content: enhance device requirements for feature bits
PATCH v3 1/1 content: enhance device requirements for feature bits
---
v2:
- Refine the wording (Cornelia);
v3:
- Refine the wording (MST);
content.tex | 7 +++++++ 1 file changed, 7 insertions(+)
content.tex +7 -0
@@ -125,6 +125,13 @@ which was not offered. The device SHOULD accept any valid subset of features the driver accepts, otherwise it MUST fail to set the FEATURES_OK \field{device status} bit when the driver writes it. +If a device has successfully negotiated a set of features +at least once (by accepting the FEATURES_OK \field{device +status} bit during device initialization), then it SHOULD +NOT fail re-negotiation of the same set of features after +a device or system reset. Failure to do so would interfere +with resuming from suspend and error recovery. + \subsection{Legacy Interface: A Note on Feature Bits}\label{sec:Basic Facilities of a Virtio Device / Feature Bits / Legacy Interface: A Note on Feature Bits}
Signature · 6 lines
2.17.0 --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
Quoted 21 lines · On Mon, 11 Jun 2018 15:56:40 +0800 Tiwei Bie <[email protected]> wrote:
Suggested-by Michael S. Tsirkin <[email protected]>Signed-off-by Tiwei Bie <[email protected]>---
v2:
- Refine the wording (Cornelia);v3:
- Refine the wording (MST);content.tex | 7 +++++++ 1 file changed, 7 insertions(+)content.tex +0
-0@@ -125,6 +125,13 @@ which was not offered. The device SHOULD accept any valid subset of features the driver accepts, otherwise it MUST fail to set the FEATURES_OK \field{device status} bit when the driver writes it.
content.tex which was not offered. The device SHOULD accept any valid subset · line 137
Quoted 10 lines
content.tex +7
-0+If a device has successfully negotiated a set of features +at least once (by accepting the FEATURES_OK \field{device +status} bit during device initialization), then it SHOULD +NOT fail re-negotiation of the same set of features after +a device or system reset. Failure to do so would interfere +with resuming from suspend and error recovery. + \subsection{Legacy Interface: A Note on Feature Bits}\label{sec:Basic Facilities of a Virtio Device / Feature Bits / Legacy Interface: A Note on Feature Bits}
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
Quoted 21 lines · Tiwei Bie
Suggested-by Michael S. Tsirkin <[email protected]>Signed-off-by Tiwei Bie <[email protected]>---
v2:
- Refine the wording (Cornelia);v3:
- Refine the wording (MST);content.tex | 7 +++++++ 1 file changed, 7 insertions(+)content.tex +0
-0@@ -125,6 +125,13 @@ which was not offered. The device SHOULD accept any valid subset of features the driver accepts, otherwise it MUST fail to set the FEATURES_OK \field{device status} bit when the driver writes it.
content.tex which was not offered. The device SHOULD accept any valid subset · line 137
Quoted 10 lines
content.tex +7
-0+If a device has successfully negotiated a set of features +at least once (by accepting the FEATURES_OK \field{device +status} bit during device initialization), then it SHOULD +NOT fail re-negotiation of the same set of features after +a device or system reset. Failure to do so would interfere +with resuming from suspend and error recovery. + \subsection{Legacy Interface: A Note on Feature Bits}\label{sec:Basic Facilities of a Virtio Device / Feature Bits / Legacy Interface: A Note on Feature Bits}
OK but there's no \field{device status} anywhere
else I think.
We only have \field{status} in spec.
Quoted 2 lines
Signature · 1 line
2.17.0
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
Quoted 22 lines · On Mon, 11 Jun 2018 16:24:51 +0300 "Michael S. Tsirkin" <[email protected]> wrote:
Quoted 21 lines · Tiwei Bie
Suggested-by Michael S. Tsirkin <[email protected]>Signed-off-by Tiwei Bie <[email protected]>---
v2:
- Refine the wording (Cornelia);v3:
- Refine the wording (MST);content.tex | 7 +++++++ 1 file changed, 7 insertions(+)content.tex +0
-0@@ -125,6 +125,13 @@ which was not offered. The device SHOULD accept any valid subset of features the driver accepts, otherwise it MUST fail to set the FEATURES_OK \field{device status} bit when the driver writes it.
Quoted 14 lines
content.tex which was not offered. The device SHOULD accept any valid subset · line 137
Quoted 10 lines
content.tex +7
-0+If a device has successfully negotiated a set of features +at least once (by accepting the FEATURES_OK \field{device +status} bit during device initialization), then it SHOULD +NOT fail re-negotiation of the same set of features after +a device or system reset. Failure to do so would interfere +with resuming from suspend and error recovery. + \subsection{Legacy Interface: A Note on Feature Bits}\label{sec:Basic Facilities of a Virtio Device / Feature Bits / Legacy Interface: A Note on Feature Bits}OK but there's no \field{device status} anywhere
else I think.
We only have \field{status} in spec.
Should probably be "device \field{status} bit", then?
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
Quoted 40 lines · Cornelia Huck
Quoted 22 lines · On Mon, 11 Jun 2018 16:24:51 +0300 "Michael S. Tsirkin" <[email protected]> wrote:
Quoted 21 lines · Tiwei Bie
Suggested-by Michael S. Tsirkin <[email protected]>Signed-off-by Tiwei Bie <[email protected]>---
v2:
- Refine the wording (Cornelia);v3:
- Refine the wording (MST);content.tex | 7 +++++++ 1 file changed, 7 insertions(+)content.tex +0
-0@@ -125,6 +125,13 @@ which was not offered. The device SHOULD accept any valid subset of features the driver accepts, otherwise it MUST fail to set the FEATURES_OK \field{device status} bit when the driver writes it.Quoted 14 lines
content.tex which was not offered. The device SHOULD accept any valid subset · line 137
Quoted 10 lines
content.tex +7
-0+If a device has successfully negotiated a set of features +at least once (by accepting the FEATURES_OK \field{device +status} bit during device initialization), then it SHOULD +NOT fail re-negotiation of the same set of features after +a device or system reset. Failure to do so would interfere +with resuming from suspend and error recovery. + \subsection{Legacy Interface: A Note on Feature Bits}\label{sec:Basic Facilities of a Virtio Device / Feature Bits / Legacy Interface: A Note on Feature Bits}OK but there's no \field{device status} anywhere
else I think.
We only have \field{status} in spec.
Quoted 1 line
Should probably be "device \field{status} bit", then?
I'm actually wrong, it is called device status in "Basic Facilities of a Virtio Device". We probably should fix the places that
call it \field{status}.
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
Quoted 3 lines · Tiwei Bie
Suggested-by Michael S. Tsirkin <[email protected]>Signed-off-by Tiwei Bie <[email protected]>
Quoted 30 lines
---
v2:
- Refine the wording (Cornelia);v3:
- Refine the wording (MST);content.tex | 7 +++++++ 1 file changed, 7 insertions(+)content.tex +7
-0@@ -125,6 +125,13 @@ which was not offered. The device SHOULD accept any valid subset of features the driver accepts, otherwise it MUST fail to set the FEATURES_OK \field{device status} bit when the driver writes it. +If a device has successfully negotiated a set of features +at least once (by accepting the FEATURES_OK \field{device +status} bit during device initialization), then it SHOULD +NOT fail re-negotiation of the same set of features after +a device or system reset. Failure to do so would interfere +with resuming from suspend and error recovery. + \subsection{Legacy Interface: A Note on Feature Bits}\label{sec:Basic Facilities of a Virtio Device / Feature Bits / Legacy Interface: A Note on Feature Bits}Signature · 1 line
2.17.0
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
Quoted 18 lines · Tiwei Bie
Suggested-by Michael S. Tsirkin <[email protected]>Signed-off-by Tiwei Bie <[email protected]>---
v2:
- Refine the wording (Cornelia);v3:
- Refine the wording (MST);content.tex | 7 +++++++ 1 file changed, 7 insertions(+)content.tex +0
-0
content.tex which was not offered. The device SHOULD accept any valid subset · line 134
Quoted 10 lines
content.tex +7
-0@@ -125,6 +125,13 @@ which was not offered. The device SHOULD accept any valid subset of features the driver accepts, otherwise it MUST fail to set the FEATURES_OK \field{device status} bit when the driver writes it. +If a device has successfully negotiated a set of features +at least once (by accepting the FEATURES_OK \field{device +status} bit during device initialization), then it SHOULD +NOT fail re-negotiation of the same set of features after +a device or system reset. Failure to do so would interfere +with resuming from suspend and error recovery. +
Sorry people but I don't get it. I mean it is kind of reasonable to assume that with a given device and a given driver (given, i.e. nothing changes) the two will always negotiate the same features (including the extremal case where the negotiation fails).
Either the device or a driver rolling a dice to make feature negotiation more fun seems quite unreasonable. So I assume this is not what we are bothering to soft prohibit here.
So the interesting scenario seems to be when stuff changes. When migrating the implementation of the device could change. Or something changes regarding the resources used to provide the virtual device.
But then, if the device really can not support the set of features it used to be able, I guess the SHOULD does not take effect (I guess that is the difference compared to MUST).
Bottom line is: I tried to figure out what is this about, but I failed. I've read https://github.com/oasis-tcs/virtio-spec/issues/14 too but it did not click. I would appreciate some assistance.
content.tex which was not offered. The device SHOULD accept any valid subset · line 137
Quoted 4 lines
content.tex +0
-0\subsection{Legacy Interface: A Note on Feature Bits}\label{sec:Basic Facilities of a Virtio Device / Feature Bits / Legacy Interface: A Note on Feature Bits}
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
Quoted 45 lines · Halil Pasic
Quoted 18 lines · Tiwei Bie
Suggested-by Michael S. Tsirkin <[email protected]>Signed-off-by Tiwei Bie <[email protected]>---
v2:
- Refine the wording (Cornelia);v3:
- Refine the wording (MST);content.tex | 7 +++++++ 1 file changed, 7 insertions(+)content.tex +0
-0content.tex which was not offered. The device SHOULD accept any valid subset · line 133
Quoted 9 lines
content.tex +7
-0@@ -125,6 +125,13 @@ which was not offered. The device SHOULD accept any valid subset of features the driver accepts, otherwise it MUST fail to set the FEATURES_OK \field{device status} bit when the driver writes it. +If a device has successfully negotiated a set of features +at least once (by accepting the FEATURES_OK \field{device +status} bit during device initialization), then it SHOULD +NOT fail re-negotiation of the same set of features after +a device or system reset. Failure to do so would interfere +with resuming from suspend and error recovery. +Sorry people but I don't get it. I mean it is kind of reasonable to assume that with a given device and a given driver (given, i.e. nothing changes) the two will always negotiate the same features (including the extremal case where the negotiation fails).
Either the device or a driver rolling a dice to make feature negotiation more fun seems quite unreasonable. So I assume this is not what we are bothering to soft prohibit here.
So the interesting scenario seems to be when stuff changes. When migrating the implementation of the device could change. Or something changes regarding the resources used to provide the virtual device.
