spapr_vscsi: do not allow device hotplug
PATCH 1/1 spapr_vscsi: do not allow device hotplug
We do not implement hotplug in the vscsi bus, but we forgot to tell qdev about it. The result is that users are able to hotplug devices in the vscsi bus, the devices appear in qdev, but they aren't usable by the guest OS unless the user reboots it first.
Setting qbus hotplug_handler to NULL will tell qdev-monitor, via qbus_is_hotpluggable(), that we do not support hotplug operations in spapr_vscsi.
--- hw/scsi/spapr_vscsi.c | 3 +++ 1 file changed, 3 insertions(+)
hw/scsi/spapr_vscsi.c +3 -0
@@ -1219,6 +1219,9 @@ static void spapr_vscsi_realize(SpaprVioDevice *dev, Error **errp)scsi_bus_new (&s->bus,sizeof (s->bus),DEVICE (dev), &vscsi_scsi_info,NULL ); + +/* ibmvscsi SCSI bus does not allow hotplug. */ +qbus_set_hotplug_handler (BUS (&s->bus),NULL ); }void spapr_vscsi_create (SpaprVioBus *bus)
Signature · 1 line
2.26.2
David Gibson
AppliedQuoted 5 lines · Daniel Henrique Barboza
We do not implement hotplug in the vscsi bus, but we forgot to tell qdev about it. The result is that users are able to hotplug devices in the vscsi bus, the devices appear in qdev, but they aren't usable by the guest OS unless the user reboots it first.
Quoted 7 lines
Setting qbus
hotplug_handlerto NULL will tell qdev-monitor, viaqbus_is_hotpluggable(), that we do not support hotplug operations inspapr_vscsi.Signed-off-by Daniel Henrique Barboza <[email protected]>
That was much easier than I feared. Nice!
Applied to ppc-for-5.2.
Quoted 18 lines
--- hw/scsi/spapr_vscsi.c | 3 +++ 1 file changed, 3 insertions(+)hw/scsi/spapr_vscsi.c +3
-0@@ -1219,6 +1219,9 @@ static void spapr_vscsi_realize(SpaprVioDevice *dev, Error **errp)scsi_bus_new (&s->bus,sizeof (s->bus),DEVICE (dev), &vscsi_scsi_info,NULL ); + +/* ibmvscsi SCSI bus does not allow hotplug. */ +qbus_set_hotplug_handler (BUS (&s->bus),NULL ); }void spapr_vscsi_create (SpaprVioBus *bus)
Signature · 4 lines
David Gibson | I'll have my music baroque, and my code david AT gibson.dropbear.id.au | minimalist, thank you. NOT _the_ _other_ | _way_ _around_! http://www.ozlabs.org/~dgibson
Cc'ing Markus
Quoted 21 lines · Daniel Henrique Barboza
We do not implement hotplug in the vscsi bus, but we forgot to tell qdev about it. The result is that users are able to hotplug devices in the vscsi bus, the devices appear in qdev, but they aren't usable by the guest OS unless the user reboots it first.
Setting qbus
hotplug_handlerto NULL will tell qdev-monitor, viaqbus_is_hotpluggable(), that we do not support hotplug operations inspapr_vscsi.Signed-off-by Daniel Henrique Barboza <[email protected]>--- hw/scsi/spapr_vscsi.c | 3 +++ 1 file changed, 3 insertions(+)hw/scsi/spapr_vscsi.c +0
-0
hw/scsi/spapr_vscsi.c static void spapr_vscsi_realize(SpaprVioDevice *dev, Error **errp) · line 1224
Quoted 6 lines
hw/scsi/spapr_vscsi.c +3
-0@@ -1219,6 +1219,9 @@ static void spapr_vscsi_realize(SpaprVioDevice *dev, Error **errp)scsi_bus_new (&s->bus,sizeof (s->bus),DEVICE (dev), &vscsi_scsi_info,NULL ); + +/* ibmvscsi SCSI bus does not allow hotplug. */ +qbus_set_hotplug_handler (BUS (&s->bus),NULL );
Can't this be a problem later in DeviceClass::unrealize()?
