Skip to contentSkip to conversation

vdpa/vdpa_sim: Use the negotiated features when calling vringh_init_iotlb

PATCH In review Eli Cohen 3 participants 5 messages Jul 20, 2021 – Aug 10, 2021 virtio core

lore
Share & tools
lore

Eli Cohen

author
#0 lore raw

PATCH 1/1 vdpa/vdpa_sim: Use the negotiated features when calling vringh_init_iotlb

When calling vringh_init_iotlb(), use the negotiated features which might be different than the supported features.

Fixes 2c53d0f64c06f ("vdpasim: vDPA device simulator")

---
v0 --> v1:
Update "Fixes" line

 drivers/vdpa/vdpa_sim/vdpa_sim.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/vdpa/vdpa_sim/vdpa_sim.c +2 -2
index 14e024de5cbf..89a474c7a096 100644
@@ -66,7 +66,7 @@ static void vdpasim_queue_ready(struct vdpasim *vdpasim, unsigned int idx)
 {
 	struct vdpasim_virtqueue *vq = &vdpasim->vqs[idx];
 
-	vringh_init_iotlb(&vq->vring, vdpasim->dev_attr.supported_features,
+	vringh_init_iotlb(&vq->vring, vdpasim->features,
 			  VDPASIM_QUEUE_MAX, false,
 			  (struct vring_desc *)(uintptr_t)vq->desc_addr,
 			  (struct vring_avail *)
@@ -86,7 +86,7 @@ static void vdpasim_vq_reset(struct vdpasim *vdpasim,
 	vq->device_addr = 0;
 	vq->cb = NULL;
 	vq->private = NULL;
-	vringh_init_iotlb(&vq->vring, vdpasim->dev_attr.supported_features,
+	vringh_init_iotlb(&vq->vring, vdpasim->features,
 			  VDPASIM_QUEUE_MAX, false, NULL, NULL, NULL);
 
 	vq->vring.notify = NULL;
Signature · 1 line
2.30.1
#1 lore raw
Quoted 26 lines · Eli Cohen

When calling vringh_init_iotlb(), use the negotiated features which might be different than the supported features.

Fixes 2c53d0f64c06f ("vdpasim: vDPA device simulator")

---
v0 --> v1:
Update "Fixes" line

drivers/vdpa/vdpa_sim/vdpa_sim.c | 4 ++--

1 file changed, 2 insertions(+), 2 deletions(-)
drivers/vdpa/vdpa_sim/vdpa_sim.c +1 -1
index 14e024de5cbf..89a474c7a096 100644
@@ -66,7 +66,7 @@ static void vdpasim_queue_ready(struct vdpasim *vdpasim, unsigned int idx)
 {
 	struct vdpasim_virtqueue *vq = &vdpasim->vqs[idx];
 
-	vringh_init_iotlb(&vq->vring, vdpasim->dev_attr.supported_features,
+	vringh_init_iotlb(&vq->vring, vdpasim->features,
 			  VDPASIM_QUEUE_MAX, false,
 			  (struct vring_desc *)(uintptr_t)vq->desc_addr,
 			  (struct vring_avail *)

drivers/vdpa/vdpa_sim/vdpa_sim.c static void vdpasim_vq_reset(struct vdpasim *vdpasim, · line 89

Quoted 5 lines
drivers/vdpa/vdpa_sim/vdpa_sim.c +1 -1
@@ -86,7 +86,7 @@ static void vdpasim_vq_reset(struct vdpasim *vdpasim,
 	vq->device_addr = 0;
 	vq->cb = NULL;
 	vq->private = NULL;
-	vringh_init_iotlb(&vq->vring, vdpasim->dev_attr.supported_features,
+	vringh_init_iotlb(&vq->vring, vdpasim->features,

vdpasim_vq_reset() is called while resetting the device in vdpasim_reset() where we also set vdpasim->features = 0 after resetting the vqs, so maybe it's better to use the supported features here, since the negotiated ones are related to the previous instance.

