allow to deprecate objects and devices
v2 cover allow to deprecate objects and devices
Put some infrastructure in place to allow tagging objects (including devices) as deprected. Use it to mark the ohci pci host adapter and the usb hub as deprecated.
v2:
- pick up reviews.
- drop ohci patch.
- add cirrus vga patch.
Gerd Hoffmann (4):
qom: allow to mark objects (including devices) as deprecated.
usb: add config options for the hub and hid devices usb/hub: deprecate, don't build by default vga/cirrus: deprecate, don't build by default
include/qom/object.h | 1 + hw/display/cirrus_vga.c | 1 + hw/display/cirrus_vga_isa.c | 1 + hw/usb/dev-hub.c | 1 + qom/qom-qmp-cmds.c | 4 ++++ system/qdev-monitor.c | 5 +++++ hw/display/Kconfig | 1 - hw/usb/Kconfig | 9 +++++++++ hw/usb/meson.build | 4 ++-- qapi/qom.json | 4 +++- 10 files changed, 27 insertions(+), 4 deletions(-)
Signature · 1 line
2.45.1
PATCH v2 1/4 qom: allow to mark objects (including devices) as deprecated.
Add deprecation_note field (string) to ObjectClass.
Add deprecated bool to ObjectTypeInfo, report in 'qom-list-types'. Print the note when listing devices via '-device help'.
--- include/qom/object.h | 1 + qom/qom-qmp-cmds.c | 4 ++++ system/qdev-monitor.c | 5 +++++ qapi/qom.json | 4 +++- 4 files changed, 13 insertions(+), 1 deletion(-)
include/qom/object.h +1 -0
@@ -136,6 +136,7 @@ struct ObjectClassObjectUnparent *unparent;GHashTable *properties; +const char *deprecation_note; };/**
qom/qom-qmp-cmds.c +4 -0
@@ -101,6 +101,10 @@ static void qom_list_types_tramp(ObjectClass *klass, void *data)if (parent) { info->parent =g_strdup (object_class_get_name (parent)); } +if (klass->deprecation_note) { + info->has_deprecated = true; + info->deprecated = true; + }QAPI_LIST_PREPEND (*pret, info); }
system/qdev-monitor.c +5 -0
@@ -144,6 +144,8 @@ static bool qdev_class_has_alias(DeviceClass *dc)static void qdev_print_devinfo (DeviceClass *dc) { +ObjectClass *klass =OBJECT_CLASS (dc); +qemu_printf ("name \"%s\"" ,object_class_get_name (OBJECT_CLASS (dc)));if (dc->bus_type) {qemu_printf (", bus %s" , dc->bus_type);
@@ -157,6 +159,9 @@ static void qdev_print_devinfo(DeviceClass *dc)if (!dc->user_creatable) {qemu_printf (", no-user" ); } +if (klass->deprecation_note) { +qemu_printf (", deprecated \"%s\"" , klass->deprecation_note); + }qemu_printf ("\n" ); }
qapi/qom.json +3 -1
@@ -163,10 +163,12 @@ # # @parent: Name of parent type, if any (since 2.10) # +# @deprecated: the type is deprecated (since 9.1) +# # Since: 1.1 ## { 'struct': 'ObjectTypeInfo', - 'data': { 'name': 'str', '*abstract': 'bool', '*parent': 'str' } } + 'data': { 'name': 'str', '*abstract': 'bool', '*parent': 'str', '*deprecated': 'bool' } } ## # @qom-list-types:
Signature · 1 line
2.45.1
Re: [PATCH v2 1/4] qom: allow to mark objects (including devices) as deprecated.
Quoted 3 lines · Gerd Hoffmann
Add
deprecation_notefield (string) to ObjectClass.
Add deprecated bool to ObjectTypeInfo, report in 'qom-list-types'. Print the note when listing devices via '-device help'.
In the subject line, I suggest s/allow to mark/allow marking/
Quoted 9 lines
Signed-off-by Gerd Hoffmann <[email protected]>--- include/qom/object.h | 1 + qom/qom-qmp-cmds.c | 4 ++++ system/qdev-monitor.c | 5 +++++ qapi/qom.json | 4 +++- 4 files changed, 13 insertions(+), 1 deletion(-)
Signature · 3 lines
Eric Blake, Principal Software Engineer Red Hat, Inc. Virtualization: qemu.org | libguestfs.org
Re: [PATCH v2 1/4] qom: allow to mark objects (including devices) as deprecated.
In reply to Gerd HoffmannQuoted 11 lines · Gerd Hoffmann
Add
deprecation_notefield (string) to ObjectClass.