I was expecting something like, overwriting the parent bus type:
-- >8 --
@@ -1271,6 +1271,7 @@ static void spapr_vscsi_class_init(ObjectClass *klass, void *data)
DeviceClass *dc = DEVICE_CLASS(klass);
SpaprVioDeviceClass *k = VIO_SPAPR_DEVICE_CLASS(klass);+ k->bus_type = NULL; /* ibmvscsi SCSI bus does not allow hotplug. */
k->realize = spapr_vscsi_realize;
k->reset = spapr_vscsi_reset;
k->devnode = spapr_vscsi_devnode;---
Quoted 4 lines
hw/scsi/spapr_vscsi.c +0
-0}void spapr_vscsi_create (SpaprVioBus *bus)
Quoted 25 lines · Philippe Mathieu-Daudé
Cc'ing Markus
Quoted 22 lines · Daniel Henrique Barboza
We do not implement hotplug in the vscsi bus, but we forgot to tell qdev about it. The result is that users are able to hotplug devices in the vscsi bus, the devices appear in qdev, but they aren't usable by the guest OS unless the user reboots it first.
Setting qbus
hotplug_handlerto NULL will tell qdev-monitor, viaqbus_is_hotpluggable(), that we do not support hotplug operations inspapr_vscsi.Signed-off-by Daniel Henrique Barboza <[email protected]>--- hw/scsi/spapr_vscsi.c | 3 +++ 1 file changed, 3 insertions(+)hw/scsi/spapr_vscsi.c +0
-0@@ -1219,6 +1219,9 @@ static void spapr_vscsi_realize(SpaprVioDevice *dev, Error **errp)
Quoted 7 lines
hw/scsi/spapr_vscsi.c static void spapr_vscsi_realize(SpaprVioDevice *dev, Error **errp) · line 1224
Quoted 5 lines
hw/scsi/spapr_vscsi.c +3
-0scsi_bus_new (&s->bus,sizeof (s->bus),DEVICE (dev), &vscsi_scsi_info,NULL ); + +/* ibmvscsi SCSI bus does not allow hotplug. */ +qbus_set_hotplug_handler (BUS (&s->bus),NULL );Can't this be a problem later in
DeviceClass::unrealize()?
Not as far as I've tested. A call to qbus_set_hotplug_handler(bus,NULL) after setting it to NULL isn't breaking anything either (just tested).
I verified before sending the patch that setting hotplug_handler to NULL is done in some unrealize() calls in buses, but not on devices. And I'm not sure which instance would cause an unrealize() in the device to fail if the hotplug_handler of the bus is NULL. As far as I'm concerned this shouldn't be happening in our case here, where we're not dealing with hotplug devices in the bus at all.
Which potential problems are you referring to?
Quoted 13 lines
I was expecting something like, overwriting the parent bus type:
-- >8 --
(quoted diff) +1
-0@@ -1271,6 +1271,7 @@ static void spapr_vscsi_class_init(ObjectClass *klass, void *data) DeviceClass *dc = DEVICE_CLASS(klass); SpaprVioDeviceClass *k = VIO_SPAPR_DEVICE_CLASS(klass); + k->bus_type = NULL; /* ibmvscsi SCSI bus does not allow hotplug. */ k->realize = spapr_vscsi_realize; k->reset = spapr_vscsi_reset;k->devnode = spapr_vscsi_devnode;
Quoted 1 line
---
spapr_vscsi is not a bus, is an interface. Setting NULL to bus_type in spapr_vio breaks guest init:
qemu-system-ppc64: /home/danielhb/qemu/hw/core/qdev.c:102: qdev_set_parent_bus: Assertion `dc->bus_type && object_dynamic_cast(OBJECT(bus), dc->bus_type)' failed. Aborted
I'm not so sure this would be better than what I'm doing either. qdev_device_add() calls qbus_is_hotpluggable() to see if the chosen bus allows hotplug. This function verifies if bus->hotplug_handler is NULL. What I'm doing is simply setting hotplug_handler to NULL in the SCSI bus instance that belongs to spapr_vscsi. As far as I understand this is a valid use of the qdev API - I should be able to set hotplug_handler to NULL if I don't want devices being hotplugged in the bus I'm instantiating. Either that, or qbus_is_hotpluggable() must check for something else that I can safely turn off.
