Re: [PATCH V2 4/6] cpr: relax vhost migration blockers
Copy qemu-block, virtio
On Wed, Oct 25, 2023 at 12:44:27PM -0700, Steve Sistare wrote:
> vhost blocks migration if logging is not supported to track dirty
> memory, and vhost-user blocks it if the log cannot be saved to a shm fd.
>
> vhost-vdpa blocks migration if both hosts do not support all the device's
> features using a shadow VQ, for tracking requests and dirty memory.
>
> vhost-scsi blocks migration if storage cannot be shared across hosts,
> or if state cannot be migrated.
>
> None of these conditions apply if the old and new qemu processes do
> not run concurrently, and if new qemu starts on the same host as old,
> which is the case for cpr.
>
> Narrow the scope of these blockers so they only apply to normal mode.
> They will not block cpr modes when they are added in subsequent patches.
>
> No functional change until a new mode is added.
>
> Signed-off-by: Steve Sistare
> Reviewed-by: Juan Quintela
> ---
> hw/scsi/vhost-scsi.c | 2 +-
> hw/virtio/vhost.c| 2 +-
> 2 files changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/hw/scsi/vhost-scsi.c b/hw/scsi/vhost-scsi.c
> index 14e23cc..bf528d5 100644
> --- a/hw/scsi/vhost-scsi.c
> +++ b/hw/scsi/vhost-scsi.c
> @@ -208,7 +208,7 @@ static void vhost_scsi_realize(DeviceState *dev, Error
> **errp)
> "When external environment supports it (Orchestrator
> migrates "
> "target SCSI device state or use shared storage over
> network), "
> "set 'migratable' property to true to enable migration.");
> -if (migrate_add_blocker(>migration_blocker, errp) < 0) {
> +if (migrate_add_blocker_normal(>migration_blocker, errp) < 0) {
> goto free_virtio;
> }
> }
> diff --git a/hw/virtio/vhost.c b/hw/virtio/vhost.c
> index d737671..f5e9625 100644
> --- a/hw/virtio/vhost.c
> +++ b/hw/virtio/vhost.c
> @@ -1527,7 +1527,7 @@ int vhost_dev_init(struct vhost_dev *hdev, void *opaque,
> }
>
> if (hdev->migration_blocker != NULL) {
> -r = migrate_add_blocker(>migration_blocker, errp);
> +r = migrate_add_blocker_normal(>migration_blocker, errp);
> if (r < 0) {
> goto fail_busyloop;
> }
> --
> 1.8.3.1
>
--
Peter Xu
[PATCH V2 4/6] cpr: relax vhost migration blockers
vhost blocks migration if logging is not supported to track dirty
memory, and vhost-user blocks it if the log cannot be saved to a shm fd.
vhost-vdpa blocks migration if both hosts do not support all the device's
features using a shadow VQ, for tracking requests and dirty memory.
vhost-scsi blocks migration if storage cannot be shared across hosts,
or if state cannot be migrated.
None of these conditions apply if the old and new qemu processes do
not run concurrently, and if new qemu starts on the same host as old,
which is the case for cpr.
Narrow the scope of these blockers so they only apply to normal mode.
They will not block cpr modes when they are added in subsequent patches.
No functional change until a new mode is added.
Signed-off-by: Steve Sistare
Reviewed-by: Juan Quintela
---
hw/scsi/vhost-scsi.c | 2 +-
hw/virtio/vhost.c| 2 +-
2 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/hw/scsi/vhost-scsi.c b/hw/scsi/vhost-scsi.c
index 14e23cc..bf528d5 100644
--- a/hw/scsi/vhost-scsi.c
+++ b/hw/scsi/vhost-scsi.c
@@ -208,7 +208,7 @@ static void vhost_scsi_realize(DeviceState *dev, Error
**errp)
"When external environment supports it (Orchestrator migrates "
"target SCSI device state or use shared storage over network),
"
"set 'migratable' property to true to enable migration.");
-if (migrate_add_blocker(>migration_blocker, errp) < 0) {
+if (migrate_add_blocker_normal(>migration_blocker, errp) < 0) {
goto free_virtio;
}
}
diff --git a/hw/virtio/vhost.c b/hw/virtio/vhost.c
index d737671..f5e9625 100644
--- a/hw/virtio/vhost.c
+++ b/hw/virtio/vhost.c
@@ -1527,7 +1527,7 @@ int vhost_dev_init(struct vhost_dev *hdev, void *opaque,
}
if (hdev->migration_blocker != NULL) {
-r = migrate_add_blocker(>migration_blocker, errp);
+r = migrate_add_blocker_normal(>migration_blocker, errp);
if (r < 0) {
goto fail_busyloop;
}
--
1.8.3.1
[PATCH V2 4/6] cpr: relax vhost migration blockers
vhost blocks migration if logging is not supported to track dirty
memory, and vhost-user blocks it if the log cannot be saved to a shm fd.
vhost-vdpa blocks migration if both hosts do not support all the device's
features using a shadow VQ, for tracking requests and dirty memory.
vhost-scsi blocks migration if storage cannot be shared across hosts,
or if state cannot be migrated.
None of these conditions apply if the old and new qemu processes do
not run concurrently, and if new qemu starts on the same host as old,
which is the case for cpr.
Narrow the scope of these blockers so they only apply to normal mode.
They will not block cpr modes when they are added in subsequent patches.
No functional change until a new mode is added.
Signed-off-by: Steve Sistare
Reviewed-by: Juan Quintela
---
hw/scsi/vhost-scsi.c | 2 +-
hw/virtio/vhost.c| 2 +-
2 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/hw/scsi/vhost-scsi.c b/hw/scsi/vhost-scsi.c
index 14e23cc..bf528d5 100644
--- a/hw/scsi/vhost-scsi.c
+++ b/hw/scsi/vhost-scsi.c
@@ -208,7 +208,7 @@ static void vhost_scsi_realize(DeviceState *dev, Error
**errp)
"When external environment supports it (Orchestrator migrates "
"target SCSI device state or use shared storage over network),
"
"set 'migratable' property to true to enable migration.");
-if (migrate_add_blocker(>migration_blocker, errp) < 0) {
+if (migrate_add_blocker_normal(>migration_blocker, errp) < 0) {
goto free_virtio;
}
}
diff --git a/hw/virtio/vhost.c b/hw/virtio/vhost.c
index d737671..f5e9625 100644
--- a/hw/virtio/vhost.c
+++ b/hw/virtio/vhost.c
@@ -1527,7 +1527,7 @@ int vhost_dev_init(struct vhost_dev *hdev, void *opaque,
}
if (hdev->migration_blocker != NULL) {
-r = migrate_add_blocker(>migration_blocker, errp);
+r = migrate_add_blocker_normal(>migration_blocker, errp);
if (r < 0) {
goto fail_busyloop;
}
--
1.8.3.1