Quoted 7 lines
But then, if the device really can not support the set of features it used to be able, I guess the SHOULD does not take effect (I guess that is the difference compared to MUST).
Bottom line is: I tried to figure out what is this about, but I failed. I've read https://github.com/oasis-tcs/virtio-spec/issues/14 too but it did not click. I would appreciate some assistance.
It's exactly what it says. Let's say you negotiated a feature and then device sets NEED_RESET. Driver must now reset the device and put it back in the same state it had before the reset, then resubmit requests that were available but never used.
What if any of the features changed? Device suddenly needs to check for requests which do not match the features.
Suspend is similar: guests tend to assume hardware does not change across suspend/resume, any changes tend to make resume fail.
Quoted 5 lines
content.tex which was not offered. The device SHOULD accept any valid subset
Quoted 5 lines
content.tex +0
-0\subsection{Legacy Interface: A Note on Feature Bits}\label{sec:Basic Facilities of a Virtio Device / Feature Bits / Legacy Interface: A Note on Feature Bits}
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
Quoted 59 lines · Michael S. Tsirkin
Quoted 45 lines · Halil Pasic
Quoted 18 lines · Tiwei Bie
Suggested-by Michael S. Tsirkin <[email protected]>Signed-off-by Tiwei Bie <[email protected]>---
v2:
- Refine the wording (Cornelia);v3:
- Refine the wording (MST);content.tex | 7 +++++++ 1 file changed, 7 insertions(+)content.tex +0
-0content.tex which was not offered. The device SHOULD accept any valid subset · line 133
Quoted 9 lines
content.tex +7
-0@@ -125,6 +125,13 @@ which was not offered. The device SHOULD accept any valid subset of features the driver accepts, otherwise it MUST fail to set the FEATURES_OK \field{device status} bit when the driver writes it. +If a device has successfully negotiated a set of features +at least once (by accepting the FEATURES_OK \field{device +status} bit during device initialization), then it SHOULD +NOT fail re-negotiation of the same set of features after +a device or system reset. Failure to do so would interfere +with resuming from suspend and error recovery. +Sorry people but I don't get it. I mean it is kind of reasonable to assume that with a given device and a given driver (given, i.e. nothing changes) the two will always negotiate the same features (including the extremal case where the negotiation fails).
Either the device or a driver rolling a dice to make feature negotiation more fun seems quite unreasonable. So I assume this is not what we are bothering to soft prohibit here.
So the interesting scenario seems to be when stuff changes. When migrating the implementation of the device could change. Or something changes regarding the resources used to provide the virtual device.
Quoted 7 lines
But then, if the device really can not support the set of features it used to be able, I guess the SHOULD does not take effect (I guess that is the difference compared to MUST).
Bottom line is: I tried to figure out what is this about, but I failed. I've read https://github.com/oasis-tcs/virtio-spec/issues/14 too but it did not click. I would appreciate some assistance.
It's exactly what it says. Let's say you negotiated a feature and then device sets
NEED_RESET. Driver must now reset the device and put it back in the same state it had before the reset, then resubmit requests that were available but never used.
Quoted 8 lines
What if any of the features changed? Device suddenly needs to check for requests which do not match the features.
Suspend is similar: guests tend to assume hardware does not change across suspend/resume, any changes tend to make resume fail.
Thank you very much! But it still does not answer why would a device want to do that (fail to negotiate a feature that it was able to negotiate before). So I'm still in the dark about what are we trading for what.
Is there somewhere a patch that fixes such a bug? Maybe that would help me understand what can be done at the device to avoid the problem.
Regards,
Halil
Quoted 10 lines
Quoted 5 lines
content.tex which was not offered. The device SHOULD accept any valid subset
Quoted 5 lines
content.tex +0
-0\subsection{Legacy Interface: A Note on Feature Bits}\label{sec:Basic Facilities of a Virtio Device / Feature Bits / Legacy Interface: A Note on Feature Bits}---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
Quoted 66 lines · Halil Pasic
Quoted 63 lines · Michael S. Tsirkin
Quoted 45 lines · Halil Pasic
Quoted 18 lines · Tiwei Bie
Suggested-by Michael S. Tsirkin <[email protected]>Signed-off-by Tiwei Bie <[email protected]>---
v2:
- Refine the wording (Cornelia);v3:
- Refine the wording (MST);content.tex | 7 +++++++ 1 file changed, 7 insertions(+)content.tex +0
-0content.tex which was not offered. The device SHOULD accept any valid subset · line 133
Quoted 9 lines
content.tex +7
-0@@ -125,6 +125,13 @@ which was not offered. The device SHOULD accept any valid subset of features the driver accepts, otherwise it MUST fail to set the FEATURES_OK \field{device status} bit when the driver writes it. +If a device has successfully negotiated a set of features +at least once (by accepting the FEATURES_OK \field{device +status} bit during device initialization), then it SHOULD +NOT fail re-negotiation of the same set of features after +a device or system reset. Failure to do so would interfere +with resuming from suspend and error recovery. +Sorry people but I don't get it. I mean it is kind of reasonable to assume that with a given device and a given driver (given, i.e. nothing changes) the two will always negotiate the same features (including the extremal case where the negotiation fails).
Either the device or a driver rolling a dice to make feature negotiation more fun seems quite unreasonable. So I assume this is not what we are bothering to soft prohibit here.
So the interesting scenario seems to be when stuff changes. When migrating the implementation of the device could change. Or something changes regarding the resources used to provide the virtual device.
Quoted 7 lines
But then, if the device really can not support the set of features it used to be able, I guess the SHOULD does not take effect (I guess that is the difference compared to MUST).
Bottom line is: I tried to figure out what is this about, but I failed. I've read https://github.com/oasis-tcs/virtio-spec/issues/14 too but it did not click. I would appreciate some assistance.
It's exactly what it says. Let's say you negotiated a feature and then device sets
NEED_RESET. Driver must now reset the device and put it back in the same state it had before the reset, then resubmit requests that were available but never used.What if any of the features changed? Device suddenly needs to check for requests which do not match the features.
Quoted 8 lines
Quoted 3 lines
Suspend is similar: guests tend to assume hardware does not change across suspend/resume, any changes tend to make resume fail.
Thank you very much! But it still does not answer why would a device want to do that (fail to negotiate a feature that it was able to negotiate before). So I'm still in the dark about what are we trading for what.
It would be a mis-configured device. For example QEMU does not migrate the device features so if you misconfigure QEMU with different flags on source and destination (not a supported configuration), features might seem to change from guest POV.
Quoted 4 lines
Is there somewhere a patch that fixes such a bug? Maybe that would help me understand what can be done at the device to avoid the problem.
Quoted 14 lines
Regards,
HalilQuoted 10 lines
Quoted 5 lines
content.tex which was not offered. The device SHOULD accept any valid subset
Quoted 5 lines
content.tex +0
-0\subsection{Legacy Interface: A Note on Feature Bits}\label{sec:Basic Facilities of a Virtio Device / Feature Bits / Legacy Interface: A Note on Feature Bits}---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
Quoted 69 lines · Halil Pasic
Quoted 59 lines · Michael S. Tsirkin
Quoted 45 lines · Halil Pasic
Quoted 18 lines · Tiwei Bie
Suggested-by Michael S. Tsirkin <[email protected]>Signed-off-by Tiwei Bie <[email protected]>---
v2:
- Refine the wording (Cornelia);v3:
- Refine the wording (MST);content.tex | 7 +++++++ 1 file changed, 7 insertions(+)content.tex +0
-0content.tex which was not offered. The device SHOULD accept any valid subset · line 133
Quoted 9 lines
content.tex +7
-0@@ -125,6 +125,13 @@ which was not offered. The device SHOULD accept any valid subset of features the driver accepts, otherwise it MUST fail to set the FEATURES_OK \field{device status} bit when the driver writes it. +If a device has successfully negotiated a set of features +at least once (by accepting the FEATURES_OK \field{device +status} bit during device initialization), then it SHOULD +NOT fail re-negotiation of the same set of features after +a device or system reset. Failure to do so would interfere +with resuming from suspend and error recovery. +Sorry people but I don't get it. I mean it is kind of reasonable to assume that with a given device and a given driver (given, i.e. nothing changes) the two will always negotiate the same features (including the extremal case where the negotiation fails).
Either the device or a driver rolling a dice to make feature negotiation more fun seems quite unreasonable. So I assume this is not what we are bothering to soft prohibit here.
So the interesting scenario seems to be when stuff changes. When migrating the implementation of the device could change. Or something changes regarding the resources used to provide the virtual device.
Quoted 7 lines
But then, if the device really can not support the set of features it used to be able, I guess the SHOULD does not take effect (I guess that is the difference compared to MUST).
Bottom line is: I tried to figure out what is this about, but I failed. I've read https://github.com/oasis-tcs/virtio-spec/issues/14 too but it did not click. I would appreciate some assistance.
It's exactly what it says. Let's say you negotiated a feature and then device sets
NEED_RESET. Driver must now reset the device and put it back in the same state it had before the reset, then resubmit requests that were available but never used.Quoted 8 lines
What if any of the features changed? Device suddenly needs to check for requests which do not match the features.
Suspend is similar: guests tend to assume hardware does not change across suspend/resume, any changes tend to make resume fail.
Quoted 4 lines
Thank you very much! But it still does not answer why would a device want to do that (fail to negotiate a feature that it was able to negotiate before). So I'm still in the dark about what are we trading for what.
Hi Halil,
Just like what you said, normally there is no reason for a device to fail to negotiate a feature that it was able to negotiate before. But the spec doesn't forbid devices to do this , i.e. the spec allows a device to fail to negotiate a feature that it was able to negotiate before, which could cause problems in some cases. Although everything works fine in reality because there is no device would really do this, it would be better to make spec to explicitly forbid devices to do this in the necessary cases.
Best regards,
Tiwei Bie
Quoted 20 lines
Is there somewhere a patch that fixes such a bug? Maybe that would help me understand what can be done at the device to avoid the problem.
Regards,
HalilQuoted 10 lines
Quoted 5 lines
content.tex which was not offered. The device SHOULD accept any valid subset
Quoted 5 lines
content.tex +0
-0\subsection{Legacy Interface: A Note on Feature Bits}\label{sec:Basic Facilities of a Virtio Device / Feature Bits / Legacy Interface: A Note on Feature Bits}---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
Quoted 88 lines · Tiwei Bie
Quoted 69 lines · Halil Pasic
Quoted 59 lines · Michael S. Tsirkin
Quoted 45 lines · Halil Pasic
Quoted 18 lines · Tiwei Bie
Suggested-by Michael S. Tsirkin <[email protected]>Signed-off-by Tiwei Bie <[email protected]>---
v2:
- Refine the wording (Cornelia);v3:
- Refine the wording (MST);content.tex | 7 +++++++ 1 file changed, 7 insertions(+)content.tex +0
-0content.tex which was not offered. The device SHOULD accept any valid subset · line 133
Quoted 9 lines
content.tex +7
-0@@ -125,6 +125,13 @@ which was not offered. The device SHOULD accept any valid subset of features the driver accepts, otherwise it MUST fail to set the FEATURES_OK \field{device status} bit when the driver writes it. +If a device has successfully negotiated a set of features +at least once (by accepting the FEATURES_OK \field{device +status} bit during device initialization), then it SHOULD +NOT fail re-negotiation of the same set of features after +a device or system reset. Failure to do so would interfere +with resuming from suspend and error recovery. +Sorry people but I don't get it. I mean it is kind of reasonable to assume that with a given device and a given driver (given, i.e. nothing changes) the two will always negotiate the same features (including the extremal case where the negotiation fails).