Add deprecated bool to ObjectTypeInfo, report in 'qom-list-types'. Print the note when listing devices via '-device help'.Signed-off-by Gerd Hoffmann <[email protected]>--- include/qom/object.h | 1 + qom/qom-qmp-cmds.c | 4 ++++ system/qdev-monitor.c | 5 +++++ qapi/qom.json | 4 +++- 4 files changed, 13 insertions(+), 1 deletion(-)
I replied to v1 before noticing your v2, so see my comment there about printing a warning on stderr when creating an instance.
With regards,
Daniel
Signature · 3 lines
|: https://berrange.com -o- https://www.flickr.com/photos/dberrange :| |: https://libvirt.org -o- https://fstop138.berrange.com :| |: https://entangle-photo.org -o- https://www.instagram.com/dberrange :|
PATCH v2 2/4 usb: add config options for the hub and hid devices
In reply to Gerd Hoffmann--- hw/usb/Kconfig | 10 ++++++++++ hw/usb/meson.build | 4 ++-- 2 files changed, 12 insertions(+), 2 deletions(-)
hw/usb/Kconfig +10 -0
@@ -65,6 +65,16 @@ config TUSB6010 bool select USB_MUSB +config USB_HUB + bool + default y + depends on USB + +config USB_HID + bool + default y + depends on USB + config USB_TABLET_WACOM bool default y
hw/usb/meson.build +2 -2
@@ -35,8 +35,8 @@ system_ss.add(when: 'CONFIG_XLNX_VERSAL', if_true: files('xlnx-versal-usb2-ctrl- system_ss.add(when: 'CONFIG_XLNX_USB_SUBSYS', if_true: files('xlnx-usb-subsystem.c')) # emulated usb devices -system_ss.add(when: 'CONFIG_USB', if_true: files('dev-hub.c')) -system_ss.add(when: 'CONFIG_USB', if_true: files('dev-hid.c')) +system_ss.add(when: 'CONFIG_USB_HUB', if_true: files('dev-hub.c')) +system_ss.add(when: 'CONFIG_USB_HID', if_true: files('dev-hid.c')) system_ss.add(when: 'CONFIG_USB_TABLET_WACOM', if_true: files('dev-wacom.c')) system_ss.add(when: 'CONFIG_USB_STORAGE_CORE', if_true: files('dev-storage.c')) system_ss.add(when: 'CONFIG_USB_STORAGE_BOT', if_true: files('dev-storage-bot.c'))
Signature · 1 line
2.45.1
Re: [PATCH v2 2/4] usb: add config options for the hub and hid devices
Quoted 6 lines · Gerd Hoffmann
Signed-off-by Gerd Hoffmann <[email protected]>Reviewed-by Thomas Huth <[email protected]>--- hw/usb/Kconfig | 10 ++++++++++ hw/usb/meson.build | 4 ++-- 2 files changed, 12 insertions(+), 2 deletions(-)
PATCH v2 3/4 usb/hub: deprecate, don't build by default
In reply to Gerd HoffmannThe hub supports only USB 1.1. When running out of usb ports it is in almost all cases the much better choice to add another usb host adapter (or increase the number of root ports when using xhci) instead of using the usb hub.
--- hw/usb/dev-hub.c | 1 + hw/usb/Kconfig | 1 - 2 files changed, 1 insertion(+), 1 deletion(-)
hw/usb/dev-hub.c +1 -0
@@ -686,6 +686,7 @@ static void usb_hub_class_initfn(ObjectClass *klass, void *data)set_bit (DEVICE_CATEGORY_BRIDGE , dc->categories); dc->fw_name ="hub" ; dc->vmsd = &vmstate_usb_hub; + klass->deprecation_note ="use more root ports or additional hostadapters instead" ;device_class_set_props (dc, usb_hub_properties); }
hw/usb/Kconfig +0 -1
@@ -67,7 +67,6 @@ config TUSB6010 config USB_HUB bool - default y depends on USB config USB_HID
Signature · 1 line
2.45.1
PATCH v2 4/4 vga/cirrus: deprecate, don't build by default
In reply to Gerd Hoffmannstdvga is the much better option.
