virtio: introduce SUSPEND bit in device status
PATCH 1/1 virtio: introduce SUSPEND bit in device status
This commit allows the driver to suspend the device by introducing a new status bit SUSPEND in device_status.
This commit also introduce a new feature bit VIRTIO_F_SUSPEND which indicating whether the device support SUSPEND.
This SUSPEND bit is transport-independent.
--- content.tex | 34 ++++++++++++++++++++++++++++++++-- 1 file changed, 32 insertions(+), 2 deletions(-)
content.tex +32 -2
@@ -49,6 +49,9 @@ \section{\field{Device Status} Field}\label{sec:Basic Facilities of a Virtio Dev \item[DEVICE_NEEDS_RESET (64)] Indicates that the device has experienced an error from which it can't recover. + +\item[SUSPEND (16)] When VIRTIO_F_SUSPEND is negotiated, indicates that the + device has been suspended by the driver. \end{description} The \field{device status} field starts out as 0, and is reinitialized to 0 by
@@ -73,6 +76,10 @@ \section{\field{Device Status} Field}\label{sec:Basic Facilities of a Virtio Dev recover by issuing a reset. \end{note} +The driver SHOULD NOT set SUSPEND if FEATURES_OK is not set. + +When setting SUSPEND, the driver MUST re-read \field{device status} to ensure the SUSPEND bit is set. + \devicenormative{\subsection}{Device Status Field}{Basic Facilities of a Virtio Device / Device Status Field} The device MUST NOT consume buffers or send any used buffer
@@ -82,6 +89,25 @@ \section{\field{Device Status} Field}\label{sec:Basic Facilities of a Virtio Dev that a reset is needed. If DRIVER_OK is set, after it sets DEVICE_NEEDS_RESET, the device MUST send a device configuration change notification to the driver. +The device MUST ignore SUSPEND if FEATURES_OK is not set. + +The device MUST ignore SUSPEND if VIRTIO_F_SUSPEND is not negotiated. + +The device SHOULD allow settings to \field{device status} even when SUSPEND is set. + +If VIRTIO_F_SUSPEND is negotiated and SUSPEND is set, the device SHOULD clear SUSPEND +and resumes operation upon DRIVER_OK. + +If VIRTIO_F_SUSPEND is negotiated, when the driver sets SUSPEND, +the device SHOULD perform the following actions before presenting SUSPEND bit in the \field{device status}: + +\begin{itemize} +\item Stop consuming buffers of any virtqueues and mark all finished descriptors as used. +\item Wait until all descriptors that being processed to finish and mark them as used. +\item Flush all used buffer and send used buffer notifications to the driver. +\item Pause its operation except \field{device status} and preserve configurations in its Device Configuration Space, see \ref{sec:Basic Facilities of a Virtio Device / Device Configuration Space} +\end{itemize} + \section{Feature Bits}\label{sec:Basic Facilities of a Virtio Device / Feature Bits} Each virtio device offers all the features it understands. During
@@ -99,10 +125,10 @@ \section{Feature Bits}\label{sec:Basic Facilities of a Virtio Device / Feature B \begin{description} \item[0 to 23, and 50 to 127] Feature bits for the specific device type -\item[24 to 41] Feature bits reserved for extensions to the queue and +\item[24 to 42] Feature bits reserved for extensions to the queue and feature negotiation mechanisms -\item[42 to 49, and 128 and above] Feature bits reserved for future extensions. +\item[43 to 49, and 128 and above] Feature bits reserved for future extensions. \end{description} \begin{note}
@@ -872,6 +898,10 @@ \chapter{Reserved Feature Bits}\label{sec:Reserved Feature Bits} \ref{devicenormative:Basic Facilities of a Virtio Device / Feature Bits} for handling features reserved for future use. + \item[VIRTIO_F_SUSPEND(42)] This feature indicates that the driver can + SUSPEND the device. + See \ref{sec:Basic Facilities of a Virtio Device / Device Status Field}. + \end{description} \drivernormative{\section}{Reserved Feature Bits}{Reserved Feature Bits}
Signature · 6 lines
2.35.3 --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
Quoted 11 lines · Zhu Lingshan
This commit allows the driver to suspend the device by introducing a new status bit SUSPEND in
device_status.This commit also introduce a new feature bit
VIRTIO_F_SUSPENDwhich indicating whether the device support SUSPEND.This SUSPEND bit is transport-independent.
Signed-off-by Zhu Lingshan <[email protected]>Signed-off-by Jason Wang <[email protected]>Signed-off-by Eugenio Pérez <[email protected]>
Could we get some kind of dscription how this has taken into consideration the proposal from David Stevens?
I find it really tiring when there are competing patches with authors ignoring each other's work and leaving it up to reviewers to figure out how do the patches compare.
Quoted 81 lines
--- content.tex | 34 ++++++++++++++++++++++++++++++++-- 1 file changed, 32 insertions(+), 2 deletions(-)content.tex +32
-2@@ -49,6 +49,9 @@ \section{\field{Device Status} Field}\label{sec:Basic Facilities of a Virtio Dev \item[DEVICE_NEEDS_RESET (64)] Indicates that the device has experienced an error from which it can't recover. + +\item[SUSPEND (16)] When VIRTIO_F_SUSPEND is negotiated, indicates that the + device has been suspended by the driver. \end{description} The \field{device status} field starts out as 0, and is reinitialized to 0 by@@ -73,6 +76,10 @@ \section{\field{Device Status} Field}\label{sec:Basic Facilities of a Virtio Dev recover by issuing a reset. \end{note} +The driver SHOULD NOT set SUSPEND if FEATURES_OK is not set. + +When setting SUSPEND, the driver MUST re-read \field{device status} to ensure the SUSPEND bit is set. + \devicenormative{\subsection}{Device Status Field}{Basic Facilities of a Virtio Device / Device Status Field} The device MUST NOT consume buffers or send any used buffer@@ -82,6 +89,25 @@ \section{\field{Device Status} Field}\label{sec:Basic Facilities of a Virtio Dev that a reset is needed. If DRIVER_OK is set, after it sets DEVICE_NEEDS_RESET, the device MUST send a device configuration change notification to the driver. +The device MUST ignore SUSPEND if FEATURES_OK is not set. + +The device MUST ignore SUSPEND if VIRTIO_F_SUSPEND is not negotiated. + +The device SHOULD allow settings to \field{device status} even when SUSPEND is set. + +If VIRTIO_F_SUSPEND is negotiated and SUSPEND is set, the device SHOULD clear SUSPEND +and resumes operation upon DRIVER_OK. + +If VIRTIO_F_SUSPEND is negotiated, when the driver sets SUSPEND, +the device SHOULD perform the following actions before presenting SUSPEND bit in the \field{device status}: + +\begin{itemize} +\item Stop consuming buffers of any virtqueues and mark all finished descriptors as used. +\item Wait until all descriptors that being processed to finish and mark them as used. +\item Flush all used buffer and send used buffer notifications to the driver. +\item Pause its operation except \field{device status} and preserve configurations in its Device Configuration Space, see \ref{sec:Basic Facilities of a Virtio Device / Device Configuration Space} +\end{itemize} + \section{Feature Bits}\label{sec:Basic Facilities of a Virtio Device / Feature Bits} Each virtio device offers all the features it understands. During@@ -99,10 +125,10 @@ \section{Feature Bits}\label{sec:Basic Facilities of a Virtio Device / Feature B \begin{description} \item[0 to 23, and 50 to 127] Feature bits for the specific device type -\item[24 to 41] Feature bits reserved for extensions to the queue and +\item[24 to 42] Feature bits reserved for extensions to the queue and feature negotiation mechanisms -\item[42 to 49, and 128 and above] Feature bits reserved for future extensions. +\item[43 to 49, and 128 and above] Feature bits reserved for future extensions. \end{description} \begin{note}@@ -872,6 +898,10 @@ \chapter{Reserved Feature Bits}\label{sec:Reserved Feature Bits} \ref{devicenormative:Basic Facilities of a Virtio Device / Feature Bits} for handling features reserved for future use. + \item[VIRTIO_F_SUSPEND(42)] This feature indicates that the driver can + SUSPEND the device. + See \ref{sec:Basic Facilities of a Virtio Device / Device Status Field}. + \end{description} \drivernormative{\section}{Reserved Feature Bits}{Reserved Feature Bits}Signature · 1 line
2.35.3
This publicly archived list offers a means to provide input to the OASIS Virtual I/O Device (VIRTIO) TC.
In order to verify user consent to the Feedback License terms and to minimize spam in the list archive, subscription is required before posting.
Subscribe: [email protected]
Unsubscribe: [email protected]
List help: [email protected]
List archive: https://lists.oasis-open.org/archives/virtio-comment/
Feedback License: https://www.oasis-open.org/who/ipr/feedback_license.pdf
List Guidelines: https://www.oasis-open.org/policies-guidelines/mailing-lists
Committee: https://www.oasis-open.org/committees/virtio/
Join OASIS: https://www.oasis-open.org/join/
Quoted 14 lines · Michael S. Tsirkin
Quoted 11 lines · Zhu Lingshan
This commit allows the driver to suspend the device by introducing a new status bit SUSPEND in
device_status.This commit also introduce a new feature bit
VIRTIO_F_SUSPENDwhich indicating whether the device support SUSPEND.This SUSPEND bit is transport-independent.
