIO_NET_F_MQ && VIRTIO_NET_F_MTU, the config
will look like:
struct virtio_net_config {
u8 mac[6];
le16 status;
le16 mtu;
};
> and may be read by the driver during feature negotiation, and
> accessed by the driver after the feature has been successfully
> negotiated. A shorthand for this is a statement that a field only
> exists if a certain feature bit is set."
I'm not sure using "shorthand" is good for the spec, at least we can
limit the its scope only to the configuration space part.
Thanks
>
Powered by blists - more mailing lists