Either the device or a driver rolling a dice to make feature negotiation more fun seems quite unreasonable. So I assume this is not what we are bothering to soft prohibit here.
So the interesting scenario seems to be when stuff changes. When migrating the implementation of the device could change. Or something changes regarding the resources used to provide the virtual device.
Quoted 7 lines
But then, if the device really can not support the set of features it used to be able, I guess the SHOULD does not take effect (I guess that is the difference compared to MUST).
Bottom line is: I tried to figure out what is this about, but I failed. I've read https://github.com/oasis-tcs/virtio-spec/issues/14 too but it did not click. I would appreciate some assistance.
It's exactly what it says. Let's say you negotiated a feature and then device sets
NEED_RESET. Driver must now reset the device and put it back in the same state it had before the reset, then resubmit requests that were available but never used.Quoted 8 lines
What if any of the features changed? Device suddenly needs to check for requests which do not match the features.
Suspend is similar: guests tend to assume hardware does not change across suspend/resume, any changes tend to make resume fail.
Quoted 4 lines
Thank you very much! But it still does not answer why would a device want to do that (fail to negotiate a feature that it was able to negotiate before). So I'm still in the dark about what are we trading for what.
Hi Halil,
Just like what you said, normally there is no reason for a device to fail to negotiate a feature that it was able to negotiate before. But the spec doesn't forbid devices to do this , i.e. the spec allows a device to fail to negotiate a feature that it was able to negotiate before, which could cause problems in some cases. Although everything works fine in reality because there is no device would really do this, it would be better to make spec to explicitly forbid devices to do this in the necessary cases.
Quoted 3 lines
Best regards,
Tiwei Bie
I think we have most of it already covered with 'The device SHOULD accept any valid subset of features the driver accepts'.
IMHO what we add with your proposed normative statement is that if the device used to offer a feature bit it SHOULD keep offering it. That's clearly not covered by the by what I've cited.
But it's kind of covered by a non-normative statement 'Each virtio device offers all the features it understands.'
This seems most relevant in case of migration. That is device implementation S(ource) and device implementation T(arget) are migration compatible. But hey, features that are present
in S and not present in T are of concern for migration compatibility. AFAIK the VIRTIO specification does not make claims about migration compatibility.
So if I think QEMU, and somebody (maintainer) is deciding to remove support for of a certain device for a certain feature bit in the next version, he better thinks hard how could this breakmigration. I don't think the proposed normative statement with it's SHOULD would make the the guy more careful.
What is even more interesting is the scenario where the new version of the device does not remove support for a feature, but adds support for one, let's call it F_N.
The scenario is the following we have systems O(ld) and N(ew). We start on O then we migrate to new. There some reset of concern happens. Features get re-negotiated and we start exploiting F_N. In my reading of your addition, this is legit. But then we migrate back from N to O. No re-negotiation happens (because it is not obligatory), and things explode (hopefully, just migration fails, and not guest dies) because O does not have support for F_N. Your normative statement was nowhere violated as far as I can tell.
Bottom line is, I still don't know what benefit does this addition to the standard have to the implementer of the standard. In my opinion it's just another chunk of text that is hard to figure out. It's hard to tell what is the device and what is before, what is system reset. If we were to make the spec complete with spelling out every 'don't make anything stupid' I'm under the impression there is a lot of work to do. I had a discussion here on the completeness of this spec, and completeness does not seem to be a primary goal. I'm still not sold on this one.
Regards,
Halil
Quoted 15 lines
Quoted 15 lines
Is there somewhere a patch that fixes such a bug? Maybe that would help me understand what can be done at the device to avoid the problem.
Regards,
HalilQuoted 6 lines
Quoted 5 lines
Quoted 5 lines
content.tex +0
-0\subsection{Legacy Interface: A Note on Feature Bits}\label{sec:Basic Facilities of a Virtio Device / Feature Bits / Legacy Interface: A Note on Feature Bits}
Quoted 6 lines
Quoted 5 lines
Quoted 4 lines
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
Quoted 76 lines · Michael S. Tsirkin
Quoted 66 lines · Halil Pasic
Quoted 63 lines · Michael S. Tsirkin
Quoted 45 lines · Halil Pasic
Quoted 18 lines · Tiwei Bie
Suggested-by Michael S. Tsirkin <[email protected]>Signed-off-by Tiwei Bie <[email protected]>---
v2:
- Refine the wording (Cornelia);v3:
- Refine the wording (MST);content.tex | 7 +++++++ 1 file changed, 7 insertions(+)content.tex +0
-0content.tex which was not offered. The device SHOULD accept any valid subset · line 133
Quoted 9 lines
content.tex +7
-0@@ -125,6 +125,13 @@ which was not offered. The device SHOULD accept any valid subset of features the driver accepts, otherwise it MUST fail to set the FEATURES_OK \field{device status} bit when the driver writes it. +If a device has successfully negotiated a set of features +at least once (by accepting the FEATURES_OK \field{device +status} bit during device initialization), then it SHOULD +NOT fail re-negotiation of the same set of features after +a device or system reset. Failure to do so would interfere +with resuming from suspend and error recovery. +Sorry people but I don't get it. I mean it is kind of reasonable to assume that with a given device and a given driver (given, i.e. nothing changes) the two will always negotiate the same features (including the extremal case where the negotiation fails).
Either the device or a driver rolling a dice to make feature negotiation more fun seems quite unreasonable. So I assume this is not what we are bothering to soft prohibit here.
So the interesting scenario seems to be when stuff changes. When migrating the implementation of the device could change. Or something changes regarding the resources used to provide the virtual device.
Quoted 7 lines
But then, if the device really can not support the set of features it used to be able, I guess the SHOULD does not take effect (I guess that is the difference compared to MUST).
Bottom line is: I tried to figure out what is this about, but I failed. I've read https://github.com/oasis-tcs/virtio-spec/issues/14 too but it did not click. I would appreciate some assistance.
It's exactly what it says. Let's say you negotiated a feature and then device sets
NEED_RESET. Driver must now reset the device and put it back in the same state it had before the reset, then resubmit requests that were available but never used.What if any of the features changed? Device suddenly needs to check for requests which do not match the features.
Quoted 8 lines
Quoted 3 lines
Suspend is similar: guests tend to assume hardware does not change across suspend/resume, any changes tend to make resume fail.
Thank you very much! But it still does not answer why would a device want to do that (fail to negotiate a feature that it was able to negotiate before). So I'm still in the dark about what are we trading for what.
Quoted 5 lines
It would be a mis-configured device. For example QEMU does not migrate the device features so if you misconfigure QEMU with different flags on source and destination (not a supported configuration), features might seem to change from guest POV.
Do you mean set (or rather restrict) what QEMU calls the host_features?
AFAIR there is no reset right after the migration. But yes if then there is a reset and another migration. After a lots of thinking, it seems you speak about the scenario I described in the answer to Tiwei Bie. But there I also say that this statement you add here is not good enough for that. Still puzzled.
Quoted 14 lines
Quoted 14 lines
Is there somewhere a patch that fixes such a bug? Maybe that would help me understand what can be done at the device to avoid the problem.
Regards,
HalilQuoted 6 lines
Quoted 5 lines
Quoted 5 lines
content.tex +0
-0\subsection{Legacy Interface: A Note on Feature Bits}\label{sec:Basic Facilities of a Virtio Device / Feature Bits / Legacy Interface: A Note on Feature Bits}
Quoted 9 lines
Quoted 4 lines
Quoted 4 lines
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
Quoted 98 lines · Halil Pasic
Quoted 88 lines · Tiwei Bie
Quoted 69 lines · Halil Pasic
Quoted 59 lines · Michael S. Tsirkin
Quoted 45 lines · Halil Pasic
Quoted 18 lines · Tiwei Bie
Suggested-by Michael S. Tsirkin <[email protected]>Signed-off-by Tiwei Bie <[email protected]>---
v2:
- Refine the wording (Cornelia);v3:
- Refine the wording (MST);content.tex | 7 +++++++ 1 file changed, 7 insertions(+)content.tex +0
-0content.tex which was not offered. The device SHOULD accept any valid subset · line 133
Quoted 9 lines
content.tex +7
-0@@ -125,6 +125,13 @@ which was not offered. The device SHOULD accept any valid subset of features the driver accepts, otherwise it MUST fail to set the FEATURES_OK \field{device status} bit when the driver writes it. +If a device has successfully negotiated a set of features +at least once (by accepting the FEATURES_OK \field{device +status} bit during device initialization), then it SHOULD +NOT fail re-negotiation of the same set of features after +a device or system reset. Failure to do so would interfere +with resuming from suspend and error recovery. +Sorry people but I don't get it. I mean it is kind of reasonable to assume that with a given device and a given driver (given, i.e. nothing changes) the two will always negotiate the same features (including the extremal case where the negotiation fails).
Either the device or a driver rolling a dice to make feature negotiation more fun seems quite unreasonable. So I assume this is not what we are bothering to soft prohibit here.
So the interesting scenario seems to be when stuff changes. When migrating the implementation of the device could change. Or something changes regarding the resources used to provide the virtual device.
Quoted 7 lines
But then, if the device really can not support the set of features it used to be able, I guess the SHOULD does not take effect (I guess that is the difference compared to MUST).
Bottom line is: I tried to figure out what is this about, but I failed. I've read https://github.com/oasis-tcs/virtio-spec/issues/14 too but it did not click. I would appreciate some assistance.
It's exactly what it says. Let's say you negotiated a feature and then device sets
NEED_RESET. Driver must now reset the device and put it back in the same state it had before the reset, then resubmit requests that were available but never used.Quoted 8 lines
What if any of the features changed? Device suddenly needs to check for requests which do not match the features.
Suspend is similar: guests tend to assume hardware does not change across suspend/resume, any changes tend to make resume fail.
Quoted 4 lines
Thank you very much! But it still does not answer why would a device want to do that (fail to negotiate a feature that it was able to negotiate before). So I'm still in the dark about what are we trading for what.
Hi Halil,
Just like what you said, normally there is no reason for a device to fail to negotiate a feature that it was able to negotiate before. But the spec doesn't forbid devices to do this , i.e. the spec allows a device to fail to negotiate a feature that it was able to negotiate before, which could cause problems in some cases. Although everything works fine in reality because there is no device would really do this, it would be better to make spec to explicitly forbid devices to do this in the necessary cases.
Quoted 3 lines
Best regards,
Tiwei BieI think we have most of it already covered with 'The device SHOULD accept any valid subset of features the driver accepts'.