Signed-off-by Zhu Lingshan <[email protected]>Signed-off-by Jason Wang <[email protected]>Signed-off-by Eugenio Pérez <[email protected]>
Quoted 6 lines
Could we get some kind of dscription how this has taken into consideration the proposal from David Stevens?
I find it really tiring when there are competing patches with authors ignoring each other's work and leaving it up to reviewers to figure out how do the patches compare.
This patch looks like it could be used to implement my use case. However, parts of it are a bit vague and imprecise, so it's hard to actually say whether my use case would actually be covered by a specific implementation of this proposal.
To give specifics on what I'm trying to do, I need to allow a guest with virtio-pci devices (including stateful devices like virtio-fs and virtio-gpu) to be put into S1/S3, and to allow the virtio-pci devices to wake up the guest (currently via an ACPI GPE, but eventually via a native PCI PME). This is all done on a consumer device, so there is no need for snapshotting or for live migration.
Quoted 33 lines
Quoted 33 lines
--- content.tex | 34 ++++++++++++++++++++++++++++++++-- 1 file changed, 32 insertions(+), 2 deletions(-)content.tex +7
-0@@ -49,6 +49,9 @@ \section{\field{Device Status} Field}\label{sec:Basic Facilities of a Virtio Dev \item[DEVICE_NEEDS_RESET (64)] Indicates that the device has experienced an error from which it can't recover. + +\item[SUSPEND (16)] When VIRTIO_F_SUSPEND is negotiated, indicates that the + device has been suspended by the driver. \end{description} The \field{device status} field starts out as 0, and is reinitialized to 0 by@@ -73,6 +76,10 @@ \section{\field{Device Status} Field}\label{sec:Basic Facilities of a Virtio Dev recover by issuing a reset. \end{note} +The driver SHOULD NOT set SUSPEND if FEATURES_OK is not set. + +When setting SUSPEND, the driver MUST re-read \field{device status} to ensure the SUSPEND bit is set. + \devicenormative{\subsection}{Device Status Field}{Basic Facilities of a Virtio Device / Device Status Field} The device MUST NOT consume buffers or send any used buffer@@ -82,6 +89,25 @@ \section{\field{Device Status} Field}\label{sec:Basic Facilities of a Virtio Dev that a reset is needed. If DRIVER_OK is set, after it sets DEVICE_NEEDS_RESET, the device MUST send a device configuration change notification to the driver.
Quoted 16 lines
Quoted 6 lines
content.tex +6
-0+The device MUST ignore SUSPEND if FEATURES_OK is not set. + +The device MUST ignore SUSPEND if VIRTIO_F_SUSPEND is not negotiated. + +The device SHOULD allow settings to \field{device status} even when SUSPEND is set. +content.tex \section{\field{Device Status} Field}\label{sec:Basic Facilities of a Virtio Dev · line 107
Quoted 10 lines
content.tex +10
-0+If VIRTIO_F_SUSPEND is negotiated and SUSPEND is set, the device SHOULD clear SUSPEND +and resumes operation upon DRIVER_OK. + +If VIRTIO_F_SUSPEND is negotiated, when the driver sets SUSPEND, +the device SHOULD perform the following actions before presenting SUSPEND bit in the \field{device status}: + +\begin{itemize} +\item Stop consuming buffers of any virtqueues and mark all finished descriptors as used. +\item Wait until all descriptors that being processed to finish and mark them as used. +\item Flush all used buffer and send used buffer notifications to the driver.
What's the difference between the previous three lines? They might be trying to say different things, but there is a lot of overlap in how they're phrased. That makes it hard to figure out exactly what they're mandating. Is it something like: "stop processing new buffers", "finish processing any buffers that are currently in flight", "mark all finished buffers as used and send a used buffer notification"?
Also, what does this mean for devices where the driver places empty buffers into the virtqueue that the device then holds on to for an indeterminate period of time (e.g. network receiveq, balloon statsq, etc)?
Quoted 1 line
content.tex \section{\field{Device Status} Field}\label{sec:Basic Facilities of a Virtio Dev · line 108
Quoted 1 line
content.tex +1
-0+\item Pause its operation except \field{device status} and preserve configurations in its Device Configuration Space, see \ref{sec:Basic Facilities of a Virtio Device / Device Configuration Space}
What does "Pause its operation except device status" mean? The word "operation" is vague, what exactly does it include? Also what does "operate its device status" mean?
-David
This publicly archived list offers a means to provide input to the OASIS Virtual I/O Device (VIRTIO) TC.
In order to verify user consent to the Feedback License terms and to minimize spam in the list archive, subscription is required before posting.
Subscribe: [email protected]
Unsubscribe: [email protected]
List help: [email protected]
List archive: https://lists.oasis-open.org/archives/virtio-comment/
Feedback License: https://www.oasis-open.org/who/ipr/feedback_license.pdf
List Guidelines: https://www.oasis-open.org/policies-guidelines/mailing-lists
Committee: https://www.oasis-open.org/committees/virtio/
Join OASIS: https://www.oasis-open.org/join/
Quoted 16 lines · David Stevens
Quoted 15 lines · Michael S. Tsirkin
Quoted 11 lines · Zhu Lingshan
This commit allows the driver to suspend the device by introducing a new status bit SUSPEND in
device_status.This commit also introduce a new feature bit
VIRTIO_F_SUSPENDwhich indicating whether the device support SUSPEND.This SUSPEND bit is transport-independent.
Signed-off-by Zhu Lingshan <[email protected]>Signed-off-by Jason Wang <[email protected]>Signed-off-by Eugenio Pérez <[email protected]>Could we get some kind of dscription how this has taken into consideration the proposal from David Stevens?
Quoted 7 lines
Quoted 3 lines
I find it really tiring when there are competing patches with authors ignoring each other's work and leaving it up to reviewers to figure out how do the patches compare.
This patch looks like it could be used to implement my use case. However, parts of it are a bit vague and imprecise, so it's hard to actually say whether my use case would actually be covered by a specific implementation of this proposal.
I am on vacation till this Friday. Shall we co-work on this and post something new together?
Quoted 7 lines
To give specifics on what I'm trying to do, I need to allow a guest with virtio-pci devices (including stateful devices like virtio-fs and virtio-gpu) to be put into S1/S3, and to allow the virtio-pci devices to wake up the guest (currently via an ACPI GPE, but eventually via a native PCI PME). This is all done on a consumer device, so there is no need for snapshotting or for live migration.
Suspending is not dedicated only for live migration.
For your use case, shall we add a new PCI section like saying: when entering SUSPEND, the device should get into S1/S3 and other PCI specific steps in PCI transport charter.
That is because PCI is a transport layer of virtio, controlling virtio states by PCI sounds like a layer violation.
Quoted 48 lines
Quoted 48 lines
Quoted 48 lines
--- content.tex | 34 ++++++++++++++++++++++++++++++++-- 1 file changed, 32 insertions(+), 2 deletions(-)content.tex +21
-0@@ -49,6 +49,9 @@ \section{\field{Device Status} Field}\label{sec:Basic Facilities of a Virtio Dev \item[DEVICE_NEEDS_RESET (64)] Indicates that the device has experienced an error from which it can't recover. + +\item[SUSPEND (16)] When VIRTIO_F_SUSPEND is negotiated, indicates that the + device has been suspended by the driver. \end{description} The \field{device status} field starts out as 0, and is reinitialized to 0 by@@ -73,6 +76,10 @@ \section{\field{Device Status} Field}\label{sec:Basic Facilities of a Virtio Dev recover by issuing a reset. \end{note} +The driver SHOULD NOT set SUSPEND if FEATURES_OK is not set. + +When setting SUSPEND, the driver MUST re-read \field{device status} to ensure the SUSPEND bit is set. + \devicenormative{\subsection}{Device Status Field}{Basic Facilities of a Virtio Device / Device Status Field} The device MUST NOT consume buffers or send any used buffer@@ -82,6 +89,25 @@ \section{\field{Device Status} Field}\label{sec:Basic Facilities of a Virtio Dev that a reset is needed. If DRIVER_OK is set, after it sets DEVICE_NEEDS_RESET, the device MUST send a device configuration change notification to the driver. +The device MUST ignore SUSPEND if FEATURES_OK is not set. + +The device MUST ignore SUSPEND if VIRTIO_F_SUSPEND is not negotiated. + +The device SHOULD allow settings to \field{device status} even when SUSPEND is set. + +If VIRTIO_F_SUSPEND is negotiated and SUSPEND is set, the device SHOULD clear SUSPEND +and resumes operation upon DRIVER_OK. + +If VIRTIO_F_SUSPEND is negotiated, when the driver sets SUSPEND, +the device SHOULD perform the following actions before presenting SUSPEND bit in the \field{device status}: + +\begin{itemize} +\item Stop consuming buffers of any virtqueues and mark all finished descriptors as used.
Quoted 8 lines
Quoted 2 lines
content.tex \section{\field{Device Status} Field}\label{sec:Basic Facilities of a Virtio Dev · line 107
Quoted 2 lines
content.tex +2
-0+\item Wait until all descriptors that being processed to finish and mark them as used. +\item Flush all used buffer and send used buffer notifications to the driver.What's the difference between the previous three lines? They might be trying to say different things, but there is a lot of overlap in how they're phrased. That makes it hard to figure out exactly what they're mandating. Is it something like: "stop processing new buffers", "finish processing any buffers that are currently in flight", "mark all finished buffers as used and send a used buffer notification"?
sure, the "mark used" parts can merge.