--- hw/display/cirrus_vga.c | 1 + hw/display/cirrus_vga_isa.c | 1 + hw/display/Kconfig | 1 - 3 files changed, 2 insertions(+), 1 deletion(-)
hw/display/cirrus_vga.c +1 -0
@@ -3007,6 +3007,7 @@ static void cirrus_vga_class_init(ObjectClass *klass, void *data) dc->vmsd = &vmstate_pci_cirrus_vga;device_class_set_props (dc, pci_vga_cirrus_properties); dc->hotpluggable = false; + klass->deprecation_note ="use stdvga instead" ; }static const TypeInfo cirrus_vga_info = {
hw/display/cirrus_vga_isa.c +1 -0
@@ -85,6 +85,7 @@ static void isa_cirrus_vga_class_init(ObjectClass *klass, void *data) dc->realize = isa_cirrus_vga_realizefn;device_class_set_props (dc, isa_cirrus_vga_properties);set_bit (DEVICE_CATEGORY_DISPLAY , dc->categories); + klass->deprecation_note ="use stdvga instead" ; }static const TypeInfo isa_cirrus_vga_info = {
hw/display/Kconfig +0 -1
@@ -11,7 +11,6 @@ config FW_CFG_DMA config VGA_CIRRUS bool - default y if PCI_DEVICES depends on PCI select VGA
Signature · 1 line
2.45.1
Re: [PATCH v2 4/4] vga/cirrus: deprecate, don't build by default
Quoted 26 lines · Gerd Hoffmann
stdvga is the much better option.
Signed-off-by Gerd Hoffmann <[email protected]>---
hw/display/cirrus_vga.c | 1 + hw/display/cirrus_vga_isa.c | 1 + hw/display/Kconfig | 1 - 3 files changed, 2 insertions(+), 1 deletion(-)hw/display/cirrus_vga.c +1
-0@@ -3007,6 +3007,7 @@ static void cirrus_vga_class_init(ObjectClass *klass, void *data) dc->vmsd = &vmstate_pci_cirrus_vga;device_class_set_props (dc, pci_vga_cirrus_properties); dc->hotpluggable = false; + klass->deprecation_note ="use stdvga instead" ; }static const TypeInfo cirrus_vga_info = {hw/display/cirrus_vga_isa.c +0
-0
hw/display/cirrus_vga_isa.c static void isa_cirrus_vga_class_init(ObjectClass *klass, void *data) · line 88
Quoted 4 lines
hw/display/cirrus_vga_isa.c +1
-0@@ -85,6 +85,7 @@ static void isa_cirrus_vga_class_init(ObjectClass *klass, void *data) dc->realize = isa_cirrus_vga_realizefn;device_class_set_props (dc, isa_cirrus_vga_properties);set_bit (DEVICE_CATEGORY_DISPLAY , dc->categories); + klass->deprecation_note ="use stdvga instead" ;
Excepr some old OSes work better with this than stdvga so could this be left and not removed? Does it cause a lot of work to keep this device? I thought it's stable already and were not many changes for it lately. If something works why drop it?
Regards,
BALATON Zoltan
Quoted 16 lines
hw/display/cirrus_vga_isa.c +0
-0}static const TypeInfo isa_cirrus_vga_info = {hw/display/Kconfig +0
-0@@ -11,7 +11,6 @@ config FW_CFG_DMAconfig
VGA_CIRRUS
bool
- default y ifPCI_DEVICES
depends on PCI
select VGA
Re: [PATCH v2 4/4] vga/cirrus: deprecate, don't build by default
Quoted 22 lines · BALATON Zoltan
Quoted 21 lines · Gerd Hoffmann
stdvga is the much better option.
Signed-off-by Gerd Hoffmann <[email protected]>---
hw/display/cirrus_vga.c | 1 +hw/display/cirrus_vga_isa.c| 1 + hw/display/Kconfig | 1 -3 files changed, 2 insertions(+), 1 deletion(-)hw/display/cirrus_vga.c +1
-0@@ -3007,6 +3007,7 @@ static void cirrus_vga_class_init(ObjectClass *klass, void *data) dc->vmsd = &vmstate_pci_cirrus_vga; device_class_set_props(dc, pci_vga_cirrus_properties); dc->hotpluggable = false; + klass->deprecation_note = "use stdvga instead"; }
Quoted 13 lines
hw/display/cirrus_vga_isa.c static void isa_cirrus_vga_class_init(ObjectClass *klass, void · line 89
Quoted 11 lines
static const TypeInfo
cirrus_vga_info= {hw/display/cirrus_vga_isa.c +1
-0@@ -85,6 +85,7 @@ static void isa_cirrus_vga_class_init(ObjectClass *klass, void *data) dc->realize = isa_cirrus_vga_realizefn; device_class_set_props(dc, isa_cirrus_vga_properties); set_bit(DEVICE_CATEGORY_DISPLAY, dc->categories); + klass->deprecation_note = "use stdvga instead";Excepr some old OSes work better with this than stdvga so could this be left and not removed? Does it cause a lot of work to keep this device? I thought it's stable already and were not many changes for it lately. If something works why drop it?
Seconded: whilst stdvga is preferred, there are a lot of older OSs that work well in QEMU using the Cirrus emulation. I appreciate that the code could do with a bit of work, but is there a more specific reason that it should be deprecated?
ATB,
Mark.