Quoted 6 lines
IMHO what we add with your proposed normative statement is that if the device used to offer a feature bit it SHOULD keep offering it. That's clearly not covered by the by what I've cited.
But it's kind of covered by a non-normative statement 'Each virtio device offers all the features it understands.'
Well one has to squint very hard to understand it.
And note that "understands" is not the same as "supports". Device can still fail to set FEATURES_OK.
Quoted 17 lines
This seems most relevant in case of migration. That is device implementation S(ource) and device implementation T(arget) are migration compatible. But hey, features that are present
in S and not present in T are of concern for migration compatibility. AFAIK the VIRTIO specification does not make claims about migration compatibility.So if I think QEMU, and somebody (maintainer) is deciding to remove support for of a certain device for a certain feature bit in the next version, he better thinks hard how could this breakmigration. I don't think the proposed normative statement with it's SHOULD would make the the guy more careful.
What is even more interesting is the scenario where the new version of the device does not remove support for a feature, but adds support for one, let's call it
F_N.
Quoted 8 lines
The scenario is the following we have systems O(ld) and N(ew). We start on O then we migrate to new. There some reset of concern happens. Features get re-negotiated and we start exploiting
F_N. In my reading of your addition, this is legit. But then we migrate back from N to O. No re-negotiation happens (because it is not obligatory), and things explode (hopefully, just migration fails, and not guest dies) because O does not have support forF_N. Your normative statement was nowhere violated as far as I can tell.
Oops I shouldn't even have started about migration. Let's forget migration. It's a simple question on what we can assume after we reset device.
Some people want to be able to change features dynamically. Is that OK? This text clarifies that no, it isn't.
Quoted 2 lines
Bottom line is, I still don't know what benefit does this addition to the standard have to the implementer of the standard.
A question was asked. On suspend we save features and try to restore them. Should driver handle device not offering some of these features after resume? What this offers is a simple answer: don't worry about it too much, devices have been warned that it's not a good idea.
Quoted 3 lines
In my opinion
it's just another chunk of text that is hard to figure out. It's hard to tell what is the device
Most people know this I think
Quoted 1 line
and what is before
Sorry before what?
Quoted 1 line
, what is system reset.
I think many people do know what is a system reset. It's an attempt to cover suspend to disk. How would you put it?
Quoted 9 lines
If
we were to make the spec complete with spelling out every 'don't make anything stupid' I'm under the impression there is a lot of work to do. I had a discussion here on the completeness of this spec, and completeness does not seem to be a primary goal. I'm still not sold on this one.Regards,
Halil
Yea, it's just that it's not clear that changing feature bits when device is reset is all that stupid, since it does after all lose its state.
Quoted 15 lines
Quoted 15 lines
Quoted 15 lines
Is there somewhere a patch that fixes such a bug? Maybe that would help me understand what can be done at the device to avoid the problem.
Regards,
HalilQuoted 6 lines
Quoted 5 lines
Quoted 5 lines
content.tex +0
-0\subsection{Legacy Interface: A Note on Feature Bits}\label{sec:Basic Facilities of a Virtio Device / Feature Bits / Legacy Interface: A Note on Feature Bits}
Quoted 6 lines
Quoted 6 lines
Quoted 5 lines
Quoted 4 lines
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
Quoted 85 lines · Halil Pasic
Quoted 76 lines · Michael S. Tsirkin
Quoted 66 lines · Halil Pasic
Quoted 63 lines · Michael S. Tsirkin
Quoted 45 lines · Halil Pasic
Quoted 18 lines · Tiwei Bie
Suggested-by Michael S. Tsirkin <[email protected]>Signed-off-by Tiwei Bie <[email protected]>---
v2:
- Refine the wording (Cornelia);v3:
- Refine the wording (MST);content.tex | 7 +++++++ 1 file changed, 7 insertions(+)content.tex +0
-0content.tex which was not offered. The device SHOULD accept any valid subset · line 133
Quoted 9 lines
content.tex +7
-0@@ -125,6 +125,13 @@ which was not offered. The device SHOULD accept any valid subset of features the driver accepts, otherwise it MUST fail to set the FEATURES_OK \field{device status} bit when the driver writes it. +If a device has successfully negotiated a set of features +at least once (by accepting the FEATURES_OK \field{device +status} bit during device initialization), then it SHOULD +NOT fail re-negotiation of the same set of features after +a device or system reset. Failure to do so would interfere +with resuming from suspend and error recovery. +Sorry people but I don't get it. I mean it is kind of reasonable to assume that with a given device and a given driver (given, i.e. nothing changes) the two will always negotiate the same features (including the extremal case where the negotiation fails).
Either the device or a driver rolling a dice to make feature negotiation more fun seems quite unreasonable. So I assume this is not what we are bothering to soft prohibit here.
So the interesting scenario seems to be when stuff changes. When migrating the implementation of the device could change. Or something changes regarding the resources used to provide the virtual device.
Quoted 7 lines
But then, if the device really can not support the set of features it used to be able, I guess the SHOULD does not take effect (I guess that is the difference compared to MUST).
Bottom line is: I tried to figure out what is this about, but I failed. I've read https://github.com/oasis-tcs/virtio-spec/issues/14 too but it did not click. I would appreciate some assistance.
It's exactly what it says. Let's say you negotiated a feature and then device sets
NEED_RESET. Driver must now reset the device and put it back in the same state it had before the reset, then resubmit requests that were available but never used.What if any of the features changed? Device suddenly needs to check for requests which do not match the features.
Quoted 8 lines
Quoted 3 lines
Suspend is similar: guests tend to assume hardware does not change across suspend/resume, any changes tend to make resume fail.
Thank you very much! But it still does not answer why would a device want to do that (fail to negotiate a feature that it was able to negotiate before). So I'm still in the dark about what are we trading for what.
Quoted 5 lines
It would be a mis-configured device. For example QEMU does not migrate the device features so if you misconfigure QEMU with different flags on source and destination (not a supported configuration), features might seem to change from guest POV.
Quoted 7 lines
Do you mean set (or rather restrict) what QEMU calls the
host_features?AFAIR there is no reset right after the migration. But yes if then there is a reset and another migration. After a lots of thinking, it seems you speak about the scenario I described in the answer to Tiwei Bie. But there I also say that this statement you add here is not good enough for that. Still puzzled.
What would a good enough statement look like?
Quoted 19 lines
Quoted 19 lines
Quoted 18 lines
Is there somewhere a patch that fixes such a bug? Maybe that would help me understand what can be done at the device to avoid the problem.
Regards,
HalilQuoted 10 lines
Quoted 5 lines
content.tex which was not offered. The device SHOULD accept any valid subset
Quoted 5 lines
content.tex +0
-0\subsection{Legacy Interface: A Note on Feature Bits}\label{sec:Basic Facilities of a Virtio Device / Feature Bits / Legacy Interface: A Note on Feature Bits}---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
Quoted 4 lines
Quoted 4 lines
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
Quoted 103 lines · Michael S. Tsirkin
Quoted 102 lines · Halil Pasic
Quoted 88 lines · Tiwei Bie
Quoted 69 lines · Halil Pasic
Quoted 59 lines · Michael S. Tsirkin
Quoted 45 lines · Halil Pasic
Quoted 18 lines · Tiwei Bie
Suggested-by Michael S. Tsirkin <[email protected]>Signed-off-by Tiwei Bie <[email protected]>---
v2:
- Refine the wording (Cornelia);v3:
- Refine the wording (MST);content.tex | 7 +++++++ 1 file changed, 7 insertions(+)content.tex +0
-0content.tex which was not offered. The device SHOULD accept any valid subset · line 133
Quoted 9 lines
content.tex +7
-0@@ -125,6 +125,13 @@ which was not offered. The device SHOULD accept any valid subset of features the driver accepts, otherwise it MUST fail to set the FEATURES_OK \field{device status} bit when the driver writes it. +If a device has successfully negotiated a set of features +at least once (by accepting the FEATURES_OK \field{device +status} bit during device initialization), then it SHOULD +NOT fail re-negotiation of the same set of features after +a device or system reset. Failure to do so would interfere +with resuming from suspend and error recovery. +Sorry people but I don't get it. I mean it is kind of reasonable to assume that with a given device and a given driver (given, i.e. nothing changes) the two will always negotiate the same features (including the extremal case where the negotiation fails).
Either the device or a driver rolling a dice to make feature negotiation more fun seems quite unreasonable. So I assume this is not what we are bothering to soft prohibit here.
So the interesting scenario seems to be when stuff changes. When migrating the implementation of the device could change. Or something changes regarding the resources used to provide the virtual device.
Quoted 7 lines
But then, if the device really can not support the set of features it used to be able, I guess the SHOULD does not take effect (I guess that is the difference compared to MUST).
Bottom line is: I tried to figure out what is this about, but I failed. I've read https://github.com/oasis-tcs/virtio-spec/issues/14 too but it did not click. I would appreciate some assistance.
It's exactly what it says. Let's say you negotiated a feature and then device sets
NEED_RESET. Driver must now reset the device and put it back in the same state it had before the reset, then resubmit requests that were available but never used.Quoted 8 lines
What if any of the features changed? Device suddenly needs to check for requests which do not match the features.
Suspend is similar: guests tend to assume hardware does not change across suspend/resume, any changes tend to make resume fail.
Quoted 4 lines
Thank you very much! But it still does not answer why would a device want to do that (fail to negotiate a feature that it was able to negotiate before). So I'm still in the dark about what are we trading for what.
Hi Halil,
Just like what you said, normally there is no reason for a device to fail to negotiate a feature that it was able to negotiate before. But the spec doesn't forbid devices to do this , i.e. the spec allows a device to fail to negotiate a feature that it was able to negotiate before, which could cause problems in some cases. Although everything works fine in reality because there is no device would really do this, it would be better to make spec to explicitly forbid devices to do this in the necessary cases.
Quoted 3 lines
Best regards,
Tiwei BieI think we have most of it already covered with 'The device SHOULD accept any valid subset of features the driver accepts'.
IMHO what we add with your proposed normative statement is that if the device used to offer a feature bit it SHOULD keep offering it. That's clearly not covered by the by what I've cited.
Quoted 7 lines
Quoted 2 lines
But it's kind of covered by a non-normative statement 'Each virtio device offers all the features it understands.'
Well one has to squint very hard to understand it.
And note that "understands" is not the same as "supports". Device can still fail to setFEATURES_OK.
But I guess it should not. I don't know what is the driver supposed to do in the scenario you describe: The device offered me (the driver) a set of features, I the driver accepted them *all*. The device failed to set FEATURES_OK, because there was *one feature that it "understands" but does not "support". Should I (the driver) start a backtracking feature negotiation to figure out the difference between "understands" and "supports".
Quoted 27 lines
Quoted 18 lines
This seems most relevant in case of migration. That is device implementation S(ource) and device implementation T(arget) are migration compatible. But hey, features that are present
in S and not present in T are of concern for migration compatibility. AFAIK the VIRTIO specification does not make claims about migration compatibility.So if I think QEMU, and somebody (maintainer) is deciding to remove support for of a certain device for a certain feature bit in the next version, he better thinks hard how could this breakmigration. I don't think the proposed normative statement with it's SHOULD would make the the guy more careful.