Quoted 5 lines
Also, what does this mean for devices where the driver places empty buffers into the virtqueue that the device then holds on to for an indeterminate period of time (e.g. network receiveq, balloon statsq, etc)?
The device doesn't process them.
Oh, I should add: The driver should not make any more buffers available to the device.
Quoted 5 lines
Quoted 2 lines
content.tex \section{\field{Device Status} Field}\label{sec:Basic Facilities of a Virtio Dev · line 108
Quoted 2 lines
content.tex +1
-0+\item Pause its operation except \field{device status} and preserve configurations in its Device Configuration Space, see \ref{sec:Basic Facilities of a Virtio Device / Device Configuration Space}What does "Pause its operation except device status" mean? The word "operation" is vague, what exactly does it include? Also what does "operate its device status" mean?
Because we need to resume the device after suspending it by setting DRIVER_OK, so we need the device status filed keel alive.
Thanks,
Zhu Lingshan
Quoted 2 lines
-David
This publicly archived list offers a means to provide input to the OASIS Virtual I/O Device (VIRTIO) TC.
In order to verify user consent to the Feedback License terms and to minimize spam in the list archive, subscription is required before posting.
Subscribe: [email protected]
Unsubscribe: [email protected]
List help: [email protected]
List archive: https://lists.oasis-open.org/archives/virtio-comment/
Feedback License: https://www.oasis-open.org/who/ipr/feedback_license.pdf
List Guidelines: https://www.oasis-open.org/policies-guidelines/mailing-lists
Committee: https://www.oasis-open.org/committees/virtio/
Join OASIS: https://www.oasis-open.org/join/
Quoted 20 lines · Lingshan
Quoted 19 lines · David Stevens
Quoted 18 lines · Michael S. Tsirkin
Quoted 11 lines · Zhu Lingshan
This commit allows the driver to suspend the device by introducing a new status bit SUSPEND in
device_status.This commit also introduce a new feature bit
VIRTIO_F_SUSPENDwhich indicating whether the device support SUSPEND.This SUSPEND bit is transport-independent.
Signed-off-by Zhu Lingshan <[email protected]>Signed-off-by Jason Wang <[email protected]>Signed-off-by Eugenio Pérez <[email protected]>Could we get some kind of dscription how this has taken into consideration the proposal from David Stevens?
I find it really tiring when there are competing patches with authors ignoring each other's work and leaving it up to reviewers to figure out how do the patches compare.
Quoted 6 lines
Quoted 4 lines
This patch looks like it could be used to implement my use case. However, parts of it are a bit vague and imprecise, so it's hard to actually say whether my use case would actually be covered by a specific implementation of this proposal.
I am on vacation till this Friday. Shall we co-work on this and post something new together?
That works for me.
Quoted 7 lines
Quoted 7 lines
To give specifics on what I'm trying to do, I need to allow a guest with virtio-pci devices (including stateful devices like virtio-fs and virtio-gpu) to be put into S1/S3, and to allow the virtio-pci devices to wake up the guest (currently via an ACPI GPE, but eventually via a native PCI PME). This is all done on a consumer device, so there is no need for snapshotting or for live migration.
Quoted 7 lines
Suspending is not dedicated only for live migration.
For your use case, shall we add a new PCI section like saying: when entering SUSPEND, the device should get into S1/S3 and other PCI specific steps in PCI transport charter.That is because PCI is a transport layer of virtio, controlling virtio states by PCI sounds like a layer violation.
I don't know if it's really necessary to mandate that in the spec. Sure, most systems are probably going to want to put the virtio-pci device into S1/S3 after suspending virtio operation. But nothing really breaks or goes wrong if we don't do that.
Quoted 66 lines
Quoted 48 lines
Quoted 48 lines
Quoted 48 lines
--- content.tex | 34 ++++++++++++++++++++++++++++++++-- 1 file changed, 32 insertions(+), 2 deletions(-)content.tex +21
-0@@ -49,6 +49,9 @@ \section{\field{Device Status} Field}\label{sec:Basic Facilities of a Virtio Dev \item[DEVICE_NEEDS_RESET (64)] Indicates that the device has experienced an error from which it can't recover. + +\item[SUSPEND (16)] When VIRTIO_F_SUSPEND is negotiated, indicates that the + device has been suspended by the driver. \end{description} The \field{device status} field starts out as 0, and is reinitialized to 0 by@@ -73,6 +76,10 @@ \section{\field{Device Status} Field}\label{sec:Basic Facilities of a Virtio Dev recover by issuing a reset. \end{note} +The driver SHOULD NOT set SUSPEND if FEATURES_OK is not set. + +When setting SUSPEND, the driver MUST re-read \field{device status} to ensure the SUSPEND bit is set. + \devicenormative{\subsection}{Device Status Field}{Basic Facilities of a Virtio Device / Device Status Field} The device MUST NOT consume buffers or send any used buffer@@ -82,6 +89,25 @@ \section{\field{Device Status} Field}\label{sec:Basic Facilities of a Virtio Dev that a reset is needed. If DRIVER_OK is set, after it sets DEVICE_NEEDS_RESET, the device MUST send a device configuration change notification to the driver. +The device MUST ignore SUSPEND if FEATURES_OK is not set. + +The device MUST ignore SUSPEND if VIRTIO_F_SUSPEND is not negotiated. + +The device SHOULD allow settings to \field{device status} even when SUSPEND is set. + +If VIRTIO_F_SUSPEND is negotiated and SUSPEND is set, the device SHOULD clear SUSPEND +and resumes operation upon DRIVER_OK. + +If VIRTIO_F_SUSPEND is negotiated, when the driver sets SUSPEND, +the device SHOULD perform the following actions before presenting SUSPEND bit in the \field{device status}: + +\begin{itemize} +\item Stop consuming buffers of any virtqueues and mark all finished descriptors as used.Quoted 8 lines
Quoted 2 lines
content.tex \section{\field{Device Status} Field}\label{sec:Basic Facilities of a Virtio Dev · line 107
Quoted 2 lines
content.tex +2
-0+\item Wait until all descriptors that being processed to finish and mark them as used. +\item Flush all used buffer and send used buffer notifications to the driver.What's the difference between the previous three lines? They might be trying to say different things, but there is a lot of overlap in how they're phrased. That makes it hard to figure out exactly what they're mandating. Is it something like: "stop processing new buffers", "finish processing any buffers that are currently in flight", "mark all finished buffers as used and send a used buffer notification"?
sure, the "mark used" parts can merge.
Quoted 5 lines
Also, what does this mean for devices where the driver places empty buffers into the virtqueue that the device then holds on to for an indeterminate period of time (e.g. network receiveq, balloon statsq, etc)?
The device doesn't process them.
Oh, I should add: The driver should not make any more buffers available to the device.Quoted 1 line
Quoted 7 lines
Quoted 4 lines
Quoted 1 line
content.tex \section{\field{Device Status} Field}\label{sec:Basic Facilities of a Virtio Dev · line 108
Quoted 1 line
content.tex +1
-0+\item Pause its operation except \field{device status} and preserve configurations in its Device Configuration Space, see \ref{sec:Basic Facilities of a Virtio Device / Device Configuration Space}What does "Pause its operation except device status" mean? The word "operation" is vague, what exactly does it include? Also what does "operate its device status" mean?
Because we need to resume the device after suspending it by setting
DRIVER_OK, so we need the device status
filed keel alive.
This doesn't answer what "Pause its operation" means. If the specification is not precise, then the best that can be implemented is for a specific device to be compatible with a specific driver. If we actually want portable, spec-compliant devices, we need explicitly defined MUST/MUST NOT requirements. Presumably we want something like:
A device MUST not send notifications while suspended.
A device MUST ignore writes to its device configuration while suspended, except for writes to the device status byte.
A driver MUST NOT write to a suspended device's configuration space, except for the device status byte.
We also need to specify what a suspended device is allowed to do with buffers it owns. For my use case of just suspending/resuming a VM, I don't think it particularly matters if a suspended device writes to buffers or descriptors while it is suspended. But based on what you wrote about finishing any in-flight processing of buffers and not processing any empty buffers, presumably you would prefer that the device not be allowed to access buffer/descriptors while it is suspended? That approach is probably cleaner from a pure specification standpoint, but it's also more restrictive for devices and thus harder to properly implement.
-David
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
Quoted 42 lines · David Stevens
Quoted 20 lines · Lingshan
Quoted 19 lines · David Stevens
Quoted 18 lines · Michael S. Tsirkin
Quoted 11 lines · Zhu Lingshan
This commit allows the driver to suspend the device by introducing a new status bit SUSPEND in
device_status.This commit also introduce a new feature bit
VIRTIO_F_SUSPENDwhich indicating whether the device support SUSPEND.This SUSPEND bit is transport-independent.
Signed-off-by Zhu Lingshan <[email protected]>Signed-off-by Jason Wang <[email protected]>Signed-off-by Eugenio Pérez <[email protected]>Could we get some kind of dscription how this has taken into consideration the proposal from David Stevens?
I find it really tiring when there are competing patches with authors ignoring each other's work and leaving it up to reviewers to figure out how do the patches compare.
Quoted 6 lines
Quoted 4 lines
This patch looks like it could be used to implement my use case. However, parts of it are a bit vague and imprecise, so it's hard to actually say whether my use case would actually be covered by a specific implementation of this proposal.