Thanks,
DHB
Quoted 6 lines
Quoted 5 lines
hw/scsi/spapr_vscsi.c +0
-0}void spapr_vscsi_create (SpaprVioBus *bus)
Quoted 39 lines · Daniel Henrique Barboza
Quoted 19 lines · Philippe Mathieu-Daudé
Cc'ing Markus
Quoted 16 lines · Daniel Henrique Barboza
We do not implement hotplug in the vscsi bus, but we forgot to tell qdev about it. The result is that users are able to hotplug devices in the vscsi bus, the devices appear in qdev, but they aren't usable by the guest OS unless the user reboots it first.
Setting qbus
hotplug_handlerto NULL will tell qdev-monitor, viaqbus_is_hotpluggable(), that we do not support hotplug operations inspapr_vscsi.Signed-off-by Daniel Henrique Barboza <[email protected]>---
hw/scsi/spapr_vscsi.c| 3 +++1 file changed, 3 insertions(+)Quoted 13 lines
hw/scsi/spapr_vscsi.c static void spapr_vscsi_realize(SpaprVioDevice · line 1224
Quoted 11 lines
hw/scsi/spapr_vscsi.c +3
-0@@ -1219,6 +1219,9 @@ static void spapr_vscsi_realize(SpaprVioDevice *dev, Error **errp) scsi_bus_new(&s->bus, sizeof(s->bus), DEVICE(dev), &vscsi_scsi_info, NULL); + + /* ibmvscsi SCSI bus does not allow hotplug. */ + qbus_set_hotplug_handler(BUS(&s->bus), NULL);Can't this be a problem later in
DeviceClass::unrealize()?Not as far as I've tested. A call to
qbus_set_hotplug_handler(bus,NULL)after setting it to NULL isn't breaking anything either (just tested).
Quoted 8 lines
I verified before sending the patch that setting
hotplug_handlerto NULL is done in someunrealize()calls in buses, but not on devices. And I'm not sure which instance would cause anunrealize()in the device to fail if thehotplug_handlerof the bus is NULL. As far as I'm concerned this shouldn't be happening in our case here, where we're not dealing with hotplug devices in the bus at all.Which potential problems are you referring to?
I was wondering if qdev_unrealize() runs similar checks than qdev_realize().
Quoted 27 lines
Quoted 13 lines
I was expecting something like, overwriting the parent bus type:
-- >8 --
(quoted diff) +1
-0@@ -1271,6 +1271,7 @@ static void spapr_vscsi_class_init(ObjectClass *klass, void *data) DeviceClass *dc = DEVICE_CLASS(klass); SpaprVioDeviceClass *k = VIO_SPAPR_DEVICE_CLASS(klass); + k->bus_type = NULL; /* ibmvscsi SCSI bus does not allow hotplug. */ k->realize = spapr_vscsi_realize; k->reset = spapr_vscsi_reset;Quoted 2 lines
k->devnode=spapr_vscsi_devnode;
---
spapr_vscsiis not a bus, is an interface. Setting NULL tobus_typeinspapr_vio
breaks guest init:qemu-system-ppc64:
/home/danielhb/qemu/hw/core/qdev.c:102:qdev_set_parent_bus: Assertion `dc->bus_type &&object_dynamic_cast(OBJECT(bus), dc->bus_type)' failed. Aborted
Quoted 10 lines
I'm not so sure this would be better than what I'm doing either.
qdev_device_add()
callsqbus_is_hotpluggable()to see if the chosen bus allows hotplug. This function verifies ifbus->hotplug_handleris NULL. What I'm doing is simply settinghotplug_handlerto NULL in the SCSI bus instance that belongs tospapr_vscsi. As far as I understand this is a valid use of the qdev API - I should be able to sethotplug_handlerto NULL if I don't want devices being hotplugged in the bus I'm instantiating. Either that, orqbus_is_hotpluggable()
must check for something else that I can safely turn off.
I'm not saying I know the correct way to do that, this is why I added Markus in the loop :) I'm wondering from an API point of view what is the best way to achieve what you want. I'm glad this works this way.