Thanks,
Stefano

_______________________________________________
Virtualization mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/virtualization

author
#2 lore raw
Quoted 21 lines · Stefano Garzarella
Quoted 20 lines · Eli Cohen

When calling vringh_init_iotlb(), use the negotiated features which might be different than the supported features.

Fixes 2c53d0f64c06f ("vdpasim: vDPA device simulator")

---
v0 --> v1:
Update "Fixes" line

drivers/vdpa/vdpa_sim/vdpa_sim.c | 4 ++--

1 file changed, 2 insertions(+), 2 deletions(-)
drivers/vdpa/vdpa_sim/vdpa_sim.c +0 -0
index 14e024de5cbf..89a474c7a096 100644
@@ -66,7 +66,7 @@ static void vdpasim_queue_ready(struct vdpasim *vdpasim, unsigned int idx)
 {
 	struct vdpasim_virtqueue *vq = &vdpasim->vqs[idx];
 
Quoted 17 lines

drivers/vdpa/vdpa_sim/vdpa_sim.c static void vdpasim_vq_reset(struct vdpasim *vdpasim, · line 89

Quoted 11 lines
drivers/vdpa/vdpa_sim/vdpa_sim.c +2 -2
-	vringh_init_iotlb(&vq->vring, vdpasim->dev_attr.supported_features,
+	vringh_init_iotlb(&vq->vring, vdpasim->features,
 			  VDPASIM_QUEUE_MAX, false,
 			  (struct vring_desc *)(uintptr_t)vq->desc_addr,
 			  (struct vring_avail *)
@@ -86,7 +86,7 @@ static void vdpasim_vq_reset(struct vdpasim *vdpasim,
 	vq->device_addr = 0;
 	vq->cb = NULL;
 	vq->private = NULL;
-	vringh_init_iotlb(&vq->vring, vdpasim->dev_attr.supported_features,
+	vringh_init_iotlb(&vq->vring, vdpasim->features,

vdpasim_vq_reset() is called while resetting the device in vdpasim_reset() where we also set vdpasim->features = 0 after resetting the vqs, so maybe it's better to use the supported features here, since the negotiated ones are related to the previous instance.

I don't think using supported features is valid. Better to make sure vringh_init_iotlb() is called after the features have been negotiated.

Quoted 3 lines

Thanks,
Stefano

#3 lore raw
Quoted 40 lines · Eli Cohen
Quoted 21 lines · Stefano Garzarella
Quoted 20 lines · Eli Cohen

When calling vringh_init_iotlb(), use the negotiated features which might be different than the supported features.

Fixes 2c53d0f64c06f ("vdpasim: vDPA device simulator")

---
v0 --> v1:
Update "Fixes" line

drivers/vdpa/vdpa_sim/vdpa_sim.c | 4 ++--

1 file changed, 2 insertions(+), 2 deletions(-)
drivers/vdpa/vdpa_sim/vdpa_sim.c +0 -0
index 14e024de5cbf..89a474c7a096 100644
@@ -66,7 +66,7 @@ static void vdpasim_queue_ready(struct vdpasim *vdpasim, unsigned int idx)
 {
 	struct vdpasim_virtqueue *vq = &vdpasim->vqs[idx];
 
Quoted 17 lines

drivers/vdpa/vdpa_sim/vdpa_sim.c static void vdpasim_vq_reset(struct vdpasim *vdpasim, · line 89