What is even more interesting is the scenario where the new version of the device does not remove support for a feature, but adds support for one, let's call it
F_N.Quoted 8 lines
The scenario is the following we have systems O(ld) and N(ew). We start on O then we migrate to new. There some reset of concern happens. Features get re-negotiated and we start exploiting
F_N. In my reading of your addition, this is legit. But then we migrate back from N to O. No re-negotiation happens (because it is not obligatory), and things explode (hopefully, just migration fails, and not guest dies) because O does not have support forF_N. Your normative statement was nowhere violated as far as I can tell.
Quoted 7 lines
Oops I shouldn't even have started about migration. Let's forget migration. It's a simple question on what we can assume after we reset device.
Some people want to be able to change features dynamically. Is that OK? This text clarifies that no, it isn't.
That's a very reasonable question, and a straight answer. Yet I think the normative statement is not good enough. In a sense, that it does not say 'it is not OK to change features dynamically'. IMHO to express that we should state something like: 'For a life-time of a virtio device (which transcends device resets) each subsequent feature re-negotiation SHOULD result in the exact same set of features being negotiated as the first successful negotiation.'
In my reading the normative statement discussed here says features are not allowed to 'disappear' dynamically. But does not say a thing about new features 'appearing' dynamically.
About features 'appearing' dynamically, AFAIR there was a virtio-crypto feature that changed the request format (if negotiated). So IFAIR if we were to re-submit the requests unchanged after gaining this feature, we would end up having the problem you described.
However if both add and remove are unsafe, then 'The only way to renegotiate is to reset the device.' is misleading IMHO.
Quoted 9 lines
Quoted 2 lines
Bottom line is, I still don't know what benefit does this addition to the standard have to the implementer of the standard.
A question was asked. On suspend we save features and try to restore them. Should driver handle device not offering some of these features after resume? What this offers is a simple answer: don't worry about it too much, devices have been warned that it's not a good idea.
I don't know enough about suspend/resume. I will try to catch up. But I think I'm slowly starting to understand the problem. My guess is that there is some sort of reset involved in the procedure that could affect what QEMU calls the host_features, but would not affect the requests on the queue.
The questions still remains: Why would the device want to take away a feature? What should the device do (respecting the warning given here) instead of taking away the feature (if the need arises) ?
Quoted 8 lines
Quoted 5 lines
In my opinion
it's just another chunk of text that is hard to figure out. It's hard to tell what is the deviceMost people know this I think
I mean the same device. If I migrate back and forth in the spirit of the normative statement the device is still the same device. When I think QEMU however,we would end up realizing a device each time we spin up a QEMU at the target host. So the life-cycle of the QEMU device and of the virtio device ends up being a different one.
Quoted 4 lines
Quoted 1 line
and what is before
Sorry before what?
My bad. The original text does not use 'before' just 'after'. For some strange reason I started thinking about sequences of re-negotiations and there 'before' slipped in...
Quoted 6 lines
Quoted 2 lines
, what is system reset.
I think many people do know what is a system reset. It's an attempt to cover suspend to disk. How would you put it?
At this point I think I have enough understanding of what is behind, to make a step back, and do the research and the thinking. Thanks for your patience.
But let me carry on with my answer without doing the research for now. Having a notion of system reset and specifying how virio facilities relate to it (affected, unaffected) seems very reasonable. But I think it is a new thing in the spec. I don't think solely adding tihs normative statement is sufficient to achieve that.
Quoted 8 lines
Quoted 8 lines
If
we were to make the spec complete with spelling out every 'don't make anything stupid' I'm under the impression there is a lot of work to do. I had a discussion here on the completeness of this spec, and completeness does not seem to be a primary goal. I'm still not sold on this one.
Quoted 7 lines
Quoted 2 lines
Regards,
HalilYea, it's just that it's not clear that changing feature bits when device is reset is all that stupid, since it does after all lose its state.
My intuition was that this should be a part of describing what a device reset is. It seems the device does not loose all state though -- otherwise I don't understand the problem with the available but not yet used requests.
Anyway many thanks for having this discussion with me. My initial problem was that I could not relate this to anything sane. Now I have to learn more about suspend/resume.
A quick recap at the end. This is about 'Should driver handle device not offering some of these features after resume?' This paragraph is supposed to tell the driver developer don't bother. And I guess it's also supposed to tell the device developer: fail to resume (e.g. migrate) the device if you realize if some features negotiated before can not be supported any more.
Like this if it is a suspend/resume we still end up not being able to resume the device or the whole guest. But at least no funny things will happen if the driver does try to use the feature that went away.
My intuitions is, that handling such feature changes in the guest would be cleaner. The guest has all the information it needs at it's disposal (e.g. are requests in flight, do these depend on some feature that went away or the opposite, can we let the upper layer re-submit the requests and just give up on the ones that stuck available). But I have to the gaps in my understanding before having any.
Regards,
Halil
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
Quoted 88 lines · Michael S. Tsirkin
Quoted 87 lines · Halil Pasic
Quoted 76 lines · Michael S. Tsirkin
Quoted 66 lines · Halil Pasic
Quoted 63 lines · Michael S. Tsirkin
Quoted 45 lines · Halil Pasic
Quoted 18 lines · Tiwei Bie
Suggested-by Michael S. Tsirkin <[email protected]>Signed-off-by Tiwei Bie <[email protected]>---
v2:
- Refine the wording (Cornelia);v3:
- Refine the wording (MST);content.tex | 7 +++++++ 1 file changed, 7 insertions(+)content.tex +0
-0content.tex which was not offered. The device SHOULD accept any valid subset · line 133
Quoted 9 lines
content.tex +7
-0@@ -125,6 +125,13 @@ which was not offered. The device SHOULD accept any valid subset of features the driver accepts, otherwise it MUST fail to set the FEATURES_OK \field{device status} bit when the driver writes it. +If a device has successfully negotiated a set of features +at least once (by accepting the FEATURES_OK \field{device +status} bit during device initialization), then it SHOULD +NOT fail re-negotiation of the same set of features after +a device or system reset. Failure to do so would interfere +with resuming from suspend and error recovery. +Sorry people but I don't get it. I mean it is kind of reasonable to assume that with a given device and a given driver (given, i.e. nothing changes) the two will always negotiate the same features (including the extremal case where the negotiation fails).
Either the device or a driver rolling a dice to make feature negotiation more fun seems quite unreasonable. So I assume this is not what we are bothering to soft prohibit here.
So the interesting scenario seems to be when stuff changes. When migrating the implementation of the device could change. Or something changes regarding the resources used to provide the virtual device.
Quoted 7 lines
But then, if the device really can not support the set of features it used to be able, I guess the SHOULD does not take effect (I guess that is the difference compared to MUST).
Bottom line is: I tried to figure out what is this about, but I failed. I've read https://github.com/oasis-tcs/virtio-spec/issues/14 too but it did not click. I would appreciate some assistance.
It's exactly what it says. Let's say you negotiated a feature and then device sets
NEED_RESET. Driver must now reset the device and put it back in the same state it had before the reset, then resubmit requests that were available but never used.What if any of the features changed? Device suddenly needs to check for requests which do not match the features.
Quoted 8 lines
Quoted 3 lines
Suspend is similar: guests tend to assume hardware does not change across suspend/resume, any changes tend to make resume fail.
Thank you very much! But it still does not answer why would a device want to do that (fail to negotiate a feature that it was able to negotiate before). So I'm still in the dark about what are we trading for what.
Quoted 5 lines
It would be a mis-configured device. For example QEMU does not migrate the device features so if you misconfigure QEMU with different flags on source and destination (not a supported configuration), features might seem to change from guest POV.
Do you mean set (or rather restrict) what QEMU calls the
host_features?
Quoted 9 lines
Quoted 5 lines
AFAIR there is no reset right after the migration. But yes if then there is a reset and another migration. After a lots of thinking, it seems you speak about the scenario I described in the answer to Tiwei Bie. But there I also say that this statement you add here is not good enough for that. Still puzzled.
What would a good enough statement look like?
I did some reading and some thinking on the weekend. AFAIU the situation
is tricky. To mitigate that let me establish the terminology I'm going to
use. For vm lifecycle I'm going to use the definitions form libvirt as
defined by https://libvirt.org/guide/html/Application_Development_Guide-Guest_Domains-Lifecycle.html.
You explained, the motivation for this addition to the VIRTIO specification is hibernate (aka suspend to disk).
(1) AFAIU on hibernate the VM goes from 'running' to (most likely) 'defined'. The first step of the resume from hibernate is to start the VM. From the guest OS life-cycle perspective however we don't start a completely new cycle (like the VM life-cycle does) with complete re-initialization. After resuming form hibernate the system is expected to be put in essentially the same state (but not exactly) as it was before hibernate.
(2) From VM (life-cycle) perspective we can not distinguish between a 'shutdown' as a part of a hibernate and a 'plain shutdown'.
(3) Any rule we come up for a device (e.g. the normative statement proposed here) that regulates the effects of a 'system reset' that is a part of the hibernate cycle equally affects the normal shutdown-start cycle.
(4) Any change in the negotiated feature set can affect the validity of requests that have been constructed under different assumptions (i.e. not only features going away, but also features appearing can be a problem).
(5) The Linux implementation already has reasonable handling for both types of changes: the driver does not try to use the new features, and fails cleanly if the old ones are not accepted.
(6) Because of (3), prohibiting devices dropping support for a set of features within a hibernate cycle is only possible if we prohibit such changes in general.
(7) If I read
https://www.kernel.org/doc/html/v4.14/driver-api/pm/devices.html
correctly the driver is expected to quiesce the device before going to hibernate. AFAIU hibernating with requests in flight isn't a great idea.
(8) If there are no in-flight requests in-flight (including on the queues), then this whole feature changes might break requests story seems irrelevant.
(9) After a quick look the freeze in virtio (Linux implementation) does not seem to do anything to prevent in-flight requests though.
(10) From a VM management perspective a 'save' seems much preferable to hibernate. A VM 'save' is migration like, so even if some components of the system change between 'save' and 'restore' (e.g. QEMU up- or donwngarde) we still have mechanisms in place that (hopefully) the guest view of the system does not change. In this sense save/restore is migration like.
(11) The VIRTIO specification is a bit vague about how a reset is supposed to be handled by the guest, but it certainly does not prohibit the negotiated features from changing after reset. Here I will quote two fragments that hint this is actually something foreseen by the VIRTIO standard:
* 'During device initialization, the driver reads this and tells the device the subset that it accepts. The only way to renegotiate is to reset the device.'
* 'If the driver sets the FAILED bit, the driver MUST later reset the device before attempting to re-initialize.' If re-initialize is in a sense of '3.1.1 Driver Requirements: Device Initialization' then full feature negotiation seems to be compulsory. Linux does not do this. But since setting up queues seems to be a part of the 3.1.1 initialization sequence (even if formulated somewhat vague), my best guess after reset the driver is not supposed to perform 3.1.1 to the letter.