I am on vacation till this Friday. Shall we co-work on this and post something new together?
That works for me.
Quoted 12 lines
Quoted 7 lines
To give specifics on what I'm trying to do, I need to allow a guest with virtio-pci devices (including stateful devices like virtio-fs and virtio-gpu) to be put into S1/S3, and to allow the virtio-pci devices to wake up the guest (currently via an ACPI GPE, but eventually via a native PCI PME). This is all done on a consumer device, so there is no need for snapshotting or for live migration.
Suspending is not dedicated only for live migration.
For your use case, shall we add a new PCI section like saying: when entering SUSPEND, the device should get into S1/S3 and other PCI specific steps in PCI transport charter.
Quoted 7 lines
Quoted 2 lines
That is because PCI is a transport layer of virtio, controlling virtio states by PCI sounds like a layer violation.
I don't know if it's really necessary to mandate that in the spec. Sure, most systems are probably going to want to put the virtio-pci device into S1/S3 after suspending virtio operation. But nothing really breaks or goes wrong if we don't do that.
What I feel is worth mentioning in virtio spec is that just setting the status bit does not reduce power by itself, standard pci control is required for that.
Quoted 97 lines
Quoted 66 lines
Quoted 48 lines
Quoted 48 lines
Quoted 48 lines
--- content.tex | 34 ++++++++++++++++++++++++++++++++-- 1 file changed, 32 insertions(+), 2 deletions(-)content.tex +21
-0@@ -49,6 +49,9 @@ \section{\field{Device Status} Field}\label{sec:Basic Facilities of a Virtio Dev \item[DEVICE_NEEDS_RESET (64)] Indicates that the device has experienced an error from which it can't recover. + +\item[SUSPEND (16)] When VIRTIO_F_SUSPEND is negotiated, indicates that the + device has been suspended by the driver. \end{description} The \field{device status} field starts out as 0, and is reinitialized to 0 by@@ -73,6 +76,10 @@ \section{\field{Device Status} Field}\label{sec:Basic Facilities of a Virtio Dev recover by issuing a reset. \end{note} +The driver SHOULD NOT set SUSPEND if FEATURES_OK is not set. + +When setting SUSPEND, the driver MUST re-read \field{device status} to ensure the SUSPEND bit is set. + \devicenormative{\subsection}{Device Status Field}{Basic Facilities of a Virtio Device / Device Status Field} The device MUST NOT consume buffers or send any used buffer@@ -82,6 +89,25 @@ \section{\field{Device Status} Field}\label{sec:Basic Facilities of a Virtio Dev that a reset is needed. If DRIVER_OK is set, after it sets DEVICE_NEEDS_RESET, the device MUST send a device configuration change notification to the driver. +The device MUST ignore SUSPEND if FEATURES_OK is not set. + +The device MUST ignore SUSPEND if VIRTIO_F_SUSPEND is not negotiated. + +The device SHOULD allow settings to \field{device status} even when SUSPEND is set. + +If VIRTIO_F_SUSPEND is negotiated and SUSPEND is set, the device SHOULD clear SUSPEND +and resumes operation upon DRIVER_OK. + +If VIRTIO_F_SUSPEND is negotiated, when the driver sets SUSPEND, +the device SHOULD perform the following actions before presenting SUSPEND bit in the \field{device status}: + +\begin{itemize} +\item Stop consuming buffers of any virtqueues and mark all finished descriptors as used.Quoted 8 lines
Quoted 2 lines
content.tex \section{\field{Device Status} Field}\label{sec:Basic Facilities of a Virtio Dev · line 107
Quoted 2 lines
content.tex +2
-0+\item Wait until all descriptors that being processed to finish and mark them as used. +\item Flush all used buffer and send used buffer notifications to the driver.What's the difference between the previous three lines? They might be trying to say different things, but there is a lot of overlap in how they're phrased. That makes it hard to figure out exactly what they're mandating. Is it something like: "stop processing new buffers", "finish processing any buffers that are currently in flight", "mark all finished buffers as used and send a used buffer notification"?
sure, the "mark used" parts can merge.
Quoted 5 lines
Also, what does this mean for devices where the driver places empty buffers into the virtqueue that the device then holds on to for an indeterminate period of time (e.g. network receiveq, balloon statsq, etc)?
The device doesn't process them.
Oh, I should add: The driver should not make any more buffers available to the device.Quoted 1 line
Quoted 7 lines
Quoted 4 lines
Quoted 1 line
content.tex \section{\field{Device Status} Field}\label{sec:Basic Facilities of a Virtio Dev · line 108
Quoted 1 line
content.tex +1
-0+\item Pause its operation except \field{device status} and preserve configurations in its Device Configuration Space, see \ref{sec:Basic Facilities of a Virtio Device / Device Configuration Space}What does "Pause its operation except device status" mean? The word "operation" is vague, what exactly does it include? Also what does "operate its device status" mean?
Because we need to resume the device after suspending it by setting
DRIVER_OK, so we need the device status
filed keel alive.This doesn't answer what "Pause its operation" means. If the specification is not precise, then the best that can be implemented is for a specific device to be compatible with a specific driver. If we actually want portable, spec-compliant devices, we need explicitly defined MUST/MUST NOT requirements. Presumably we want something like:
A device MUST not send notifications while suspended.
A device MUST ignore writes to its device configuration while suspended, except for writes to the device status byte.
A driver MUST NOT write to a suspended device's configuration space, except for the device status byte.We also need to specify what a suspended device is allowed to do with buffers it owns. For my use case of just suspending/resuming a VM, I don't think it particularly matters if a suspended device writes to buffers or descriptors while it is suspended. But based on what you wrote about finishing any in-flight processing of buffers and not processing any empty buffers, presumably you would prefer that the device not be allowed to access buffer/descriptors while it is suspended? That approach is probably cleaner from a pure specification standpoint, but it's also more restrictive for devices and thus harder to properly implement.
Quoted 1 line
-David
Of course it's a bit harder. It does not sound too restrictive to me: as driver will not be adding new buffers, if device is still processing queues just defer setting the bit until it does not.
This publicly archived list offers a means to provide input to the OASIS Virtual I/O Device (VIRTIO) TC.
In order to verify user consent to the Feedback License terms and to minimize spam in the list archive, subscription is required before posting.
Subscribe: [email protected]
Unsubscribe: [email protected]
List help: [email protected]
List archive: https://lists.oasis-open.org/archives/virtio-comment/
Feedback License: https://www.oasis-open.org/who/ipr/feedback_license.pdf
List Guidelines: https://www.oasis-open.org/policies-guidelines/mailing-lists
Committee: https://www.oasis-open.org/committees/virtio/
Join OASIS: https://www.oasis-open.org/join/
Quoted 43 lines · Michael S. Tsirkin
Quoted 42 lines · David Stevens
Quoted 20 lines · Lingshan
Quoted 19 lines · David Stevens
Quoted 18 lines · Michael S. Tsirkin
Quoted 11 lines · Zhu Lingshan
This commit allows the driver to suspend the device by introducing a new status bit SUSPEND in
device_status.This commit also introduce a new feature bit
VIRTIO_F_SUSPENDwhich indicating whether the device support SUSPEND.This SUSPEND bit is transport-independent.
Signed-off-by Zhu Lingshan <[email protected]>Signed-off-by Jason Wang <[email protected]>Signed-off-by Eugenio Pérez <[email protected]>Could we get some kind of dscription how this has taken into consideration the proposal from David Stevens?
I find it really tiring when there are competing patches with authors ignoring each other's work and leaving it up to reviewers to figure out how do the patches compare.
Quoted 6 lines
Quoted 4 lines
This patch looks like it could be used to implement my use case. However, parts of it are a bit vague and imprecise, so it's hard to actually say whether my use case would actually be covered by a specific implementation of this proposal.
I am on vacation till this Friday. Shall we co-work on this and post something new together?
That works for me.
Quoted 13 lines
Quoted 6 lines
To give specifics on what I'm trying to do, I need to allow a guest with virtio-pci devices (including stateful devices like virtio-fs and virtio-gpu) to be put into S1/S3, and to allow the virtio-pci devices to wake up the guest (currently via an ACPI GPE, but eventually via a native PCI PME). This is all done on a consumer device, so there is no need for snapshotting or for live migration.
Suspending is not dedicated only for live migration.
For your use case, shall we add a new PCI section like saying: when entering SUSPEND, the device should get into S1/S3 and other PCI specific steps in PCI transport charter.That is because PCI is a transport layer of virtio, controlling virtio states by PCI sounds like a layer violation.
Quoted 7 lines
Quoted 4 lines
I don't know if it's really necessary to mandate that in the spec. Sure, most systems are probably going to want to put the virtio-pci device into S1/S3 after suspending virtio operation. But nothing really breaks or goes wrong if we don't do that.
What I feel is worth mentioning in virtio spec is that just setting the status bit does not reduce power by itself, standard pci control is required for that.
