2023-07-04 16:54:08

by Maxime Coquelin

[permalink] [raw]
Subject: [PATCH v2 3/3] vduse: Temporarily disable control queue features

Virtio-net driver control queue implementation is not safe
when used with VDUSE. If the VDUSE application does not
reply to control queue messages, it currently ends up
hanging the kernel thread sending this command.

Some work is on-going to make the control queue
implementation robust with VDUSE. Until it is completed,
let's disable control virtqueue and features that depend on
it.

Signed-off-by: Maxime Coquelin <[email protected]>
---
drivers/vdpa/vdpa_user/vduse_dev.c | 21 +++++++++++++++++++++
1 file changed, 21 insertions(+)

diff --git a/drivers/vdpa/vdpa_user/vduse_dev.c b/drivers/vdpa/vdpa_user/vduse_dev.c
index 1271c9796517..04367a53802b 100644
--- a/drivers/vdpa/vdpa_user/vduse_dev.c
+++ b/drivers/vdpa/vdpa_user/vduse_dev.c
@@ -1778,6 +1778,25 @@ static struct attribute *vduse_dev_attrs[] = {

ATTRIBUTE_GROUPS(vduse_dev);

+static void vduse_dev_features_fixup(struct vduse_dev_config *config)
+{
+ if (config->device_id == VIRTIO_ID_NET) {
+ /*
+ * Temporarily disable control virtqueue and features that
+ * depend on it while CVQ is being made more robust for VDUSE.
+ */
+ config->features &= ~((1ULL << VIRTIO_NET_F_CTRL_VQ) |
+ (1ULL << VIRTIO_NET_F_CTRL_RX) |
+ (1ULL << VIRTIO_NET_F_CTRL_VLAN) |
+ (1ULL << VIRTIO_NET_F_GUEST_ANNOUNCE) |
+ (1ULL << VIRTIO_NET_F_MQ) |
+ (1ULL << VIRTIO_NET_F_CTRL_MAC_ADDR) |
+ (1ULL << VIRTIO_NET_F_RSS) |
+ (1ULL << VIRTIO_NET_F_HASH_REPORT) |
+ (1ULL << VIRTIO_NET_F_NOTF_COAL));
+ }
+}
+
static int vduse_create_dev(struct vduse_dev_config *config,
void *config_buf, u64 api_version)
{
@@ -1793,6 +1812,8 @@ static int vduse_create_dev(struct vduse_dev_config *config,
if (!dev)
goto err;

+ vduse_dev_features_fixup(config);
+
dev->api_version = api_version;
dev->device_features = config->features;
dev->device_id = config->device_id;
--
2.41.0



2023-07-04 16:59:22

by Michael S. Tsirkin

[permalink] [raw]
Subject: Re: [PATCH v2 3/3] vduse: Temporarily disable control queue features

On Tue, Jul 04, 2023 at 06:40:45PM +0200, Maxime Coquelin wrote:
> Virtio-net driver control queue implementation is not safe
> when used with VDUSE. If the VDUSE application does not
> reply to control queue messages, it currently ends up
> hanging the kernel thread sending this command.
>
> Some work is on-going to make the control queue
> implementation robust with VDUSE. Until it is completed,
> let's disable control virtqueue and features that depend on
> it.
>
> Signed-off-by: Maxime Coquelin <[email protected]>
> ---
> drivers/vdpa/vdpa_user/vduse_dev.c | 21 +++++++++++++++++++++
> 1 file changed, 21 insertions(+)
>
> diff --git a/drivers/vdpa/vdpa_user/vduse_dev.c b/drivers/vdpa/vdpa_user/vduse_dev.c
> index 1271c9796517..04367a53802b 100644
> --- a/drivers/vdpa/vdpa_user/vduse_dev.c
> +++ b/drivers/vdpa/vdpa_user/vduse_dev.c
> @@ -1778,6 +1778,25 @@ static struct attribute *vduse_dev_attrs[] = {
>
> ATTRIBUTE_GROUPS(vduse_dev);
>
> +static void vduse_dev_features_fixup(struct vduse_dev_config *config)
> +{
> + if (config->device_id == VIRTIO_ID_NET) {
> + /*
> + * Temporarily disable control virtqueue and features that
> + * depend on it while CVQ is being made more robust for VDUSE.
> + */
> + config->features &= ~((1ULL << VIRTIO_NET_F_CTRL_VQ) |
> + (1ULL << VIRTIO_NET_F_CTRL_RX) |
> + (1ULL << VIRTIO_NET_F_CTRL_VLAN) |
> + (1ULL << VIRTIO_NET_F_GUEST_ANNOUNCE) |
> + (1ULL << VIRTIO_NET_F_MQ) |
> + (1ULL << VIRTIO_NET_F_CTRL_MAC_ADDR) |
> + (1ULL << VIRTIO_NET_F_RSS) |
> + (1ULL << VIRTIO_NET_F_HASH_REPORT) |
> + (1ULL << VIRTIO_NET_F_NOTF_COAL));
> + }
> +}
> +