(12) If I were to hibernate my PC and then, let's say replace my NIC with a different model, the hardware does not change assumption would not hold for a non-virtualized system either. I'm not sure this problem is ours to solve.
My conclusion is the following. I think constraining feature changes after system_reset is a bad idea. For 'normal' virtio reset some clarifications would be welcome, but this one does not seem to be a very good one. Regarding changing features, I think we are good enough with what we have today (both standard and implementation). However if we want to prohibit the features from changing after a reset in spite of my arguments presented here, IMHO we need a driver normative statement too.
Regards,
Halil
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
Quoted 158 lines · Halil Pasic
Quoted 88 lines · Michael S. Tsirkin
Quoted 87 lines · Halil Pasic
Quoted 76 lines · Michael S. Tsirkin
Quoted 66 lines · Halil Pasic
Quoted 63 lines · Michael S. Tsirkin
Quoted 45 lines · Halil Pasic
Quoted 18 lines · Tiwei Bie
Suggested-by Michael S. Tsirkin <[email protected]>Signed-off-by Tiwei Bie <[email protected]>---
v2:
- Refine the wording (Cornelia);v3:
- Refine the wording (MST);content.tex | 7 +++++++ 1 file changed, 7 insertions(+)content.tex +0
-0content.tex which was not offered. The device SHOULD accept any valid subset · line 133
Quoted 9 lines
content.tex +7
-0@@ -125,6 +125,13 @@ which was not offered. The device SHOULD accept any valid subset of features the driver accepts, otherwise it MUST fail to set the FEATURES_OK \field{device status} bit when the driver writes it. +If a device has successfully negotiated a set of features +at least once (by accepting the FEATURES_OK \field{device +status} bit during device initialization), then it SHOULD +NOT fail re-negotiation of the same set of features after +a device or system reset. Failure to do so would interfere +with resuming from suspend and error recovery. +Sorry people but I don't get it. I mean it is kind of reasonable to assume that with a given device and a given driver (given, i.e. nothing changes) the two will always negotiate the same features (including the extremal case where the negotiation fails).
Either the device or a driver rolling a dice to make feature negotiation more fun seems quite unreasonable. So I assume this is not what we are bothering to soft prohibit here.
So the interesting scenario seems to be when stuff changes. When migrating the implementation of the device could change. Or something changes regarding the resources used to provide the virtual device.
Quoted 7 lines
But then, if the device really can not support the set of features it used to be able, I guess the SHOULD does not take effect (I guess that is the difference compared to MUST).
Bottom line is: I tried to figure out what is this about, but I failed. I've read https://github.com/oasis-tcs/virtio-spec/issues/14 too but it did not click. I would appreciate some assistance.
It's exactly what it says. Let's say you negotiated a feature and then device sets
NEED_RESET. Driver must now reset the device and put it back in the same state it had before the reset, then resubmit requests that were available but never used.What if any of the features changed? Device suddenly needs to check for requests which do not match the features.
Quoted 8 lines
Quoted 3 lines
Suspend is similar: guests tend to assume hardware does not change across suspend/resume, any changes tend to make resume fail.
Thank you very much! But it still does not answer why would a device want to do that (fail to negotiate a feature that it was able to negotiate before). So I'm still in the dark about what are we trading for what.
Quoted 5 lines
It would be a mis-configured device. For example QEMU does not migrate the device features so if you misconfigure QEMU with different flags on source and destination (not a supported configuration), features might seem to change from guest POV.
Do you mean set (or rather restrict) what QEMU calls the
host_features?Quoted 9 lines
Quoted 5 lines
AFAIR there is no reset right after the migration. But yes if then there is a reset and another migration. After a lots of thinking, it seems you speak about the scenario I described in the answer to Tiwei Bie. But there I also say that this statement you add here is not good enough for that. Still puzzled.
What would a good enough statement look like?
I did some reading and some thinking on the weekend. AFAIU the situation
is tricky. To mitigate that let me establish the terminology I'm going to
use. For vm lifecycle I'm going to use the definitions form libvirt as
defined by https://libvirt.org/guide/html/Application_Development_Guide-Guest_Domains-Lifecycle.html.You explained, the motivation for this addition to the VIRTIO specification is hibernate (aka suspend to disk).
(1) AFAIU on hibernate the VM goes from 'running' to (most likely) 'defined'. The first step of the resume from hibernate is to start the VM. From the guest OS life-cycle perspective however we don't start a completely new cycle (like the VM life-cycle does) with complete re-initialization. After resuming form hibernate the system is expected to be put in essentially the same state (but not exactly) as it was before hibernate.
(2) From VM (life-cycle) perspective we can not distinguish between a 'shutdown' as a part of a hibernate and a 'plain shutdown'.
(3) Any rule we come up for a device (e.g. the normative statement proposed here) that regulates the effects of a 'system reset' that is a part of the hibernate cycle equally affects the normal shutdown-start cycle.
(4) Any change in the negotiated feature set can affect the validity of requests that have been constructed under different assumptions (i.e. not only features going away, but also features appearing can be a problem).
(5) The Linux implementation already has reasonable handling for both types of changes: the driver does not try to use the new features, and fails cleanly if the old ones are not accepted.
(6) Because of (3), prohibiting devices dropping support for a set of features within a hibernate cycle is only possible if we prohibit such changes in general.
(7) If I read
https://www.kernel.org/doc/html/v4.14/driver-api/pm/devices.html
correctly the driver is expected to quiesce the device before going to hibernate. AFAIU hibernating with requests in flight isn't a great idea.(8) If there are no in-flight requests in-flight (including on the queues), then this whole feature changes might break requests story seems irrelevant.
(9) After a quick look the freeze in virtio (Linux implementation) does not seem to do anything to prevent in-flight requests though.
(10) From a VM management perspective a 'save' seems much preferable to hibernate. A VM 'save' is migration like, so even if some components of the system change between 'save' and 'restore' (e.g. QEMU up- or donwngarde) we still have mechanisms in place that (hopefully) the guest view of the system does not change. In this sense save/restore is migration like.
Quoted 15 lines
(11) The VIRTIO specification is a bit vague about how a reset is supposed to be handled by the guest, but it certainly does not prohibit the negotiated features from changing after reset. Here I will quote two fragments that hint this is actually something foreseen by the VIRTIO standard:
* 'During device initialization, the driver reads this and tells the device the subset that it accepts. The only way to renegotiate is to reset the device.'
* 'If the driver sets the FAILED bit, the driver MUST later reset the device before attempting to re-initialize.' If re-initialize is in a sense of '3.1.1 Driver Requirements: Device Initialization' then full feature negotiation seems to be compulsory. Linux does not do this. But since setting up queues seems to be a part of the 3.1.1 initialization sequence (even if formulated somewhat vague), my best guess after reset the driver is not supposed to perform 3.1.1 to the letter.
I think frankly if we want dynamic features we should work on a mechanism that allows changing them without a system reset.
And I think the use-case that triggered this is the SRIOV feature, take a look at how that is handled across e.g. suspend/resume.
Quoted 5 lines
(12) If I were to hibernate my PC and then, let's say replace my NIC with a different model, the hardware does not change assumption would not hold for a non-virtualized system either. I'm not sure this problem is ours to solve.
Precisely and since we can't solve it, we warn people not to create this kind of configuration unless they know exactly what they are doing.
Quoted 10 lines
My conclusion is the following. I think constraining feature changes after
system_resetis a bad idea. For 'normal' virtio reset some clarifications would be welcome, but this one does not seem to be a very good one. Regarding changing features, I think we are good enough with what we have today (both standard and implementation). However if we want to prohibit the features from changing after a reset in spite of my arguments presented here, IMHO we need a driver normative statement too.Regards,
Halil
Well the motion passed with 1 abstain and 5 in favor. Tiwei was the one who proposed it so as I already did this in the past, I'll wait a day or two for him to respond and let us know whether he'd like to drop the patch, but in absence of such a response I'll have to push the proposed wording.
In that case you will need to put in a motion to revert, or make some other change on top.
Signature · 5 lines
MST --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
Quoted 202 lines · Michael S. Tsirkin
Quoted 158 lines · Halil Pasic
Quoted 88 lines · Michael S. Tsirkin
Quoted 87 lines · Halil Pasic
Quoted 76 lines · Michael S. Tsirkin
Quoted 66 lines · Halil Pasic
Quoted 63 lines · Michael S. Tsirkin
Quoted 45 lines · Halil Pasic
Quoted 18 lines · Tiwei Bie
Suggested-by Michael S. Tsirkin <[email protected]>Signed-off-by Tiwei Bie <[email protected]>---
v2:
- Refine the wording (Cornelia);v3:
- Refine the wording (MST);content.tex | 7 +++++++ 1 file changed, 7 insertions(+)content.tex +0
-0content.tex which was not offered. The device SHOULD accept any valid subset · line 133
Quoted 9 lines
content.tex +7
-0@@ -125,6 +125,13 @@ which was not offered. The device SHOULD accept any valid subset of features the driver accepts, otherwise it MUST fail to set the FEATURES_OK \field{device status} bit when the driver writes it. +If a device has successfully negotiated a set of features +at least once (by accepting the FEATURES_OK \field{device +status} bit during device initialization), then it SHOULD +NOT fail re-negotiation of the same set of features after +a device or system reset. Failure to do so would interfere +with resuming from suspend and error recovery. +Sorry people but I don't get it. I mean it is kind of reasonable to assume that with a given device and a given driver (given, i.e. nothing changes) the two will always negotiate the same features (including the extremal case where the negotiation fails).
Either the device or a driver rolling a dice to make feature negotiation more fun seems quite unreasonable. So I assume this is not what we are bothering to soft prohibit here.
So the interesting scenario seems to be when stuff changes. When migrating the implementation of the device could change. Or something changes regarding the resources used to provide the virtual device.
Quoted 7 lines
But then, if the device really can not support the set of features it used to be able, I guess the SHOULD does not take effect (I guess that is the difference compared to MUST).
Bottom line is: I tried to figure out what is this about, but I failed. I've read https://github.com/oasis-tcs/virtio-spec/issues/14 too but it did not click. I would appreciate some assistance.
It's exactly what it says. Let's say you negotiated a feature and then device sets
NEED_RESET. Driver must now reset the device and put it back in the same state it had before the reset, then resubmit requests that were available but never used.What if any of the features changed? Device suddenly needs to check for requests which do not match the features.
Quoted 8 lines
Quoted 3 lines
Suspend is similar: guests tend to assume hardware does not change across suspend/resume, any changes tend to make resume fail.
Thank you very much! But it still does not answer why would a device want to do that (fail to negotiate a feature that it was able to negotiate before). So I'm still in the dark about what are we trading for what.
Quoted 5 lines
It would be a mis-configured device. For example QEMU does not migrate the device features so if you misconfigure QEMU with different flags on source and destination (not a supported configuration), features might seem to change from guest POV.
Do you mean set (or rather restrict) what QEMU calls the
host_features?Quoted 9 lines
Quoted 5 lines
AFAIR there is no reset right after the migration. But yes if then there is a reset and another migration. After a lots of thinking, it seems you speak about the scenario I described in the answer to Tiwei Bie. But there I also say that this statement you add here is not good enough for that. Still puzzled.