I agree, only mention power management in a separate PCI transport patch in a new collaborated series
Quoted 94 lines
Quoted 94 lines
Quoted 64 lines
Quoted 48 lines
Quoted 48 lines
Quoted 48 lines
--- content.tex | 34 ++++++++++++++++++++++++++++++++-- 1 file changed, 32 insertions(+), 2 deletions(-)content.tex +21
-0@@ -49,6 +49,9 @@ \section{\field{Device Status} Field}\label{sec:Basic Facilities of a Virtio Dev \item[DEVICE_NEEDS_RESET (64)] Indicates that the device has experienced an error from which it can't recover. + +\item[SUSPEND (16)] When VIRTIO_F_SUSPEND is negotiated, indicates that the + device has been suspended by the driver. \end{description} The \field{device status} field starts out as 0, and is reinitialized to 0 by@@ -73,6 +76,10 @@ \section{\field{Device Status} Field}\label{sec:Basic Facilities of a Virtio Dev recover by issuing a reset. \end{note} +The driver SHOULD NOT set SUSPEND if FEATURES_OK is not set. + +When setting SUSPEND, the driver MUST re-read \field{device status} to ensure the SUSPEND bit is set. + \devicenormative{\subsection}{Device Status Field}{Basic Facilities of a Virtio Device / Device Status Field} The device MUST NOT consume buffers or send any used buffer@@ -82,6 +89,25 @@ \section{\field{Device Status} Field}\label{sec:Basic Facilities of a Virtio Dev that a reset is needed. If DRIVER_OK is set, after it sets DEVICE_NEEDS_RESET, the device MUST send a device configuration change notification to the driver. +The device MUST ignore SUSPEND if FEATURES_OK is not set. + +The device MUST ignore SUSPEND if VIRTIO_F_SUSPEND is not negotiated. + +The device SHOULD allow settings to \field{device status} even when SUSPEND is set. + +If VIRTIO_F_SUSPEND is negotiated and SUSPEND is set, the device SHOULD clear SUSPEND +and resumes operation upon DRIVER_OK. + +If VIRTIO_F_SUSPEND is negotiated, when the driver sets SUSPEND, +the device SHOULD perform the following actions before presenting SUSPEND bit in the \field{device status}: + +\begin{itemize} +\item Stop consuming buffers of any virtqueues and mark all finished descriptors as used.Quoted 8 lines
Quoted 2 lines
content.tex \section{\field{Device Status} Field}\label{sec:Basic Facilities of a Virtio Dev · line 107
Quoted 2 lines
content.tex +2
-0+\item Wait until all descriptors that being processed to finish and mark them as used. +\item Flush all used buffer and send used buffer notifications to the driver.What's the difference between the previous three lines? They might be trying to say different things, but there is a lot of overlap in how they're phrased. That makes it hard to figure out exactly what they're mandating. Is it something like: "stop processing new buffers", "finish processing any buffers that are currently in flight", "mark all finished buffers as used and send a used buffer notification"?
sure, the "mark used" parts can merge.
Quoted 4 lines
Also, what does this mean for devices where the driver places empty buffers into the virtqueue that the device then holds on to for an indeterminate period of time (e.g. network receiveq, balloon statsq, etc)?
The device doesn't process them.
Oh, I should add: The driver should not make any more buffers available to the device.Quoted 7 lines
Quoted 4 lines
Quoted 1 line
content.tex \section{\field{Device Status} Field}\label{sec:Basic Facilities of a Virtio Dev · line 108
Quoted 1 line
content.tex +1
-0+\item Pause its operation except \field{device status} and preserve configurations in its Device Configuration Space, see \ref{sec:Basic Facilities of a Virtio Device / Device Configuration Space}What does "Pause its operation except device status" mean? The word "operation" is vague, what exactly does it include? Also what does "operate its device status" mean?
Because we need to resume the device after suspending it by setting
DRIVER_OK, so we need the device status
filed keel alive.This doesn't answer what "Pause its operation" means. If the specification is not precise, then the best that can be implemented is for a specific device to be compatible with a specific driver. If we actually want portable, spec-compliant devices, we need explicitly defined MUST/MUST NOT requirements. Presumably we want something like:
A device MUST not send notifications while suspended.
A device MUST ignore writes to its device configuration while suspended, except for writes to the device status byte.
A driver MUST NOT write to a suspended device's configuration space, except for the device status byte.We also need to specify what a suspended device is allowed to do with buffers it owns. For my use case of just suspending/resuming a VM, I don't think it particularly matters if a suspended device writes to buffers or descriptors while it is suspended. But based on what you wrote about finishing any in-flight processing of buffers and not processing any empty buffers, presumably you would prefer that the device not be allowed to access buffer/descriptors while it is suspended? That approach is probably cleaner from a pure specification standpoint, but it's also more restrictive for devices and thus harder to properly implement.
Quoted 4 lines
Quoted 1 line
-David
Of course it's a bit harder. It does not sound too restrictive to me: as driver will not be adding new buffers, if device is still processing queues just defer setting the bit until it does not.
Yes, I agree
Thanks
Zhu Lingshan
Quoted 1 line
This publicly archived list offers a means to provide input to the OASIS Virtual I/O Device (VIRTIO) TC.
In order to verify user consent to the Feedback License terms and to minimize spam in the list archive, subscription is required before posting.
Subscribe: [email protected]
Unsubscribe: [email protected]
List help: [email protected]
List archive: https://lists.oasis-open.org/archives/virtio-comment/
Feedback License: https://www.oasis-open.org/who/ipr/feedback_license.pdf
List Guidelines: https://www.oasis-open.org/policies-guidelines/mailing-lists
Committee: https://www.oasis-open.org/committees/virtio/
Join OASIS: https://www.oasis-open.org/join/
Quoted 21 lines · David Stevens
Quoted 20 lines · Lingshan
Quoted 19 lines · David Stevens
Quoted 18 lines · Michael S. Tsirkin
Quoted 11 lines · Zhu Lingshan
This commit allows the driver to suspend the device by introducing a new status bit SUSPEND in
device_status.This commit also introduce a new feature bit
VIRTIO_F_SUSPENDwhich indicating whether the device support SUSPEND.This SUSPEND bit is transport-independent.
Signed-off-by Zhu Lingshan <[email protected]>Signed-off-by Jason Wang <[email protected]>Signed-off-by Eugenio Pérez <[email protected]>Could we get some kind of dscription how this has taken into consideration the proposal from David Stevens?
I find it really tiring when there are competing patches with authors ignoring each other's work and leaving it up to reviewers to figure out how do the patches compare.
Quoted 7 lines
Quoted 6 lines
Quoted 4 lines
This patch looks like it could be used to implement my use case. However, parts of it are a bit vague and imprecise, so it's hard to actually say whether my use case would actually be covered by a specific implementation of this proposal.
I am on vacation till this Friday. Shall we co-work on this and post something new together?
That works for me.
Cool! How about two patches in a new series:
1) a general virtio suspending patch describing suspending behaviors for all virtio devices from me, an updating version of this patch
2) a PCI transport suspending patch from you, describing PCI specific behaviors.
Dose this sound good to you?
Quoted 12 lines
Quoted 12 lines
Quoted 7 lines
To give specifics on what I'm trying to do, I need to allow a guest with virtio-pci devices (including stateful devices like virtio-fs and virtio-gpu) to be put into S1/S3, and to allow the virtio-pci devices to wake up the guest (currently via an ACPI GPE, but eventually via a native PCI PME). This is all done on a consumer device, so there is no need for snapshotting or for live migration.
Suspending is not dedicated only for live migration.
For your use case, shall we add a new PCI section like saying: when entering SUSPEND, the device should get into S1/S3 and other PCI specific steps in PCI transport charter.
Quoted 6 lines
Quoted 2 lines
That is because PCI is a transport layer of virtio, controlling virtio states by PCI sounds like a layer violation.
I don't know if it's really necessary to mandate that in the spec. Sure, most systems are probably going to want to put the virtio-pci device into S1/S3 after suspending virtio operation. But nothing really breaks or goes wrong if we don't do that.
Yes, maybe not S1/S3, I remember you want to manage power in PCI transport when suspend in a cap, S1/S3 is just an example.
Quoted 71 lines
Quoted 64 lines
Quoted 48 lines
Quoted 48 lines
Quoted 48 lines
--- content.tex | 34 ++++++++++++++++++++++++++++++++-- 1 file changed, 32 insertions(+), 2 deletions(-)content.tex +21
-0@@ -49,6 +49,9 @@ \section{\field{Device Status} Field}\label{sec:Basic Facilities of a Virtio Dev \item[DEVICE_NEEDS_RESET (64)] Indicates that the device has experienced an error from which it can't recover. + +\item[SUSPEND (16)] When VIRTIO_F_SUSPEND is negotiated, indicates that the + device has been suspended by the driver. \end{description} The \field{device status} field starts out as 0, and is reinitialized to 0 by@@ -73,6 +76,10 @@ \section{\field{Device Status} Field}\label{sec:Basic Facilities of a Virtio Dev recover by issuing a reset. \end{note} +The driver SHOULD NOT set SUSPEND if FEATURES_OK is not set. + +When setting SUSPEND, the driver MUST re-read \field{device status} to ensure the SUSPEND bit is set. + \devicenormative{\subsection}{Device Status Field}{Basic Facilities of a Virtio Device / Device Status Field} The device MUST NOT consume buffers or send any used buffer@@ -82,6 +89,25 @@ \section{\field{Device Status} Field}\label{sec:Basic Facilities of a Virtio Dev that a reset is needed. If DRIVER_OK is set, after it sets DEVICE_NEEDS_RESET, the device MUST send a device configuration change notification to the driver. +The device MUST ignore SUSPEND if FEATURES_OK is not set. + +The device MUST ignore SUSPEND if VIRTIO_F_SUSPEND is not negotiated. + +The device SHOULD allow settings to \field{device status} even when SUSPEND is set. + +If VIRTIO_F_SUSPEND is negotiated and SUSPEND is set, the device SHOULD clear SUSPEND +and resumes operation upon DRIVER_OK. + +If VIRTIO_F_SUSPEND is negotiated, when the driver sets SUSPEND, +the device SHOULD perform the following actions before presenting SUSPEND bit in the \field{device status}: + +\begin{itemize} +\item Stop consuming buffers of any virtqueues and mark all finished descriptors as used.Quoted 8 lines
Quoted 2 lines
content.tex \section{\field{Device Status} Field}\label{sec:Basic Facilities of a Virtio Dev · line 107
Quoted 2 lines
content.tex +2
-0+\item Wait until all descriptors that being processed to finish and mark them as used. +\item Flush all used buffer and send used buffer notifications to the driver.What's the difference between the previous three lines? They might be trying to say different things, but there is a lot of overlap in how they're phrased. That makes it hard to figure out exactly what they're mandating. Is it something like: "stop processing new buffers", "finish processing any buffers that are currently in flight", "mark all finished buffers as used and send a used buffer notification"?
sure, the "mark used" parts can merge.