Quoted 16 lines
Thanks,
DHB
Quoted 5 lines
Quoted 4 lines
hw/scsi/spapr_vscsi.c +0
-0} void spapr_vscsi_create(SpaprVioBus *bus)
Quoted 47 lines · Philippe Mathieu-Daudé
Quoted 46 lines · Daniel Henrique Barboza
Quoted 19 lines · Philippe Mathieu-Daudé
Cc'ing Markus
Quoted 16 lines · Daniel Henrique Barboza
We do not implement hotplug in the vscsi bus, but we forgot to tell qdev about it. The result is that users are able to hotplug devices in the vscsi bus, the devices appear in qdev, but they aren't usable by the guest OS unless the user reboots it first.
Setting qbus
hotplug_handlerto NULL will tell qdev-monitor, viaqbus_is_hotpluggable(), that we do not support hotplug operations inspapr_vscsi.Signed-off-by Daniel Henrique Barboza <[email protected]>--- hw/scsi/spapr_vscsi.c | 3 +++ 1 file changed, 3 insertions(+)Quoted 13 lines
hw/scsi/spapr_vscsi.c static void spapr_vscsi_realize(SpaprVioDevice · line 1224
Quoted 11 lines
hw/scsi/spapr_vscsi.c +3
-0@@ -1219,6 +1219,9 @@ static void spapr_vscsi_realize(SpaprVioDevice *dev, Error **errp) scsi_bus_new(&s->bus, sizeof(s->bus), DEVICE(dev), &vscsi_scsi_info, NULL); + + /* ibmvscsi SCSI bus does not allow hotplug. */ + qbus_set_hotplug_handler(BUS(&s->bus), NULL);Can't this be a problem later in
DeviceClass::unrealize()?Not as far as I've tested. A call to
qbus_set_hotplug_handler(bus,NULL)after setting it to NULL isn't breaking anything either (just tested).I verified before sending the patch that setting
hotplug_handlerto NULL is done in someunrealize()calls in buses, but not on devices. And I'm not sure which instance would cause anunrealize()in the device to fail if thehotplug_handlerof the bus is NULL. As far as I'm concerned this shouldn't be happening in our case here, where we're not dealing with hotplug devices in the bus at all.
Quoted 4 lines
Quoted 1 line
Which potential problems are you referring to?
I was wondering if
qdev_unrealize()runs similar checks thanqdev_realize().
I have no idea. I would need to check/study the code to answer that.
Quoted 39 lines
Quoted 28 lines
Quoted 14 lines
I was expecting something like, overwriting the parent bus type:
-- >8 --
(quoted diff) +1
-0@@ -1271,6 +1271,7 @@ static void spapr_vscsi_class_init(ObjectClass *klass, void *data) DeviceClass *dc = DEVICE_CLASS(klass); SpaprVioDeviceClass *k = VIO_SPAPR_DEVICE_CLASS(klass); + k->bus_type = NULL; /* ibmvscsi SCSI bus does not allow hotplug. */ k->realize = spapr_vscsi_realize; k->reset = spapr_vscsi_reset;Quoted 2 lines
k->devnode=spapr_vscsi_devnode;
---
spapr_vscsiis not a bus, is an interface. Setting NULL tobus_typeinspapr_vio
breaks guest init:qemu-system-ppc64:
/home/danielhb/qemu/hw/core/qdev.c:102:qdev_set_parent_bus: Assertion `dc->bus_type &&object_dynamic_cast(OBJECT(bus), dc->bus_type)' failed. AbortedQuoted 10 lines
I'm not so sure this would be better than what I'm doing either.
qdev_device_add()
callsqbus_is_hotpluggable()to see if the chosen bus allows hotplug. This function verifies ifbus->hotplug_handleris NULL. What I'm doing is simply settinghotplug_handlerto NULL in the SCSI bus instance that belongs tospapr_vscsi. As far as I understand this is a valid use of the qdev API - I should be able to sethotplug_handlerto NULL if I don't want devices being hotplugged in the bus I'm instantiating. Either that, orqbus_is_hotpluggable()
must check for something else that I can safely turn off.