What would a good enough statement look like?
I did some reading and some thinking on the weekend. AFAIU the situation
is tricky. To mitigate that let me establish the terminology I'm going to
use. For vm lifecycle I'm going to use the definitions form libvirt as
defined by https://libvirt.org/guide/html/Application_Development_Guide-Guest_Domains-Lifecycle.html.You explained, the motivation for this addition to the VIRTIO specification is hibernate (aka suspend to disk).
(1) AFAIU on hibernate the VM goes from 'running' to (most likely) 'defined'. The first step of the resume from hibernate is to start the VM. From the guest OS life-cycle perspective however we don't start a completely new cycle (like the VM life-cycle does) with complete re-initialization. After resuming form hibernate the system is expected to be put in essentially the same state (but not exactly) as it was before hibernate.
(2) From VM (life-cycle) perspective we can not distinguish between a 'shutdown' as a part of a hibernate and a 'plain shutdown'.
(3) Any rule we come up for a device (e.g. the normative statement proposed here) that regulates the effects of a 'system reset' that is a part of the hibernate cycle equally affects the normal shutdown-start cycle.
(4) Any change in the negotiated feature set can affect the validity of requests that have been constructed under different assumptions (i.e. not only features going away, but also features appearing can be a problem).
(5) The Linux implementation already has reasonable handling for both types of changes: the driver does not try to use the new features, and fails cleanly if the old ones are not accepted.
(6) Because of (3), prohibiting devices dropping support for a set of features within a hibernate cycle is only possible if we prohibit such changes in general.
(7) If I read
https://www.kernel.org/doc/html/v4.14/driver-api/pm/devices.html
correctly the driver is expected to quiesce the device before going to hibernate. AFAIU hibernating with requests in flight isn't a great idea.(8) If there are no in-flight requests in-flight (including on the queues), then this whole feature changes might break requests story seems irrelevant.
(9) After a quick look the freeze in virtio (Linux implementation) does not seem to do anything to prevent in-flight requests though.
(10) From a VM management perspective a 'save' seems much preferable to hibernate. A VM 'save' is migration like, so even if some components of the system change between 'save' and 'restore' (e.g. QEMU up- or donwngarde) we still have mechanisms in place that (hopefully) the guest view of the system does not change. In this sense save/restore is migration like.
Quoted 15 lines
(11) The VIRTIO specification is a bit vague about how a reset is supposed to be handled by the guest, but it certainly does not prohibit the negotiated features from changing after reset. Here I will quote two fragments that hint this is actually something foreseen by the VIRTIO standard:
* 'During device initialization, the driver reads this and tells the device the subset that it accepts. The only way to renegotiate is to reset the device.'
* 'If the driver sets the FAILED bit, the driver MUST later reset the device before attempting to re-initialize.' If re-initialize is in a sense of '3.1.1 Driver Requirements: Device Initialization' then full feature negotiation seems to be compulsory. Linux does not do this. But since setting up queues seems to be a part of the 3.1.1 initialization sequence (even if formulated somewhat vague), my best guess after reset the driver is not supposed to perform 3.1.1 to the letter.I think frankly if we want dynamic features we should work on a mechanism that allows changing them without a system reset.
And I think the use-case that triggered this is the SRIOV feature, take a look at how that is handled across e.g. suspend/resume.
Quoted 5 lines
(12) If I were to hibernate my PC and then, let's say replace my NIC with a different model, the hardware does not change assumption would not hold for a non-virtualized system either. I'm not sure this problem is ours to solve.
Precisely and since we can't solve it, we warn people not to create this kind of configuration unless they know exactly what they are doing.
Quoted 10 lines
My conclusion is the following. I think constraining feature changes after
system_resetis a bad idea. For 'normal' virtio reset some clarifications would be welcome, but this one does not seem to be a very good one. Regarding changing features, I think we are good enough with what we have today (both standard and implementation). However if we want to prohibit the features from changing after a reset in spite of my arguments presented here, IMHO we need a driver normative statement too.Regards,
Halil
Quoted 8 lines
Well the motion passed with 1 abstain and 5 in favor. Tiwei was the one who proposed it so as I already did this in the past, I'll wait a day or two for him to respond and let us know whether he'd like to drop the patch, but in absence of such a response I'll have to push the proposed wording.
In that case you will need to put in a motion to revert, or make some other change on top.
If it would be better to drop this patch,
I'm fine with dropping it. Thanks!
Best regards,
Tiwei Bie
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
Quoted 1 line · Tiwei Bie
On Mon, Jun 18, 2018 at 07:28:33PM +0300, Michael S. Tsirkin wrote:
[..]
Quoted 17 lines
Quoted 17 lines
Quoted 16 lines
(11) The VIRTIO specification is a bit vague about how a reset is supposed to be handled by the guest, but it certainly does not prohibit the negotiated features from changing after reset. Here I will quote two fragments that hint this is actually something foreseen by the VIRTIO standard:
* 'During device initialization, the driver reads this and tells the device the subset that it accepts. The only way to renegotiate is to reset the device.'
* 'If the driver sets the FAILED bit, the driver MUST later reset the device before attempting to re-initialize.' If re-initialize is in a sense of '3.1.1 Driver Requirements: Device Initialization' then full feature negotiation seems to be compulsory. Linux does not do this. But since setting up queues seems to be a part of the 3.1.1 initialization sequence (even if formulated somewhat vague), my best guess after reset the driver is not supposed to perform 3.1.1 to the letter.
Quoted 3 lines
Quoted 3 lines
I think frankly if we want dynamic features we should work on a mechanism that allows changing them without a system reset.
@Michael
I was talking abut normal virtio reset in (11). I think in Linux we have dynamic features without system reset today if a virtio device driver that is loaded as module gets replaced (e.g. rmmod/insmod new) with a more capable implementation of the same device driver.
Quoted 4 lines
Quoted 4 lines
And I think the use-case that triggered this is the SRIOV feature, take a look at how that is handled across e.g. suspend/resume.
Quoted 1 line
Quoted 8 lines
Quoted 8 lines
Quoted 4 lines
(12) If I were to hibernate my PC and then, let's say replace my NIC with a different model, the hardware does not change assumption would not hold for a non-virtualized system either. I'm not sure this problem is ours to solve.
Precisely and since we can't solve it, we warn people not to create this kind of configuration unless they know exactly what they are doing.
@Michael
I assume the various bus specifications don't bother to spell this out, and I doubt manuals of HW components do either.
If our main goal is to warn the end user to not fiddle with the features of a hibernated VM (e.g. via libvirt domain xml), and hint that if the guest is going to get hibernated, he should better configure guest as migratable even if it's not (e.g. machine type, cpu model should not be moving target) I doubt the VIRTIO spec is the right place.
IMHO neither QEMU nor KVM can detect the condition in question, and I don't think higher level management software can help either. That's why I say end-user.
Hibernate is IMHO an OS concept, and I guess some OSes don't have the concept of hibernate. I see support for hibernate out of scope for the VIRTIO spec (much like migration). But since the VIRTIO spec is supposed to be helpful above all, I'm not opposed to a note that spells the warning out.
I still oppose a device normative, as this does not seem to be something an implementer of the device should heed. And if we do want to place a note, it needs to be more direct. I could not figure out what is this about. I doubt end-users have better chances to.
Quoted 12 lines
Quoted 12 lines
Quoted 11 lines
My conclusion is the following. I think constraining feature changes after
system_resetis a bad idea. For 'normal' virtio reset some clarifications would be welcome, but this one does not seem to be a very good one. Regarding changing features, I think we are good enough with what we have today (both standard and implementation). However if we want to prohibit the features from changing after a reset in spite of my arguments presented here, IMHO we need a driver normative statement too.Regards,
Halil
Quoted 8 lines
Quoted 8 lines
Well the motion passed with 1 abstain and 5 in favor. Tiwei was the one who proposed it so as I already did this in the past, I'll wait a day or two for him to respond and let us know whether he'd like to drop the patch, but in absence of such a response I'll have to push the proposed wording.
In that case you will need to put in a motion to revert, or make some other change on top.
@Michael
If I can not convince you, nor at least some of the committee people here I'm not willing to escalate this as a motion to revert. There is no point, as I'm running out of arguments. While I'm still not convinced that this is the way to go, I'm willing to bow my head in front of the opinion of the majority. It is not like including this would have tragic consequences. I think mustered a fair effort to form an opinion and defend it. Thus there is no shame in admitting defeat.
Quoted 4 lines
If it would be better to drop this patch,
I'm fine with dropping it. Thanks!
@Tiwei Bie
Thanks for your flexibility! What is your opinion (after considering the arguments from my previous mail), is it better to include this patch in the spec or is it better to drop it? Were you able to identify mistakes in my reasoning (I mean points (1)-(12))?
Regards,
Halil
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
Quoted 2 lines · Halil Pasic
Quoted 1 line · Tiwei Bie
On Mon, Jun 18, 2018 at 07:28:33PM +0300, Michael S. Tsirkin wrote:
[...]
Quoted 11 lines
Quoted 4 lines
If it would be better to drop this patch,
I'm fine with dropping it. Thanks!@Tiwei Bie
Thanks for your flexibility! What is your opinion (after considering the arguments from my previous mail), is it better to include this patch in the spec or is it better to drop it? Were you able to identify mistakes in my reasoning (I mean points (1)-(12))?
Hi Halil,
I think maybe you thought too much about this proposal (or maybe I really missed something obvious). In my opinion, the device requirement proposed by this patch is quite simple and straightforward:
- It's just to make the spec explicitly require that a certain virtio device shouldn't fail re-negotiation of a feature set it has successfully accepted once.
- It covers the cases of virtio device reset and system reset (which includes normal shutdown and start).
I think the requirement is reasonable because for a certain virtio device, there is no reason that the feature bits it offers will change (because it should always offer all the features it understands). And we are just to add a device normative to make the spec be more explicit about that (because if a device really changes the features it offers after a device or system reset, something will go wrong). If the configs of an emulated virtio device are changed, maybe we shouldn't treat it as the same device any more, and IMO this case is not related to this proposal.
Although we have 'Each virtio device offers all the features it understands', it's not an explicit device requirement. So I don't think it's a bad idea to have an explicit device requirement about this.