This will never be exhaustive, we are adding new features.
Please add an allowlist with just legal ones instead.


> static int vduse_create_dev(struct vduse_dev_config *config,
> void *config_buf, u64 api_version)
> {
> @@ -1793,6 +1812,8 @@ static int vduse_create_dev(struct vduse_dev_config *config,
> if (!dev)
> goto err;
>
> + vduse_dev_features_fixup(config);
> +
> dev->api_version = api_version;
> dev->device_features = config->features;
> dev->device_id = config->device_id;
> --
> 2.41.0


2023-07-04 19:36:00

by Maxime Coquelin

[permalink] [raw]
Subject: Re: [PATCH v2 3/3] vduse: Temporarily disable control queue features



On 7/4/23 18:43, Michael S. Tsirkin wrote:
> On Tue, Jul 04, 2023 at 06:40:45PM +0200, Maxime Coquelin wrote:
>> Virtio-net driver control queue implementation is not safe
>> when used with VDUSE. If the VDUSE application does not
>> reply to control queue messages, it currently ends up
>> hanging the kernel thread sending this command.
>>
>> Some work is on-going to make the control queue
>> implementation robust with VDUSE. Until it is completed,
>> let's disable control virtqueue and features that depend on
>> it.
>>
>> Signed-off-by: Maxime Coquelin <[email protected]>
>> ---
>> drivers/vdpa/vdpa_user/vduse_dev.c | 21 +++++++++++++++++++++
>> 1 file changed, 21 insertions(+)
>>
>> diff --git a/drivers/vdpa/vdpa_user/vduse_dev.c b/drivers/vdpa/vdpa_user/vduse_dev.c
>> index 1271c9796517..04367a53802b 100644
>> --- a/drivers/vdpa/vdpa_user/vduse_dev.c
>> +++ b/drivers/vdpa/vdpa_user/vduse_dev.c
>> @@ -1778,6 +1778,25 @@ static struct attribute *vduse_dev_attrs[] = {
>>
>> ATTRIBUTE_GROUPS(vduse_dev);
>>
>> +static void vduse_dev_features_fixup(struct vduse_dev_config *config)
>> +{
>> + if (config->device_id == VIRTIO_ID_NET) {
>> + /*
>> + * Temporarily disable control virtqueue and features that
>> + * depend on it while CVQ is being made more robust for VDUSE.
>> + */
>> + config->features &= ~((1ULL << VIRTIO_NET_F_CTRL_VQ) |
>> + (1ULL << VIRTIO_NET_F_CTRL_RX) |
>> + (1ULL << VIRTIO_NET_F_CTRL_VLAN) |
>> + (1ULL << VIRTIO_NET_F_GUEST_ANNOUNCE) |
>> + (1ULL << VIRTIO_NET_F_MQ) |
>> + (1ULL << VIRTIO_NET_F_CTRL_MAC_ADDR) |
>> + (1ULL << VIRTIO_NET_F_RSS) |
>> + (1ULL << VIRTIO_NET_F_HASH_REPORT) |
>> + (1ULL << VIRTIO_NET_F_NOTF_COAL));
>> + }
>> +}
>> +
>
>
> This will never be exhaustive, we are adding new features.
> Please add an allowlist with just legal ones instead.

Ok, got it!
I'll post a new revision.

Thanks,
Maxime

>
>
>> static int vduse_create_dev(struct vduse_dev_config *config,
>> void *config_buf, u64 api_version)
>> {
>> @@ -1793,6 +1812,8 @@ static int vduse_create_dev(struct vduse_dev_config *config,
>> if (!dev)
>> goto err;
>>
>> + vduse_dev_features_fixup(config);
>> +
>> dev->api_version = api_version;
>> dev->device_features = config->features;
>> dev->device_id = config->device_id;
>> --
>> 2.41.0
>