Quoted 4 lines
Also, what does this mean for devices where the driver places empty buffers into the virtqueue that the device then holds on to for an indeterminate period of time (e.g. network receiveq, balloon statsq, etc)?
The device doesn't process them.
Oh, I should add: The driver should not make any more buffers available to the device.Quoted 7 lines
Quoted 4 lines
Quoted 1 line
content.tex \section{\field{Device Status} Field}\label{sec:Basic Facilities of a Virtio Dev · line 108
Quoted 1 line
content.tex +1
-0+\item Pause its operation except \field{device status} and preserve configurations in its Device Configuration Space, see \ref{sec:Basic Facilities of a Virtio Device / Device Configuration Space}What does "Pause its operation except device status" mean? The word "operation" is vague, what exactly does it include? Also what does "operate its device status" mean?
Because we need to resume the device after suspending it by setting
DRIVER_OK, so we need the device status
filed keel alive.
Quoted 7 lines
This doesn't answer what "Pause its operation" means. If the specification is not precise, then the best that can be implemented is for a specific device to be compatible with a specific driver. If we actually want portable, spec-compliant devices, we need explicitly defined MUST/MUST NOT requirements. Presumably we want something like:
A device MUST not send notifications while suspended.
The device should send a last used buffer notification if it has buffers under processing before presenting SUSPEND in the device status, but the device should not process any new buffers. These are in devicenormative section, but they may be vague, I can add more details to make it precise, thanks for the tips
Quoted 4 lines
A device MUST ignore writes to its device configuration while suspended, except for writes to the device status byte.
A driver MUST NOT write to a suspended device's configuration space, except for the device status byte.
sure, we can make it more precise than only pausing.
Quoted 11 lines
We also need to specify what a suspended device is allowed to do with buffers it owns. For my use case of just suspending/resuming a VM, I don't think it particularly matters if a suspended device writes to buffers or descriptors while it is suspended. But based on what you wrote about finishing any in-flight processing of buffers and not processing any empty buffers, presumably you would prefer that the device not be allowed to access buffer/descriptors while it is suspended? That approach is probably cleaner from a pure specification standpoint, but it's also more restrictive for devices and thus harder to properly implement.
When the driver sets SUSPEND, the device should finished all buffers that already under processing, include the in-flight buffers(or it may lose the buffers). And the device should not touch the buffers once it present SUSPEND in the device status field, because once it presents SUSPEND, the device state should be stable, people may use SUSPEND to debug the device.
The device still can process the buffer before it presenting SUSPEND.
Thanks
Zhu Lingshan
Quoted 2 lines
-David
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
Quoted 30 lines · Lingshan
Quoted 27 lines · David Stevens
Quoted 26 lines · Lingshan
Quoted 16 lines · David Stevens
Quoted 15 lines · Michael S. Tsirkin
Quoted 11 lines · Zhu Lingshan
This commit allows the driver to suspend the device by introducing a new status bit SUSPEND in
device_status.This commit also introduce a new feature bit
VIRTIO_F_SUSPENDwhich indicating whether the device support SUSPEND.This SUSPEND bit is transport-independent.
Signed-off-by Zhu Lingshan <[email protected]>Signed-off-by Jason Wang <[email protected]>Signed-off-by Eugenio Pérez <[email protected]>Could we get some kind of dscription how this has taken into consideration the proposal from David Stevens?
Quoted 7 lines
Quoted 3 lines
I find it really tiring when there are competing patches with authors ignoring each other's work and leaving it up to reviewers to figure out how do the patches compare.
This patch looks like it could be used to implement my use case. However, parts of it are a bit vague and imprecise, so it's hard to actually say whether my use case would actually be covered by a specific implementation of this proposal.
I am on vacation till this Friday. Shall we co-work on this and post something new together?
Quoted 8 lines
Quoted 1 line
That works for me.
Cool! How about two patches in a new series:
1) a general virtio suspending patch describing suspending behaviors for all virtio devices from me, an updating version of this patch
2) a PCI transport suspending patch from you, describing PCI specific behaviors.Dose this sound good to you?
Well David's patch was more precise than yours, and also better worded. So think if there's an agreement you guys would really start with that, move the functionality to the status bit and make other changes as needed.
Quoted 129 lines
Quoted 12 lines
Quoted 12 lines
Quoted 7 lines
To give specifics on what I'm trying to do, I need to allow a guest with virtio-pci devices (including stateful devices like virtio-fs and virtio-gpu) to be put into S1/S3, and to allow the virtio-pci devices to wake up the guest (currently via an ACPI GPE, but eventually via a native PCI PME). This is all done on a consumer device, so there is no need for snapshotting or for live migration.
Suspending is not dedicated only for live migration.
For your use case, shall we add a new PCI section like saying: when entering SUSPEND, the device should get into S1/S3 and other PCI specific steps in PCI transport charter.Quoted 6 lines
Quoted 2 lines
That is because PCI is a transport layer of virtio, controlling virtio states by PCI sounds like a layer violation.
I don't know if it's really necessary to mandate that in the spec. Sure, most systems are probably going to want to put the virtio-pci device into S1/S3 after suspending virtio operation. But nothing really breaks or goes wrong if we don't do that.
Yes, maybe not S1/S3, I remember you want to manage power in PCI transport when suspend in a cap, S1/S3 is just an example.
Quoted 71 lines
Quoted 64 lines
Quoted 48 lines
Quoted 48 lines
Quoted 48 lines
--- content.tex | 34 ++++++++++++++++++++++++++++++++-- 1 file changed, 32 insertions(+), 2 deletions(-)content.tex +21
-0@@ -49,6 +49,9 @@ \section{\field{Device Status} Field}\label{sec:Basic Facilities of a Virtio Dev \item[DEVICE_NEEDS_RESET (64)] Indicates that the device has experienced an error from which it can't recover. + +\item[SUSPEND (16)] When VIRTIO_F_SUSPEND is negotiated, indicates that the + device has been suspended by the driver. \end{description} The \field{device status} field starts out as 0, and is reinitialized to 0 by@@ -73,6 +76,10 @@ \section{\field{Device Status} Field}\label{sec:Basic Facilities of a Virtio Dev recover by issuing a reset. \end{note} +The driver SHOULD NOT set SUSPEND if FEATURES_OK is not set. + +When setting SUSPEND, the driver MUST re-read \field{device status} to ensure the SUSPEND bit is set. + \devicenormative{\subsection}{Device Status Field}{Basic Facilities of a Virtio Device / Device Status Field} The device MUST NOT consume buffers or send any used buffer@@ -82,6 +89,25 @@ \section{\field{Device Status} Field}\label{sec:Basic Facilities of a Virtio Dev that a reset is needed. If DRIVER_OK is set, after it sets DEVICE_NEEDS_RESET, the device MUST send a device configuration change notification to the driver. +The device MUST ignore SUSPEND if FEATURES_OK is not set. + +The device MUST ignore SUSPEND if VIRTIO_F_SUSPEND is not negotiated. + +The device SHOULD allow settings to \field{device status} even when SUSPEND is set. + +If VIRTIO_F_SUSPEND is negotiated and SUSPEND is set, the device SHOULD clear SUSPEND +and resumes operation upon DRIVER_OK. + +If VIRTIO_F_SUSPEND is negotiated, when the driver sets SUSPEND, +the device SHOULD perform the following actions before presenting SUSPEND bit in the \field{device status}: + +\begin{itemize} +\item Stop consuming buffers of any virtqueues and mark all finished descriptors as used.Quoted 8 lines
Quoted 2 lines
content.tex \section{\field{Device Status} Field}\label{sec:Basic Facilities of a Virtio Dev · line 107
Quoted 2 lines
content.tex +2
-0+\item Wait until all descriptors that being processed to finish and mark them as used. +\item Flush all used buffer and send used buffer notifications to the driver.What's the difference between the previous three lines? They might be trying to say different things, but there is a lot of overlap in how they're phrased. That makes it hard to figure out exactly what they're mandating. Is it something like: "stop processing new buffers", "finish processing any buffers that are currently in flight", "mark all finished buffers as used and send a used buffer notification"?
sure, the "mark used" parts can merge.