Best regards,
Tiwei Bie
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
Quoted 212 lines · Tiwei Bie
Quoted 202 lines · Michael S. Tsirkin
Quoted 158 lines · Halil Pasic
Quoted 88 lines · Michael S. Tsirkin
Quoted 87 lines · Halil Pasic
Quoted 76 lines · Michael S. Tsirkin
Quoted 66 lines · Halil Pasic
Quoted 63 lines · Michael S. Tsirkin
Quoted 45 lines · Halil Pasic
Quoted 18 lines · Tiwei Bie
Suggested-by Michael S. Tsirkin <[email protected]>Signed-off-by Tiwei Bie <[email protected]>---
v2:
- Refine the wording (Cornelia);v3:
- Refine the wording (MST);content.tex | 7 +++++++ 1 file changed, 7 insertions(+)content.tex +0
-0content.tex which was not offered. The device SHOULD accept any valid subset · line 133
Quoted 9 lines
content.tex +7
-0@@ -125,6 +125,13 @@ which was not offered. The device SHOULD accept any valid subset of features the driver accepts, otherwise it MUST fail to set the FEATURES_OK \field{device status} bit when the driver writes it. +If a device has successfully negotiated a set of features +at least once (by accepting the FEATURES_OK \field{device +status} bit during device initialization), then it SHOULD +NOT fail re-negotiation of the same set of features after +a device or system reset. Failure to do so would interfere +with resuming from suspend and error recovery. +Sorry people but I don't get it. I mean it is kind of reasonable to assume that with a given device and a given driver (given, i.e. nothing changes) the two will always negotiate the same features (including the extremal case where the negotiation fails).
Either the device or a driver rolling a dice to make feature negotiation more fun seems quite unreasonable. So I assume this is not what we are bothering to soft prohibit here.
So the interesting scenario seems to be when stuff changes. When migrating the implementation of the device could change. Or something changes regarding the resources used to provide the virtual device.
Quoted 7 lines
But then, if the device really can not support the set of features it used to be able, I guess the SHOULD does not take effect (I guess that is the difference compared to MUST).
Bottom line is: I tried to figure out what is this about, but I failed. I've read https://github.com/oasis-tcs/virtio-spec/issues/14 too but it did not click. I would appreciate some assistance.
It's exactly what it says. Let's say you negotiated a feature and then device sets
NEED_RESET. Driver must now reset the device and put it back in the same state it had before the reset, then resubmit requests that were available but never used.What if any of the features changed? Device suddenly needs to check for requests which do not match the features.
Quoted 8 lines
Quoted 3 lines
Suspend is similar: guests tend to assume hardware does not change across suspend/resume, any changes tend to make resume fail.
Thank you very much! But it still does not answer why would a device want to do that (fail to negotiate a feature that it was able to negotiate before). So I'm still in the dark about what are we trading for what.
Quoted 5 lines
It would be a mis-configured device. For example QEMU does not migrate the device features so if you misconfigure QEMU with different flags on source and destination (not a supported configuration), features might seem to change from guest POV.
Do you mean set (or rather restrict) what QEMU calls the
host_features?Quoted 9 lines
Quoted 5 lines
AFAIR there is no reset right after the migration. But yes if then there is a reset and another migration. After a lots of thinking, it seems you speak about the scenario I described in the answer to Tiwei Bie. But there I also say that this statement you add here is not good enough for that. Still puzzled.
What would a good enough statement look like?
I did some reading and some thinking on the weekend. AFAIU the situation
is tricky. To mitigate that let me establish the terminology I'm going to
use. For vm lifecycle I'm going to use the definitions form libvirt as
defined by https://libvirt.org/guide/html/Application_Development_Guide-Guest_Domains-Lifecycle.html.You explained, the motivation for this addition to the VIRTIO specification is hibernate (aka suspend to disk).
(1) AFAIU on hibernate the VM goes from 'running' to (most likely) 'defined'. The first step of the resume from hibernate is to start the VM. From the guest OS life-cycle perspective however we don't start a completely new cycle (like the VM life-cycle does) with complete re-initialization. After resuming form hibernate the system is expected to be put in essentially the same state (but not exactly) as it was before hibernate.
(2) From VM (life-cycle) perspective we can not distinguish between a 'shutdown' as a part of a hibernate and a 'plain shutdown'.
(3) Any rule we come up for a device (e.g. the normative statement proposed here) that regulates the effects of a 'system reset' that is a part of the hibernate cycle equally affects the normal shutdown-start cycle.
(4) Any change in the negotiated feature set can affect the validity of requests that have been constructed under different assumptions (i.e. not only features going away, but also features appearing can be a problem).
(5) The Linux implementation already has reasonable handling for both types of changes: the driver does not try to use the new features, and fails cleanly if the old ones are not accepted.
(6) Because of (3), prohibiting devices dropping support for a set of features within a hibernate cycle is only possible if we prohibit such changes in general.
(7) If I read
https://www.kernel.org/doc/html/v4.14/driver-api/pm/devices.html
correctly the driver is expected to quiesce the device before going to hibernate. AFAIU hibernating with requests in flight isn't a great idea.(8) If there are no in-flight requests in-flight (including on the queues), then this whole feature changes might break requests story seems irrelevant.
(9) After a quick look the freeze in virtio (Linux implementation) does not seem to do anything to prevent in-flight requests though.
(10) From a VM management perspective a 'save' seems much preferable to hibernate. A VM 'save' is migration like, so even if some components of the system change between 'save' and 'restore' (e.g. QEMU up- or donwngarde) we still have mechanisms in place that (hopefully) the guest view of the system does not change. In this sense save/restore is migration like.
Quoted 15 lines
(11) The VIRTIO specification is a bit vague about how a reset is supposed to be handled by the guest, but it certainly does not prohibit the negotiated features from changing after reset. Here I will quote two fragments that hint this is actually something foreseen by the VIRTIO standard:
* 'During device initialization, the driver reads this and tells the device the subset that it accepts. The only way to renegotiate is to reset the device.'
* 'If the driver sets the FAILED bit, the driver MUST later reset the device before attempting to re-initialize.' If re-initialize is in a sense of '3.1.1 Driver Requirements: Device Initialization' then full feature negotiation seems to be compulsory. Linux does not do this. But since setting up queues seems to be a part of the 3.1.1 initialization sequence (even if formulated somewhat vague), my best guess after reset the driver is not supposed to perform 3.1.1 to the letter.I think frankly if we want dynamic features we should work on a mechanism that allows changing them without a system reset.
And I think the use-case that triggered this is the SRIOV feature, take a look at how that is handled across e.g. suspend/resume.
Quoted 5 lines
(12) If I were to hibernate my PC and then, let's say replace my NIC with a different model, the hardware does not change assumption would not hold for a non-virtualized system either. I'm not sure this problem is ours to solve.
Precisely and since we can't solve it, we warn people not to create this kind of configuration unless they know exactly what they are doing.
Quoted 10 lines
My conclusion is the following. I think constraining feature changes after
system_resetis a bad idea. For 'normal' virtio reset some clarifications would be welcome, but this one does not seem to be a very good one. Regarding changing features, I think we are good enough with what we have today (both standard and implementation). However if we want to prohibit the features from changing after a reset in spite of my arguments presented here, IMHO we need a driver normative statement too.Regards,
HalilQuoted 8 lines
Well the motion passed with 1 abstain and 5 in favor. Tiwei was the one who proposed it so as I already did this in the past, I'll wait a day or two for him to respond and let us know whether he'd like to drop the patch, but in absence of such a response I'll have to push the proposed wording.
In that case you will need to put in a motion to revert, or make some other change on top.
Quoted 5 lines
If it would be better to drop this patch,
I'm fine with dropping it. Thanks!Best regards,
Tiwei Bie
Tiwei, Halil, I suspect you are just trying to be nice, but it can't really work like this - we have a process to follow. Objections were raised during the voting process. The TC still voted unanimously for the proposal
to resolve your comments to the spec, with one abstention.
If the person who proposed the change, no longer thinks it's a good idea, and requests that we postpone making the change, maybe it's best to give everyone chance to re-consider and I am not going to stand on formality and require re-voting on postponing making the change.
Anything else we'd have to re-vote.
But if you request that, please be explicit and say so. So if you want one of these, you could say for example "I withdraw my comments on the specification, the issue does not need to be addressed. Please cancel the change proposal." or "While my comments on the specification stand, I withdraw my proposal to resolve the issue, please cancel the change proposal". Or if you want the TC to decide, then you'd say e.g. "I leave that decision to the TC".
We'd then either follow a previous voting decision or re-vote if someone requests that.
Signature · 5 lines
MST --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
Quoted 42 lines · On Wed, 20 Jun 2018 00:30:59 +0800 Tiwei Bie <[email protected]> wrote:
Quoted 2 lines · Halil Pasic
Quoted 1 line · Tiwei Bie
On Mon, Jun 18, 2018 at 07:28:33PM +0300, Michael S. Tsirkin wrote:
[...]
Quoted 11 lines
Quoted 4 lines
If it would be better to drop this patch,
I'm fine with dropping it. Thanks!@Tiwei Bie
Thanks for your flexibility! What is your opinion (after considering the arguments from my previous mail), is it better to include this patch in the spec or is it better to drop it? Were you able to identify mistakes in my reasoning (I mean points (1)-(12))?Hi Halil,
I think maybe you thought too much about this proposal (or maybe I really missed something obvious). In my opinion, the device requirement proposed by this patch is quite simple and straightforward:
- It's just to make the spec explicitly require that a certain virtio device shouldn't fail re-negotiation of a feature set it has successfully accepted once.
- It covers the cases of virtio device reset and system reset (which includes normal shutdown and start).
I think the requirement is reasonable because for a certain virtio device, there is no reason that the feature bits it offers will change (because it should always offer all the features it understands). And we are just to add a device normative to make the spec be more explicit about that (because if a device really changes the features it offers after a device or system reset, something will go wrong). If the configs of an emulated virtio device are changed, maybe we shouldn't treat it as the same device any more, and IMO this case is not related to this proposal.
Quoted 4 lines
Although we have 'Each virtio device offers all the features it understands', it's not an explicit device requirement. So I don't think it's a bad idea to have an explicit device requirement about this.
I think this reasoning is sane and we really should not overthink it. The update as has been voted on looks fine to me.
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
Quoted 27 lines · Tiwei Bie
Quoted 2 lines · Halil Pasic
Quoted 1 line · Tiwei Bie
On Mon, Jun 18, 2018 at 07:28:33PM +0300, Michael S. Tsirkin wrote:
[...]
Quoted 9 lines
Quoted 3 lines
If it would be better to drop this patch,
I'm fine with dropping it. Thanks!@Tiwei Bie
Thanks for your flexibility! What is your opinion (after considering the arguments from my previous mail), is it better to include this patch in the spec or is it better to drop it? Were you able to identify mistakes in my reasoning (I mean points (1)-(12))?Hi Halil,
I think maybe you thought too much about this proposal (or maybe I really missed something obvious). In my opinion, the device requirement proposed by this patch is quite simple and straightforward:
- It's just to make the spec explicitly require that a certain virtio device shouldn't fail re-negotiation of a feature set it has successfully accepted once.
- It covers the cases of virtio device reset and system reset (which includes normal shutdown and start).
Quoted 11 lines
I think the requirement is reasonable because for a certain virtio device, there is no reason that the feature bits it offers will change (because it should always offer all the features it understands). And we are just to add a device normative to make the spec be more explicit about that (because if a device really changes the features it offers after a device or system reset, something will go wrong). If the configs of an emulated virtio device are changed, maybe we shouldn't treat it as the same device any more, and IMO this case is not related to this proposal.
Thanks for clarifying your position. I don't want to usurp any more of your valuable time. I'm not convinced but I've given up on hope to convince the opposition.
I'm giving up.
Regards,
Halil
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]