Quoted 3 lines
I'm not saying I know the correct way to do that, this is why I added Markus in the loop :) I'm wondering from an API point of view what is the best way to achieve what you want. I'm glad this works this way.
It's alright to question hehehe far too often we're remembered that 'works this way' is a looong way from 'this is the right way'. Let's see if Markus agrees with this approach or if there is a better way of doing it. I'll happily change for a safer approach that will not hit me in the back in the future.
Thanks,
DHB
Quoted 18 lines
Quoted 17 lines
Thanks,
DHB
Quoted 5 lines
Quoted 4 lines
hw/scsi/spapr_vscsi.c +0
-0} void spapr_vscsi_create(SpaprVioBus *bus)
Quoted 25 lines · Philippe Mathieu-Daudé
Cc'ing Markus
Quoted 22 lines · Daniel Henrique Barboza
We do not implement hotplug in the vscsi bus, but we forgot to tell qdev about it. The result is that users are able to hotplug devices in the vscsi bus, the devices appear in qdev, but they aren't usable by the guest OS unless the user reboots it first.
Setting qbus
hotplug_handlerto NULL will tell qdev-monitor, viaqbus_is_hotpluggable(), that we do not support hotplug operations inspapr_vscsi.Signed-off-by Daniel Henrique Barboza <[email protected]>--- hw/scsi/spapr_vscsi.c | 3 +++ 1 file changed, 3 insertions(+)hw/scsi/spapr_vscsi.c +0
-0@@ -1219,6 +1219,9 @@ static void spapr_vscsi_realize(SpaprVioDevice *dev, Error **errp)
Quoted 7 lines
hw/scsi/spapr_vscsi.c static void spapr_vscsi_realize(SpaprVioDevice *dev, Error **errp) · line 1224
Quoted 5 lines
hw/scsi/spapr_vscsi.c +3
-0scsi_bus_new (&s->bus,sizeof (s->bus),DEVICE (dev), &vscsi_scsi_info,NULL ); + +/* ibmvscsi SCSI bus does not allow hotplug. */ +qbus_set_hotplug_handler (BUS (&s->bus),NULL );Can't this be a problem later in
DeviceClass::unrealize()?
Can't say offhand. Cc'ing QOM maintainers for actual expertise.
A quick grep coughs up virtio_serial_device_realize() / virtio_serial_device_unrealize(), which set and unset the hotplug handler of a bus their device provides. Not the same as the code above, just a clue that messing with a bus's hotplug handler in a realize method can work.
qbus_set_hotplug_handler() is entirely undocumented. It sets a link. Might be intended for use by the device that provides the bus.
Quoted 11 lines
I was expecting something like, overwriting the parent bus type:
-- >8 --
(quoted diff) +1
-0@@ -1271,6 +1271,7 @@ static void spapr_vscsi_class_init(ObjectClass *klass, void *data) DeviceClass *dc = DEVICE_CLASS(klass); SpaprVioDeviceClass *k = VIO_SPAPR_DEVICE_CLASS(klass); + k->bus_type = NULL; /* ibmvscsi SCSI bus does not allow hotplug. */ k->realize = spapr_vscsi_realize; k->reset = spapr_vscsi_reset;
Quoted 1 line
k->devnode = spapr_vscsi_devnode;
k->bus_type does not exist. Do you mean dc->bus_type?
TYPE_VIO_SPAPR_VSCSI_DEVICE is a subtype of TYPE_VIO_SPAPR_DEVICE is a subtype of TYPE_DEVICE.
TYPE_DEVICE is bus-less, i.e. (concrete) instances do not plug into a bus.
TYPE_VIO_SPAPR_DEVICE has bus_type TYPE_SPAPR_VIO_BUS, i.e. (concrete) instances plug into a TYPE_SPAPR_VIO_BUS. Both hot and cold plug.
Zapping TYPE_VIO_SPAPR_VSCSI_DEVICE's bus_type makes it bus-less. I doubt that's okay. Also, it's not just about hot plug. What are you trying to accomplish?
Quoted 35 lines · Markus Armbruster
Quoted 25 lines · Philippe Mathieu-Daudé
Cc'ing Markus
Quoted 22 lines · Daniel Henrique Barboza
We do not implement hotplug in the vscsi bus, but we forgot to tell qdev about it. The result is that users are able to hotplug devices in the vscsi bus, the devices appear in qdev, but they aren't usable by the guest OS unless the user reboots it first.