Quoted 4 lines
Also, what does this mean for devices where the driver places empty buffers into the virtqueue that the device then holds on to for an indeterminate period of time (e.g. network receiveq, balloon statsq, etc)?
The device doesn't process them.
Oh, I should add: The driver should not make any more buffers available to the device.Quoted 7 lines
Quoted 4 lines
Quoted 1 line
content.tex \section{\field{Device Status} Field}\label{sec:Basic Facilities of a Virtio Dev · line 108
Quoted 1 line
content.tex +1
-0+\item Pause its operation except \field{device status} and preserve configurations in its Device Configuration Space, see \ref{sec:Basic Facilities of a Virtio Device / Device Configuration Space}What does "Pause its operation except device status" mean? The word "operation" is vague, what exactly does it include? Also what does "operate its device status" mean?
Because we need to resume the device after suspending it by setting
DRIVER_OK, so we need the device status
filed keel alive.Quoted 7 lines
This doesn't answer what "Pause its operation" means. If the specification is not precise, then the best that can be implemented is for a specific device to be compatible with a specific driver. If we actually want portable, spec-compliant devices, we need explicitly defined MUST/MUST NOT requirements. Presumably we want something like:
A device MUST not send notifications while suspended.
The device should send a last used buffer notification if it has buffers under processing before presenting SUSPEND in the device status, but the device should not process any new buffers. These are in devicenormative section, but they may be vague, I can add more details to make it precise, thanks for the tips
Quoted 4 lines
A device MUST ignore writes to its device configuration while suspended, except for writes to the device status byte.
A driver MUST NOT write to a suspended device's configuration space, except for the device status byte.sure, we can make it more precise than only pausing.
Quoted 11 lines
We also need to specify what a suspended device is allowed to do with buffers it owns. For my use case of just suspending/resuming a VM, I don't think it particularly matters if a suspended device writes to buffers or descriptors while it is suspended. But based on what you wrote about finishing any in-flight processing of buffers and not processing any empty buffers, presumably you would prefer that the device not be allowed to access buffer/descriptors while it is suspended? That approach is probably cleaner from a pure specification standpoint, but it's also more restrictive for devices and thus harder to properly implement.
When the driver sets SUSPEND, the device should finished all buffers that already under processing, include the in-flight buffers(or it may lose the buffers).
And the device should not touch the buffers once it present SUSPEND in the device status field, because once it presents SUSPEND, the device state should
be stable, people may use SUSPEND to debug the device.
Quoted 6 lines
The device still can process the buffer before it presenting SUSPEND.
Thanks
Zhu LingshanQuoted 2 lines
-David
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
Quoted 37 lines · Michael S. Tsirkin
Quoted 36 lines · Lingshan
Quoted 21 lines · David Stevens
Quoted 20 lines · Lingshan
Quoted 19 lines · David Stevens
Quoted 18 lines · Michael S. Tsirkin
Quoted 11 lines · Zhu Lingshan
This commit allows the driver to suspend the device by introducing a new status bit SUSPEND in
device_status.This commit also introduce a new feature bit
VIRTIO_F_SUSPENDwhich indicating whether the device support SUSPEND.This SUSPEND bit is transport-independent.
Signed-off-by Zhu Lingshan <[email protected]>Signed-off-by Jason Wang <[email protected]>Signed-off-by Eugenio Pérez <[email protected]>Could we get some kind of dscription how this has taken into consideration the proposal from David Stevens?
I find it really tiring when there are competing patches with authors ignoring each other's work and leaving it up to reviewers to figure out how do the patches compare.
Quoted 7 lines
Quoted 6 lines
Quoted 4 lines
This patch looks like it could be used to implement my use case. However, parts of it are a bit vague and imprecise, so it's hard to actually say whether my use case would actually be covered by a specific implementation of this proposal.
I am on vacation till this Friday. Shall we co-work on this and post something new together?
That works for me.
Cool! How about two patches in a new series:
1) a general virtio suspending patch describing suspending behaviors for all virtio devices from me, an updating version of this patch
2) a PCI transport suspending patch from you, describing PCI specific behaviors.
Quoted 4 lines
Quoted 1 line
Dose this sound good to you?
Well David's patch was more precise than yours, and also better worded. So think if there's an agreement you guys would really start with that, move the functionality to the status bit and make other changes as needed.
I think we plan to cooperate on a new series including both the status bit and PCI transport, collaborations on both of our patches.
Let's see whether this looks good to David
Thanks
Quoted 129 lines
Quoted 129 lines
Quoted 14 lines
Quoted 14 lines
Quoted 9 lines
To give specifics on what I'm trying to do, I need to allow a guest with virtio-pci devices (including stateful devices like virtio-fs and virtio-gpu) to be put into S1/S3, and to allow the virtio-pci devices to wake up the guest (currently via an ACPI GPE, but eventually via a native PCI PME). This is all done on a consumer device, so there is no need for snapshotting or for live migration.
Suspending is not dedicated only for live migration.
For your use case, shall we add a new PCI section like saying: when entering SUSPEND, the device should get into S1/S3 and other PCI specific steps in PCI transport charter.Quoted 6 lines
Quoted 2 lines
That is because PCI is a transport layer of virtio, controlling virtio states by PCI sounds like a layer violation.
I don't know if it's really necessary to mandate that in the spec. Sure, most systems are probably going to want to put the virtio-pci device into S1/S3 after suspending virtio operation. But nothing really breaks or goes wrong if we don't do that.
Yes, maybe not S1/S3, I remember you want to manage power in PCI transport when suspend in a cap, S1/S3 is just an example.
Quoted 70 lines
Quoted 63 lines
Quoted 47 lines
Quoted 47 lines
Quoted 47 lines
--- content.tex | 34 ++++++++++++++++++++++++++++++++-- 1 file changed, 32 insertions(+), 2 deletions(-)content.tex +21
-0@@ -49,6 +49,9 @@ \section{\field{Device Status} Field}\label{sec:Basic Facilities of a Virtio Dev \item[DEVICE_NEEDS_RESET (64)] Indicates that the device has experienced an error from which it can't recover. + +\item[SUSPEND (16)] When VIRTIO_F_SUSPEND is negotiated, indicates that the + device has been suspended by the driver. \end{description} The \field{device status} field starts out as 0, and is reinitialized to 0 by@@ -73,6 +76,10 @@ \section{\field{Device Status} Field}\label{sec:Basic Facilities of a Virtio Dev recover by issuing a reset. \end{note} +The driver SHOULD NOT set SUSPEND if FEATURES_OK is not set. + +When setting SUSPEND, the driver MUST re-read \field{device status} to ensure the SUSPEND bit is set. + \devicenormative{\subsection}{Device Status Field}{Basic Facilities of a Virtio Device / Device Status Field} The device MUST NOT consume buffers or send any used buffer@@ -82,6 +89,25 @@ \section{\field{Device Status} Field}\label{sec:Basic Facilities of a Virtio Dev that a reset is needed. If DRIVER_OK is set, after it sets DEVICE_NEEDS_RESET, the device MUST send a device configuration change notification to the driver. +The device MUST ignore SUSPEND if FEATURES_OK is not set. + +The device MUST ignore SUSPEND if VIRTIO_F_SUSPEND is not negotiated. + +The device SHOULD allow settings to \field{device status} even when SUSPEND is set. + +If VIRTIO_F_SUSPEND is negotiated and SUSPEND is set, the device SHOULD clear SUSPEND +and resumes operation upon DRIVER_OK. + +If VIRTIO_F_SUSPEND is negotiated, when the driver sets SUSPEND, +the device SHOULD perform the following actions before presenting SUSPEND bit in the \field{device status}: + +\begin{itemize} +\item Stop consuming buffers of any virtqueues and mark all finished descriptors as used.Quoted 8 lines
Quoted 2 lines
content.tex \section{\field{Device Status} Field}\label{sec:Basic Facilities of a Virtio Dev · line 107
Quoted 2 lines
content.tex +2
-0+\item Wait until all descriptors that being processed to finish and mark them as used. +\item Flush all used buffer and send used buffer notifications to the driver.What's the difference between the previous three lines? They might be trying to say different things, but there is a lot of overlap in how they're phrased. That makes it hard to figure out exactly what they're mandating. Is it something like: "stop processing new buffers", "finish processing any buffers that are currently in flight", "mark all finished buffers as used and send a used buffer notification"?
sure, the "mark used" parts can merge.
Quoted 4 lines
Also, what does this mean for devices where the driver places empty buffers into the virtqueue that the device then holds on to for an indeterminate period of time (e.g. network receiveq, balloon statsq, etc)?
The device doesn't process them.
Oh, I should add: The driver should not make any more buffers available to the device.Quoted 7 lines
Quoted 4 lines
Quoted 1 line
content.tex \section{\field{Device Status} Field}\label{sec:Basic Facilities of a Virtio Dev · line 108
Quoted 1 line
content.tex +1
-0+\item Pause its operation except \field{device status} and preserve configurations in its Device Configuration Space, see \ref{sec:Basic Facilities of a Virtio Device / Device Configuration Space}What does "Pause its operation except device status" mean? The word "operation" is vague, what exactly does it include? Also what does "operate its device status" mean?
Because we need to resume the device after suspending it by setting
DRIVER_OK, so we need the device status
filed keel alive.Quoted 7 lines
This doesn't answer what "Pause its operation" means. If the specification is not precise, then the best that can be implemented is for a specific device to be compatible with a specific driver. If we actually want portable, spec-compliant devices, we need explicitly defined MUST/MUST NOT requirements. Presumably we want something like:
A device MUST not send notifications while suspended.