Re: [PATCH v2 4/4] vga/cirrus: deprecate, don't build by default
Hi,
Quoted 17 lines
Quoted 8 lines
Quoted 8 lines
static const TypeInfo
cirrus_vga_info= {hw/display/cirrus_vga_isa.c +0
-0@@ -85,6 +85,7 @@ static void isa_cirrus_vga_class_init(ObjectClass *klass, void *data) dc->realize = isa_cirrus_vga_realizefn;Quoted 8 lines
hw/display/cirrus_vga_isa.c static void isa_cirrus_vga_class_init(ObjectClass *klass, void *data) · line 88
Quoted 3 lines
hw/display/cirrus_vga_isa.c +1
-0device_class_set_props(dc, isa_cirrus_vga_properties); set_bit(DEVICE_CATEGORY_DISPLAY, dc->categories); + klass->deprecation_note = "use stdvga instead";Excepr some old OSes work better with this than stdvga so could this be left and not removed? Does it cause a lot of work to keep this device? I thought it's stable already and were not many changes for it lately. If something works why drop it?
Quoted 4 lines
Seconded: whilst stdvga is preferred, there are a lot of older OSs that work well in QEMU using the Cirrus emulation. I appreciate that the code could do with a bit of work, but is there a more specific reason that it should be deprecated?
Well, the cirrus has a 2d blitter implementation which I'd rate problematic from both security and correctness point of view.
Also any guest new enough to still receive security updates can surely use stdvga instead. The "operating system museum" is IMHO pretty much the only use case where it possibly might make sense to continue using cirrus.
Having sayed that maybe the boolean classification -- be deprecated or not -- is too simple. The number of devices we can actually deprecate and remove is probably relatively small. But there also is a large number of old-ish devices which only make sense in very few use cases. When running an old OS. Or when emulating an old board. Which also tend to be not maintained very well because there are not many users. Maybe we need an "unsupported" state for them, with some infrastructure like an easy way to compile qemu without unsupported devices and an option to get warnings from qemu in case an unsupported device is used.
take care,
Gerd
Re: [PATCH v2 4/4] vga/cirrus: deprecate, don't build by default
In reply to Mark Cave-AylandQuoted 41 lines · Mark Cave-Ayland
Quoted 22 lines · BALATON Zoltan
Quoted 21 lines · Gerd Hoffmann
stdvga is the much better option.
Signed-off-by Gerd Hoffmann <[email protected]>---
hw/display/cirrus_vga.c | 1 +hw/display/cirrus_vga_isa.c| 1 + hw/display/Kconfig | 1 -3 files changed, 2 insertions(+), 1 deletion(-)hw/display/cirrus_vga.c +1
-0@@ -3007,6 +3007,7 @@ static void cirrus_vga_class_init(ObjectClass *klass, void *data) dc->vmsd = &vmstate_pci_cirrus_vga; device_class_set_props(dc, pci_vga_cirrus_properties); dc->hotpluggable = false; + klass->deprecation_note = "use stdvga instead"; }Quoted 16 lines
hw/display/cirrus_vga_isa.c static void isa_cirrus_vga_class_init(ObjectClass · line 89
Quoted 11 lines
static const TypeInfo
cirrus_vga_info= {hw/display/cirrus_vga_isa.c +1
-0@@ -85,6 +85,7 @@ static void isa_cirrus_vga_class_init(ObjectClass *klass, void *data) dc->realize = isa_cirrus_vga_realizefn; device_class_set_props(dc, isa_cirrus_vga_properties); set_bit(DEVICE_CATEGORY_DISPLAY, dc->categories); + klass->deprecation_note = "use stdvga instead";Excepr some old OSes work better with this than stdvga so could this be left and not removed? Does it cause a lot of work to keep this device? I thought it's stable already and were not many changes for it lately. If something works why drop it?
Quoted 4 lines
Seconded: whilst stdvga is preferred, there are a lot of older OSs that work well in QEMU using the Cirrus emulation. I appreciate that the code could do with a bit of work, but is there a more specific reason that it should be deprecated?
I think there's different answers here for upstream vs downstream.
Upstream QEMU's scope is to emulate pretty much arbitrary hardware that may have existed at any point in time. Emulating Cirrus is very much in scope upstream, and even if there are other better VGA devices, that doesn't make emulation of Cirrus redundant.
Downstream is a different matter - if a downstream vendor wants to be opinionated and limit the scope of devices they ship to customers, it is totally valid to cull Cirrus.
IOW, I think device deprecation *framework* is relevant to include upstream, but most actual usage of it will be downstream.
Upstream might use *object* deprecation, if we replace an backend implementation with a different one.