Setting qbus
hotplug_handlerto NULL will tell qdev-monitor, viaqbus_is_hotpluggable(), that we do not support hotplug operations inspapr_vscsi.Signed-off-by Daniel Henrique Barboza <[email protected]>--- hw/scsi/spapr_vscsi.c | 3 +++ 1 file changed, 3 insertions(+)hw/scsi/spapr_vscsi.c +0
-0@@ -1219,6 +1219,9 @@ static void spapr_vscsi_realize(SpaprVioDevice *dev, Error **errp)Quoted 7 lines
hw/scsi/spapr_vscsi.c static void spapr_vscsi_realize(SpaprVioDevice *dev, Error **errp) · line 1224
Quoted 5 lines
hw/scsi/spapr_vscsi.c +3
-0scsi_bus_new (&s->bus,sizeof (s->bus),DEVICE (dev), &vscsi_scsi_info,NULL ); + +/* ibmvscsi SCSI bus does not allow hotplug. */ +qbus_set_hotplug_handler (BUS (&s->bus),NULL );Can't this be a problem later in
DeviceClass::unrealize()?
Quoted 5 lines
Can't say offhand. Cc'ing QOM maintainers for actual expertise.
A quick grep coughs up
virtio_serial_device_realize()/virtio_serial_device_unrealize(), which set and unset the hotplug handler of a bus their device provides. Not the same as the code above,
Hm, it almost is, actually. Only difference is that virtio serial is setting the hotplug handler to something, rather than disabling it.
Quoted 5 lines
just a clue that messing with a bus's hotplug handler in a realize method can work.
qbus_set_hotplug_handler()is entirely undocumented. It sets a link. Might be intended for use by the device that provides the bus.
Heh.
Quoted 8 lines
Quoted 8 lines
I was expecting something like, overwriting the parent bus type:
-- >8 --
(quoted diff) +0
-0@@ -1271,6 +1271,7 @@ static void spapr_vscsi_class_init(ObjectClass *klass, void *data) DeviceClass *dc = DEVICE_CLASS(klass); SpaprVioDeviceClass *k = VIO_SPAPR_DEVICE_CLASS(klass);
Quoted 6 lines
Quoted 4 lines
(quoted diff) +1
-0+ k->bus_type = NULL; /* ibmvscsi SCSI bus does not allow hotplug. */ k->realize = spapr_vscsi_realize; k->reset = spapr_vscsi_reset;k->devnode = spapr_vscsi_devnode;
k->bus_typedoes not exist. Do you meandc->bus_type?
Yeah, I'm pretty sure changing the bus type is going to be worse than setting the hotplug handler.
Quoted 6 lines
TYPE_VIO_SPAPR_VSCSI_DEVICEis a subtype ofTYPE_VIO_SPAPR_DEVICEis a subtype ofTYPE_DEVICE.
TYPE_DEVICEis bus-less, i.e. (concrete) instances do not plug into a bus.
Quoted 6 lines
TYPE_VIO_SPAPR_DEVICEhasbus_typeTYPE_SPAPR_VIO_BUS, i.e. (concrete) instances plug into aTYPE_SPAPR_VIO_BUS. Both hot and cold plug.Zapping TYPE_VIO_SPAPR_VSCSI_DEVICE's
bus_typemakes it bus-less. I doubt that's okay. Also, it's not just about hot plug. What are you trying to accomplish?
I'm pretty sure all of the above is not relevant. It's the subordinate bus of TYPE_VIO_SPAPR_VSCSI_DEVICE on which we want to disable hotplugging, not the parent bus. We can't hotplug spapr-vscsi controllers, but that's fine. What we're trying to address here is that the controller supports a TYPE_SCSI_BUS, onto which disks can normally be hotplugged, but not in this case.
Signature · 4 lines
David Gibson | I'll have my music baroque, and my code david AT gibson.dropbear.id.au | minimalist, thank you. NOT _the_ _other_ | _way_ _around_! http://www.ozlabs.org/~dgibson