Quoted 11 lines
drivers/vdpa/vdpa_sim/vdpa_sim.c +2 -2
-	vringh_init_iotlb(&vq->vring, vdpasim->dev_attr.supported_features,
+	vringh_init_iotlb(&vq->vring, vdpasim->features,
 			  VDPASIM_QUEUE_MAX, false,
 			  (struct vring_desc *)(uintptr_t)vq->desc_addr,
 			  (struct vring_avail *)
@@ -86,7 +86,7 @@ static void vdpasim_vq_reset(struct vdpasim *vdpasim,
 	vq->device_addr = 0;
 	vq->cb = NULL;
 	vq->private = NULL;
-	vringh_init_iotlb(&vq->vring, vdpasim->dev_attr.supported_features,
+	vringh_init_iotlb(&vq->vring, vdpasim->features,

vdpasim_vq_reset() is called while resetting the device in vdpasim_reset() where we also set vdpasim->features = 0 after resetting the vqs, so maybe it's better to use the supported features here, since the negotiated ones are related to the previous instance.

Quoted 3 lines

I don't think using supported features is valid. Better to make sure vringh_init_iotlb() is called after the features have been negotiated.

I think the vringh_init_iotlb() call in vdpasim_vq_reset() is just used to clean up the struct vringh, then it will be initialized in vdpasim_queue_ready() when features have already been negotiated.

Maybe here we can pass 0 (to the features parameter) if we don't want to use the features supported by the device.

Thanks,
Stefano

_______________________________________________
Virtualization mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/virtualization

#4 lore raw
Quoted 49 lines · Stefano Garzarella
Quoted 40 lines · Eli Cohen
Quoted 21 lines · Stefano Garzarella
Quoted 20 lines · Eli Cohen

When calling vringh_init_iotlb(), use the negotiated features which might be different than the supported features.

Fixes 2c53d0f64c06f ("vdpasim: vDPA device simulator")

---
v0 --> v1:
Update "Fixes" line

drivers/vdpa/vdpa_sim/vdpa_sim.c | 4 ++--

1 file changed, 2 insertions(+), 2 deletions(-)
drivers/vdpa/vdpa_sim/vdpa_sim.c +0 -0
index 14e024de5cbf..89a474c7a096 100644
@@ -66,7 +66,7 @@ static void vdpasim_queue_ready(struct vdpasim *vdpasim, unsigned int idx)
 {
 	struct vdpasim_virtqueue *vq = &vdpasim->vqs[idx];
 
Quoted 17 lines

drivers/vdpa/vdpa_sim/vdpa_sim.c static void vdpasim_vq_reset(struct vdpasim *vdpasim, · line 89

Quoted 11 lines
drivers/vdpa/vdpa_sim/vdpa_sim.c +2 -2
-	vringh_init_iotlb(&vq->vring, vdpasim->dev_attr.supported_features,
+	vringh_init_iotlb(&vq->vring, vdpasim->features,
 			  VDPASIM_QUEUE_MAX, false,
 			  (struct vring_desc *)(uintptr_t)vq->desc_addr,
 			  (struct vring_avail *)
@@ -86,7 +86,7 @@ static void vdpasim_vq_reset(struct vdpasim *vdpasim,
 	vq->device_addr = 0;
 	vq->cb = NULL;
 	vq->private = NULL;
-	vringh_init_iotlb(&vq->vring, vdpasim->dev_attr.supported_features,
+	vringh_init_iotlb(&vq->vring, vdpasim->features,

vdpasim_vq_reset() is called while resetting the device in vdpasim_reset() where we also set vdpasim->features = 0 after resetting the vqs, so maybe it's better to use the supported features here, since the negotiated ones are related to the previous instance.

Quoted 3 lines

I don't think using supported features is valid. Better to make sure vringh_init_iotlb() is called after the features have been negotiated.

I think the vringh_init_iotlb() call in vdpasim_vq_reset() is just used to clean up the struct vringh, then it will be initialized in vdpasim_queue_ready() when features have already been negotiated.

Quoted 5 lines

Maybe here we can pass 0 (to the features parameter) if we don't want to use the features supported by the device.

Thanks,
Stefano

Eli? Maybe you can describe what is the observed bug the patch is trying to fix.

Thanks!

Signature · 1 line
MST