With regards,
Daniel
Signature · 3 lines
|: https://berrange.com -o- https://www.flickr.com/photos/dberrange :| |: https://libvirt.org -o- https://fstop138.berrange.com :| |: https://entangle-photo.org -o- https://www.instagram.com/dberrange :|
Re: [PATCH v2 4/4] vga/cirrus: deprecate, don't build by default
Quoted 7 lines · Daniel P. Berrangé
I think there's different answers here for upstream vs downstream.
Upstream QEMU's scope is to emulate pretty much arbitrary hardware that may have existed at any point in time. Emulating Cirrus is very much in scope upstream, and even if there are other better VGA devices, that doesn't make emulation of Cirrus redundant.
Quoted 6 lines
Downstream is a different matter - if a downstream vendor wants to be opinionated and limit the scope of devices they ship to customers, it is totally valid to cull Cirrus.
IOW, I think device deprecation *framework* is relevant to include upstream, but most actual usage of it will be downstream.
Right; for upstream we should expect that we use deprecation mostly as part of the "deprecate and then drop in a few releases" cycle. Deprecating something we don't want to drop doesn't seem to me to make much sense in an upstream context.
thanks
-- PMM
Re: [PATCH v2 4/4] vga/cirrus: deprecate, don't build by default
In reply to Daniel P. BerrangéQuoted 49 lines · Daniel P. Berrangé
Quoted 41 lines · Mark Cave-Ayland
Quoted 22 lines · BALATON Zoltan
Quoted 21 lines · Gerd Hoffmann
stdvga is the much better option.
Signed-off-by Gerd Hoffmann <[email protected]>---
hw/display/cirrus_vga.c | 1 +hw/display/cirrus_vga_isa.c| 1 + hw/display/Kconfig | 1 -3 files changed, 2 insertions(+), 1 deletion(-)hw/display/cirrus_vga.c +1
-0@@ -3007,6 +3007,7 @@ static void cirrus_vga_class_init(ObjectClass *klass, void *data) dc->vmsd = &vmstate_pci_cirrus_vga; device_class_set_props(dc, pci_vga_cirrus_properties); dc->hotpluggable = false; + klass->deprecation_note = "use stdvga instead"; }Quoted 16 lines
hw/display/cirrus_vga_isa.c static void isa_cirrus_vga_class_init(ObjectClass · line 89
Quoted 11 lines
static const TypeInfo
cirrus_vga_info= {hw/display/cirrus_vga_isa.c +1
-0@@ -85,6 +85,7 @@ static void isa_cirrus_vga_class_init(ObjectClass *klass, void *data) dc->realize = isa_cirrus_vga_realizefn; device_class_set_props(dc, isa_cirrus_vga_properties); set_bit(DEVICE_CATEGORY_DISPLAY, dc->categories); + klass->deprecation_note = "use stdvga instead";Excepr some old OSes work better with this than stdvga so could this be left and not removed? Does it cause a lot of work to keep this device? I thought it's stable already and were not many changes for it lately. If something works why drop it?
Quoted 4 lines
Seconded: whilst stdvga is preferred, there are a lot of older OSs that work well in QEMU using the Cirrus emulation. I appreciate that the code could do with a bit of work, but is there a more specific reason that it should be deprecated?
I think there's different answers here for upstream vs downstream.
Quoted 8 lines
Upstream QEMU's scope is to emulate pretty much arbitrary hardware that may have existed at any point in time. Emulating Cirrus is very much in scope upstream, and even if there are other better VGA devices, that doesn't make emulation of Cirrus redundant.
Downstream is a different matter - if a downstream vendor wants to be opinionated and limit the scope of devices they ship to customers, it is totally valid to cull Cirrus.
Few years ago I suggested qemu_security_policy_taint() "which allows unsafe (read "not very maintained") code to 'taint' QEMU security policy." (Gerd FYI):
https://lore.kernel.org/qemu-devel/[email protected]/
Upstream we could add a boolean in DeviceClass about device security status / maintenance (or enum or bitfield); then downstreams could use it to be sure unsafe devices aren't linked in.
Quoted 8 lines
IOW, I think device deprecation *framework* is relevant to include upstream, but most actual usage of it will be downstream.
Upstream might use *object* deprecation, if we replace an backend implementation with a different one.
With regards,
Daniel
Re: [PATCH v2 4/4] vga/cirrus: deprecate, don't build by default
Hi,
Quoted 9 lines
Quoted 8 lines
Upstream QEMU's scope is to emulate pretty much arbitrary hardware that may have existed at any point in time. Emulating Cirrus is very much in scope upstream, and even if there are other better VGA devices, that doesn't make emulation of Cirrus redundant.
Downstream is a different matter - if a downstream vendor wants to be opinionated and limit the scope of devices they ship to customers, it is totally valid to cull Cirrus.