The device should send a last used buffer notification if it has buffers under processing before presenting SUSPEND in the device status, but the device should not process any new buffers. These are in devicenormative section, but they may be vague, I can add more details to make it precise, thanks for the tips
Quoted 4 lines
A device MUST ignore writes to its device configuration while suspended, except for writes to the device status byte.
A driver MUST NOT write to a suspended device's configuration space, except for the device status byte.sure, we can make it more precise than only pausing.
Quoted 10 lines
We also need to specify what a suspended device is allowed to do with buffers it owns. For my use case of just suspending/resuming a VM, I don't think it particularly matters if a suspended device writes to buffers or descriptors while it is suspended. But based on what you wrote about finishing any in-flight processing of buffers and not processing any empty buffers, presumably you would prefer that the device not be allowed to access buffer/descriptors while it is suspended? That approach is probably cleaner from a pure specification standpoint, but it's also more restrictive for devices and thus harder to properly implement.
When the driver sets SUSPEND, the device should finished all buffers that already under processing, include the in-flight buffers(or it may lose the buffers).
And the device should not touch the buffers once it present SUSPEND in the device status field, because once it presents SUSPEND, the device state should
be stable, people may use SUSPEND to debug the device.
Quoted 5 lines
Quoted 5 lines
The device still can process the buffer before it presenting SUSPEND.
Thanks
Zhu LingshanQuoted 1 line
-David
This publicly archived list offers a means to provide input to the OASIS Virtual I/O Device (VIRTIO) TC.
In order to verify user consent to the Feedback License terms and to minimize spam in the list archive, subscription is required before posting.
Subscribe: [email protected]
Unsubscribe: [email protected]
List help: [email protected]
List archive: https://lists.oasis-open.org/archives/virtio-comment/
Feedback License: https://www.oasis-open.org/who/ipr/feedback_license.pdf
List Guidelines: https://www.oasis-open.org/policies-guidelines/mailing-lists
Committee: https://www.oasis-open.org/committees/virtio/
Join OASIS: https://www.oasis-open.org/join/
Quoted 39 lines · Lingshan
Quoted 38 lines · Michael S. Tsirkin
Quoted 37 lines · Lingshan
Quoted 21 lines · David Stevens
Quoted 20 lines · Lingshan
Quoted 19 lines · David Stevens
Quoted 18 lines · Michael S. Tsirkin
Quoted 11 lines · Zhu Lingshan
This commit allows the driver to suspend the device by introducing a new status bit SUSPEND in
device_status.This commit also introduce a new feature bit
VIRTIO_F_SUSPENDwhich indicating whether the device support SUSPEND.This SUSPEND bit is transport-independent.
Signed-off-by Zhu Lingshan <[email protected]>Signed-off-by Jason Wang <[email protected]>Signed-off-by Eugenio Pérez <[email protected]>Could we get some kind of dscription how this has taken into consideration the proposal from David Stevens?
I find it really tiring when there are competing patches with authors ignoring each other's work and leaving it up to reviewers to figure out how do the patches compare.
Quoted 7 lines
Quoted 6 lines
Quoted 4 lines
This patch looks like it could be used to implement my use case. However, parts of it are a bit vague and imprecise, so it's hard to actually say whether my use case would actually be covered by a specific implementation of this proposal.
I am on vacation till this Friday. Shall we co-work on this and post something new together?
That works for me.
Cool! How about two patches in a new series:
1) a general virtio suspending patch describing suspending behaviors for all virtio devices from me, an updating version of this patch
2) a PCI transport suspending patch from you, describing PCI specific behaviors.Dose this sound good to you?
Quoted 8 lines
Quoted 3 lines
Well David's patch was more precise than yours, and also better worded. So think if there's an agreement you guys would really start with that, move the functionality to the status bit and make other changes as needed.
I think we plan to cooperate on a new series including both the status bit and
PCI transport, collaborations on both of our patches.Let's see whether this looks good to David
That works for me. I can incorporate the feedback on both proposals and send out a combined version in the next day or so.
Thanks,
David
This publicly archived list offers a means to provide input to the OASIS Virtual I/O Device (VIRTIO) TC.
In order to verify user consent to the Feedback License terms and to minimize spam in the list archive, subscription is required before posting.
Subscribe: [email protected]
Unsubscribe: [email protected]
List help: [email protected]
List archive: https://lists.oasis-open.org/archives/virtio-comment/
Feedback License: https://www.oasis-open.org/who/ipr/feedback_license.pdf
List Guidelines: https://www.oasis-open.org/policies-guidelines/mailing-lists
Committee: https://www.oasis-open.org/committees/virtio/
Join OASIS: https://www.oasis-open.org/join/
Quoted 9 lines · Michael S. Tsirkin
Quoted 8 lines · Zhu Lingshan
This commit allows the driver to suspend the device by introducing a new status bit SUSPEND in
device_status.This commit also introduce a new feature bit
VIRTIO_F_SUSPENDwhich indicating whether the device support SUSPEND.This SUSPEND bit is transport-independent.
Quoted 7 lines
Quoted 3 lines
Signed-off-by Zhu Lingshan <[email protected]>Signed-off-by Jason Wang <[email protected]>Signed-off-by Eugenio Pérez <[email protected]>Could we get some kind of dscription how this has taken into consideration the proposal from David Stevens?
sure thing, we can work with David together for a solution.
Quoted 83 lines
I find it really tiring when there are competing patches with authors ignoring each other's work and leaving it up to reviewers to figure out how do the patches compare.
Quoted 78 lines
--- content.tex | 34 ++++++++++++++++++++++++++++++++-- 1 file changed, 32 insertions(+), 2 deletions(-)content.tex +32
-2@@ -49,6 +49,9 @@ \section{\field{Device Status} Field}\label{sec:Basic Facilities of a Virtio Dev \item[DEVICE_NEEDS_RESET (64)] Indicates that the device has experienced an error from which it can't recover. + +\item[SUSPEND (16)] When VIRTIO_F_SUSPEND is negotiated, indicates that the + device has been suspended by the driver. \end{description} The \field{device status} field starts out as 0, and is reinitialized to 0 by@@ -73,6 +76,10 @@ \section{\field{Device Status} Field}\label{sec:Basic Facilities of a Virtio Dev recover by issuing a reset. \end{note} +The driver SHOULD NOT set SUSPEND if FEATURES_OK is not set. + +When setting SUSPEND, the driver MUST re-read \field{device status} to ensure the SUSPEND bit is set. + \devicenormative{\subsection}{Device Status Field}{Basic Facilities of a Virtio Device / Device Status Field} The device MUST NOT consume buffers or send any used buffer@@ -82,6 +89,25 @@ \section{\field{Device Status} Field}\label{sec:Basic Facilities of a Virtio Dev that a reset is needed. If DRIVER_OK is set, after it sets DEVICE_NEEDS_RESET, the device MUST send a device configuration change notification to the driver. +The device MUST ignore SUSPEND if FEATURES_OK is not set. + +The device MUST ignore SUSPEND if VIRTIO_F_SUSPEND is not negotiated. + +The device SHOULD allow settings to \field{device status} even when SUSPEND is set. + +If VIRTIO_F_SUSPEND is negotiated and SUSPEND is set, the device SHOULD clear SUSPEND +and resumes operation upon DRIVER_OK. + +If VIRTIO_F_SUSPEND is negotiated, when the driver sets SUSPEND, +the device SHOULD perform the following actions before presenting SUSPEND bit in the \field{device status}: + +\begin{itemize} +\item Stop consuming buffers of any virtqueues and mark all finished descriptors as used. +\item Wait until all descriptors that being processed to finish and mark them as used. +\item Flush all used buffer and send used buffer notifications to the driver. +\item Pause its operation except \field{device status} and preserve configurations in its Device Configuration Space, see \ref{sec:Basic Facilities of a Virtio Device / Device Configuration Space} +\end{itemize} + \section{Feature Bits}\label{sec:Basic Facilities of a Virtio Device / Feature Bits} Each virtio device offers all the features it understands. During@@ -99,10 +125,10 @@ \section{Feature Bits}\label{sec:Basic Facilities of a Virtio Device / Feature B \begin{description} \item[0 to 23, and 50 to 127] Feature bits for the specific device type -\item[24 to 41] Feature bits reserved for extensions to the queue and +\item[24 to 42] Feature bits reserved for extensions to the queue and feature negotiation mechanisms -\item[42 to 49, and 128 and above] Feature bits reserved for future extensions. +\item[43 to 49, and 128 and above] Feature bits reserved for future extensions. \end{description} \begin{note}@@ -872,6 +898,10 @@ \chapter{Reserved Feature Bits}\label{sec:Reserved Feature Bits} \ref{devicenormative:Basic Facilities of a Virtio Device / Feature Bits} for handling features reserved for future use. + \item[VIRTIO_F_SUSPEND(42)] This feature indicates that the driver can + SUSPEND the device. + See \ref{sec:Basic Facilities of a Virtio Device / Device Status Field}. + \end{description}
Quoted 3 lines
Quoted 3 lines
content.tex +0
-0\drivernormative{\section}{Reserved Feature Bits}{Reserved Feature Bits}Signature · 1 line
2.35.3
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]