Quoted 8 lines
Few years ago I suggested
qemu_security_policy_taint()"which allows unsafe (read "not very maintained") code to 'taint' QEMU security policy." (Gerd FYI):
https://lore.kernel.org/qemu-devel/[email protected]/Upstream we could add a boolean in DeviceClass about device security status / maintenance (or enum or bitfield); then downstreams could use it to be sure unsafe devices aren't linked in.
Yes, I think it makes sense to maintain that information upstream. It is useful information to have. Even if upstream isn't going to enforce something qemu could print useful hints.
So, the question is whenever we want go for a simple bool, or go for a bitfield giving more detailed information. Bits I think could be useful:
(1) OBJECT_STATUS_DEPRECATED
Stuff we actually want remove. Very few cases, maybe usb-hub.
(2) OBJECT_STATUS_UNSAFE
"not very maintained". Probably need a better name for this.
(3) OBJECT_STATUS_OUTDATED
Old device where newer / better alternatives exist.
Looking at the USB host adapters I'd attach 2+3 to ohci and 3 to uhci and ehci. In general there is a lot of overlap for (2) + (3) though.
We might also have recommendation bits such as:
(4) OBJECT_STATUS_VIRTUALIZATON
Devices recommended for virtualization use cases (all virtio, xhci, ...).
Quoted 2 lines
Quoted 2 lines
IOW, I think device deprecation *framework* is relevant to include upstream, but most actual usage of it will be downstream.
When doing it at ObjectClass not DeviceClass level we get introspection support for (almost) free, so I'd do it for ObjectClass even if the vast majority of users will be devices.
Quoted 2 lines
Quoted 2 lines
Upstream might use *object* deprecation, if we replace an backend implementation with a different one.
Sounds like having OBJECT_STATUS_DEPRECATED makes sense even if we don't have any device actually using that.
take care,
Gerd
Re: [PATCH v2 4/4] vga/cirrus: deprecate, don't build by default
Quoted 28 lines · Gerd Hoffmann
Hi,
Quoted 9 lines
Quoted 8 lines
Upstream QEMU's scope is to emulate pretty much arbitrary hardware that may have existed at any point in time. Emulating Cirrus is very much in scope upstream, and even if there are other better VGA devices, that doesn't make emulation of Cirrus redundant.
Downstream is a different matter - if a downstream vendor wants to be opinionated and limit the scope of devices they ship to customers, it is totally valid to cull Cirrus.
Quoted 8 lines
Few years ago I suggested
qemu_security_policy_taint()"which allows unsafe (read "not very maintained") code to 'taint' QEMU security policy." (Gerd FYI):
https://lore.kernel.org/qemu-devel/[email protected]/Upstream we could add a boolean in DeviceClass about device security status / maintenance (or enum or bitfield); then downstreams could use it to be sure unsafe devices aren't linked in.
Yes, I think it makes sense to maintain that information upstream. It is useful information to have. Even if upstream isn't going to enforce something qemu could print useful hints.
So, the question is whenever we want go for a simple bool, or go for a bitfield giving more detailed information. Bits I think could be useful:
Quoted 5 lines
(1)
OBJECT_STATUS_DEPRECATED
Stuff we actually want remove. Very few cases, maybe usb-hub.(2)
OBJECT_STATUS_UNSAFE
"not very maintained". Probably need a better name for this.
If it reflects maintainer status, then we're obvioulsy overlapping with the MAINTAINERS file info, but "UNSAFE" suggests something different....
So what I think would be valuable is marking whether a device is considered to provide a guest security boundary. This would in turn indicate whether we would treat flaws in its impl as being worthy of triaging as CVEs, or merely be normal bugs.
We already document in security.rst that we consider virtualization use cases only to provide a security boundary, but on every bug report we have to decide whether the device in question is considered relevant to a "virtualization" use case.
This was a known limitation, because we had no metadata to track which devices were considered "secure". It was always expected that long term we should be tagging devices/machines/accelerators/etc with their security status.
It would be nice from an end user POV to be able to have a way to tell QEMU to enforce that a given VM provides a guest security boundary, and get a hard error at startup, or hot-plug if any device cannot satisfy that requirement.
Quoted 2 lines
(3)
OBJECT_STATUS_OUTDATED
Old device where newer / better alternatives exist.
I think (3) is pretty hard to define, as "better" is very much a use case specific decision. You could say that everything, for which a virtio device exists, has a "better" alternative. That's only true if running a modern OS though. If running a vintage OS, the "outdated" thing is liley "better".
Quoted 8 lines
Looking at the USB host adapters I'd attach 2+3 to ohci and 3 to uhci and ehci. In general there is a lot of overlap for (2) + (3) though.
We might also have recommendation bits such as:
(4)
OBJECT_STATUS_VIRTUALIZATON
Devices recommended for virtualization use cases (all virtio, xhci, ...).
Even that has the same problem as (3) - the recommended devices differ depending on what OS you're intending to use.
This problem is pretty much why libosinfo exists to provide more guided hardware recommendations tailored to OS.
With regards,
Daniel
Signature · 3 lines
|: https://berrange.com -o- https://www.flickr.com/photos/dberrange :| |: https://libvirt.org -o- https://fstop138.berrange.com :| |: https://entangle-photo.org -o- https://www.instagram.com/dberrange :|
Re: [PATCH v2 4/4] vga/cirrus: deprecate, don't build by default
In reply to Daniel P. BerrangéQuoted 49 lines · Daniel P. Berrangé
Quoted 41 lines · Mark Cave-Ayland
Quoted 22 lines · BALATON Zoltan
Quoted 21 lines · Gerd Hoffmann
stdvga is the much better option.
Signed-off-by Gerd Hoffmann <[email protected]>---
hw/display/cirrus_vga.c | 1 +hw/display/cirrus_vga_isa.c| 1 + hw/display/Kconfig | 1 -3 files changed, 2 insertions(+), 1 deletion(-)hw/display/cirrus_vga.c +1
-0@@ -3007,6 +3007,7 @@ static void cirrus_vga_class_init(ObjectClass *klass, void *data) dc->vmsd = &vmstate_pci_cirrus_vga; device_class_set_props(dc, pci_vga_cirrus_properties); dc->hotpluggable = false; + klass->deprecation_note = "use stdvga instead"; }Quoted 16 lines
hw/display/cirrus_vga_isa.c static void isa_cirrus_vga_class_init(ObjectClass · line 89
Quoted 11 lines
static const TypeInfo
cirrus_vga_info= {hw/display/cirrus_vga_isa.c +1
-0@@ -85,6 +85,7 @@ static void isa_cirrus_vga_class_init(ObjectClass *klass, void *data) dc->realize = isa_cirrus_vga_realizefn; device_class_set_props(dc, isa_cirrus_vga_properties); set_bit(DEVICE_CATEGORY_DISPLAY, dc->categories); + klass->deprecation_note = "use stdvga instead";Excepr some old OSes work better with this than stdvga so could this be left and not removed? Does it cause a lot of work to keep this device? I thought it's stable already and were not many changes for it lately. If something works why drop it?
Quoted 4 lines
Seconded: whilst stdvga is preferred, there are a lot of older OSs that work well in QEMU using the Cirrus emulation. I appreciate that the code could do with a bit of work, but is there a more specific reason that it should be deprecated?
I think there's different answers here for upstream vs downstream.
Quoted 8 lines
Upstream QEMU's scope is to emulate pretty much arbitrary hardware that may have existed at any point in time. Emulating Cirrus is very much in scope upstream, and even if there are other better VGA devices, that doesn't make emulation of Cirrus redundant.
Downstream is a different matter - if a downstream vendor wants to be opinionated and limit the scope of devices they ship to customers, it is totally valid to cull Cirrus.
The concern for me is that if someone such as RedHat decided not to ship Cirrus then we'd end up in a place where command lines for some legacy OSs would work on an upstream build, but if someone was using RHEL then they would fail due to the device not being present. I can see this causing confusion for users since they would report that a command line doesn't work whilst others would shrug and report back that it works for them.
I do wonder if a solution for this would be to add a blocklist file in /etc that prevents the listed QOM types from being instantiated. The file could contain also contain a machine regex to match and a reason that can be reported to the user e.g.
# QEMU QOM type blocklist
#
# QOM type regex, machine regex list, reason "cirrus*","pc*,machine*","contains insecure blitter routines" "fdc","pc*","may not be completely secure"
This feels like a better solution because:
- It's easy to add a message that reports "The requested QOM type <foo> cannot be instantiated because <reason>" for specific machine types. The machine regex also fixes the problem where usb-ohci should be allowed for PPC machines, but not generally for PC machines.
- Downstream can ship with a secure policy for production environments but also a less restrictive policy for more casual users
- If someone really needs a device that is not enabled by default, a privileged user can simply edit the blocklist file and allow it
- It should work both with or without modules
ATB,
Mark.
Re: [PATCH v2 4/4] vga/cirrus: deprecate, don't build by default
Quoted 60 lines · Mark Cave-Ayland
Quoted 49 lines · Daniel P. Berrangé
Quoted 41 lines · Mark Cave-Ayland
Quoted 22 lines · BALATON Zoltan
Quoted 21 lines · Gerd Hoffmann
stdvga is the much better option.
Signed-off-by Gerd Hoffmann <[email protected]>---
hw/display/cirrus_vga.c | 1 +hw/display/cirrus_vga_isa.c| 1 + hw/display/Kconfig | 1 -3 files changed, 2 insertions(+), 1 deletion(-)hw/display/cirrus_vga.c +1
-0@@ -3007,6 +3007,7 @@ static void cirrus_vga_class_init(ObjectClass *klass, void *data) dc->vmsd = &vmstate_pci_cirrus_vga; device_class_set_props(dc, pci_vga_cirrus_properties); dc->hotpluggable = false; + klass->deprecation_note = "use stdvga instead"; }Quoted 16 lines
hw/display/cirrus_vga_isa.c static void isa_cirrus_vga_class_init(ObjectClass · line 89
Quoted 11 lines
static const TypeInfo
cirrus_vga_info= {hw/display/cirrus_vga_isa.c +1
-0@@ -85,6 +85,7 @@ static void isa_cirrus_vga_class_init(ObjectClass *klass, void *data) dc->realize = isa_cirrus_vga_realizefn; device_class_set_props(dc, isa_cirrus_vga_properties); set_bit(DEVICE_CATEGORY_DISPLAY, dc->categories); + klass->deprecation_note = "use stdvga instead";Excepr some old OSes work better with this than stdvga so could this be left and not removed? Does it cause a lot of work to keep this device? I thought it's stable already and were not many changes for it lately. If something works why drop it?
Quoted 4 lines
Seconded: whilst stdvga is preferred, there are a lot of older OSs that work well in QEMU using the Cirrus emulation. I appreciate that the code could do with a bit of work, but is there a more specific reason that it should be deprecated?
I think there's different answers here for upstream vs downstream.
Quoted 8 lines
Upstream QEMU's scope is to emulate pretty much arbitrary hardware that may have existed at any point in time. Emulating Cirrus is very much in scope upstream, and even if there are other better VGA devices, that doesn't make emulation of Cirrus redundant.
Downstream is a different matter - if a downstream vendor wants to be opinionated and limit the scope of devices they ship to customers, it is totally valid to cull Cirrus.
Quoted 6 lines
The concern for me is that if someone such as RedHat decided not to ship Cirrus then we'd end up in a place where command lines for some legacy OSs would work on an upstream build, but if someone was using RHEL then they would fail due to the device not being present. I can see this causing confusion for users since they would report that a command line doesn't work whilst others would shrug and report back that it works for them.
That ship sailed in RHEL over a decade and a half ago.
There is already a mountain of devices and other QEMU features that are /not/ shipped in RHEL, disabled at build time. Essentially RHEL is only targetting virtualization use cases, not emulation, so the priority is virtio devices, paired with a tiny handful of emulated devices to fill some gaps. Historically RHEL included Cirrus, but it was marked deprecated in RHEL quite a few years ago now, and will likely be removed entirely in RHEL-10. The combo of stdvga and virtio-vga/gpu is sufficient for the virtualization use case.
Yes, this does cause confusion and annoyance for users who want to try to use RHEL with QEMU cli configs they find around the web, but that's considered acceptable pain.
Quoted 18 lines
I do wonder if a solution for this would be to add a blocklist file in /etc that prevents the listed QOM types from being instantiated. The file could contain also contain a machine regex to match and a reason that can be reported to the user e.g.
# QEMU QOM type blocklist
#
# QOM type regex, machine regex list, reason "cirrus*","pc*,machine*","contains insecure blitter routines" "fdc","pc*","may not be completely secure"This feels like a better solution because:
- It's easy to add a message that reports "The requested QOM type <foo> cannot be instantiated because <reason>" for specific machine types. The machine regex also fixes the problem where usb-ohci should be allowed for PPC machines, but not generally for PC machines.
Quoted 7 lines
- Downstream can ship with a secure policy for production environments but also a less restrictive policy for more casual users
- If someone really needs a device that is not enabled by default, a privileged user can simply edit the blocklist file and allow it
- It should work both with or without modules
I don't see a user edittable config being useful in RHEL or Fedora. For RHEL we have a strict policy of only shipping what we want to support, and everything else must be fully disabled at build time. Conversely in Fedora we aim to ship everything that QEMU provides.
I would be nice to have a tagging of "quality" status for devices upstream, but that's not something that needs to be turned into a user edittable matrix against machine types. A device impl is either good or not - the code impl quality doesn't vary per machine usage, nor should upstream try to artificially block usage per machine.
With regards,
Daniel
Signature · 3 lines
|: https://berrange.com -o- https://www.flickr.com/photos/dberrange :| |: https://libvirt.org -o- https://fstop138.berrange.com :| |: https://entangle-photo.org -o- https://www.instagram.com/dberrange :|