virtio-mmio: Specify wait needed in driver during reset

Reset of a virtio-mmio device is initiated by writing 0 to its Status register. The reset operation itself may or may not be completed by the time write instruction completes. For devices that are emulated in software, writes to Status register are trapped and resumed only after the reset operation is complete. Thus a driver can be assured of reset completion as soon as its write completes. That may not be however true in other cases, such as when virtio-mmio devices implemented directly in hardware. Driver's write to Status register in such case could complete before the device reset operation is completed.

Update the specification to indicate when a driver may need to poll for reset completion before going ahead with the remaining initialization steps.

Suggested-by Jason Wang <[email protected]>

Signed-off-by Srivatsa Vaddagiri <[email protected]>

---
V2->V1:
- Use version 0x3 as an indication for drivers to poll for reset completion (rather than base it on a new feature bit)

Previous version can be found at:

https://lists.oasis-open.org/archives/virtio-dev/202107/msg00168.html

 content.tex | 11 +++++++----
 1 file changed, 7 insertions(+), 4 deletions(-)

content.tex +3 -3

index 7cec1c3..b6b30a0 100644
@@ -1730,9 +1730,9 @@ \subsection{MMIO Device Register Layout}\label{sec:Virtio Transport Options / Vi
   }
   \hline
   \mmioreg{Version}{Device version number}{0x004}{R}{%
-    0x2.
+    0x3.
     \begin{note}
-      Legacy devices (see \ref{sec:Virtio Transport Options / Virtio Over MMIO / Legacy interface}~\nameref{sec:Virtio Transport Options / Virtio Over MMIO / Legacy interface}) used 0x1.
+      Legacy devices (see \ref{sec:Virtio Transport Options / Virtio Over MMIO / Legacy interface}~\nameref{sec:Virtio Transport Options / Virtio Over MMIO / Legacy interface}) used 0x1. Devices that do not require drivers to poll for reset completion can use 0x2. See \ref{devicenormative:Virtio Transport Options / Virtio Over MMIO / MMIO Device Register Layout} for more details.
     \end{note}
   }
   \hline

@@ -1916,7 +1916,7 @@ \subsection{MMIO Device Register Layout}\label{sec:Virtio Transport Options / Vi
 
 The device MUST return 0x74726976 in \field{MagicValue}.
 
-The device MUST return value 0x2 in \field{Version}.
+The device MUST return either value 0x3 or 0x2 in \field{Version} based on its reset behavior. Drivers trigger reset of a device by writing 0 to \field{Status}. The reset operation itself may or may not be completed by the time write operation is complete. Devices whose reset operation completes synchronously with the write operation are allowed to return value of 0x2 for \field{Version}. Other devices, whose reset operation can be incomplete by the time write operation completes MUST return value 0x3 as an indication for drivers to poll for reset completion.
 
 The device MUST present each event by setting the corresponding bit in \field{InterruptStatus} from the
 moment it takes place, until the driver acknowledges the interrupt

@@ -1947,12 +1947,15 @@ \subsection{MMIO Device Register Layout}\label{sec:Virtio Transport Options / Vi
 The driver MUST ignore a device with \field{MagicValue} which is not 0x74726976,
 although it MAY report an error.
 

content.tex +0 -1

 Before reading from \field{DeviceFeatures}, the driver MUST write a value to \field{DeviceFeaturesSel}.
 
 Before writing to the \field{DriverFeatures} register, the driver MUST write a value to the \field{DriverFeaturesSel} register.
 
 
---

Qualcomm Innovation Center, Inc. is submitting the attached "feedback" as a non-member to the virtio-dev mailing list for consideration and inclusion.

content.tex \subsection{MMIO Device Register Layout}\label{sec:Virtio Transport Options / Vi · line 1958

Quoted 3 lines

content.tex +3 -0

+After writing 0 to \field{Status}, which triggers a reset of device, the driver MUST wait for a read of
+\field{Status} to return 0 before proceeding with the remaining steps of initializing the device.
+

Do we need to limit this behaviour to 0x3 only?

Quoted 5 lines · Jason Wang

content.tex \subsection{MMIO Device Register Layout}\label{sec:Virtio Transport Options / Vi · line 1958

Quoted 3 lines

content.tex +3 -0

+After writing 0 to \field{Status}, which triggers a reset of device, the driver MUST wait for a read of
+\field{Status} to return 0 before proceeding with the remaining steps of initializing the device.
+

Do we need to limit this behaviour to 0x3 only?

I felt it will be simple for drivers to unconditionally poll for reset completion (which is the right thing to do IMO), but can make it conditional to version 3 only if you say so.

- vatsa

---

Qualcomm Innovation Center, Inc. is submitting the attached "feedback" as a non-member to the virtio-dev mailing list for consideration and inclusion.

Reset of a virtio-mmio device is initiated by writing 0 to its Status register. The reset operation itself may or may not be completed by the time write instruction completes. For devices that are emulated in software, writes to Status register are trapped and resumed only after the reset operation is complete. Thus a driver can be assured of reset completion as soon as its write completes. That may not be however true in other cases, such as when virtio-mmio devices implemented directly in hardware. Driver's write to Status register in such case could complete before the device reset operation is completed.

Update the specification to indicate when a driver may need to poll for reset completion before going ahead with the remaining initialization steps.

Suggested-by Jason Wang <[email protected]>

Signed-off-by Srivatsa Vaddagiri <[email protected]>

---
V2->V1:
- Use version 0x3 as an indication for drivers to poll for reset completion (rather than base it on a new feature bit)

Previous version can be found at:

https://lists.oasis-open.org/archives/virtio-dev/202107/msg00168.html

 content.tex | 11 +++++++----
 1 file changed, 7 insertions(+), 4 deletions(-)

content.tex +3 -3

index 7cec1c3..b6b30a0 100644
@@ -1730,9 +1730,9 @@ \subsection{MMIO Device Register Layout}\label{sec:Virtio Transport Options / Vi
   } 
   \hline
   \mmioreg{Version}{Device version number}{0x004}{R}{%
-    0x2.
+    0x3.
     \begin{note}
-      Legacy devices (see \ref{sec:Virtio Transport Options / Virtio Over MMIO / Legacy interface}~\nameref{sec:Virtio Transport Options / Virtio Over MMIO / Legacy interface}) used 0x1.
+      Legacy devices (see \ref{sec:Virtio Transport Options / Virtio Over MMIO / Legacy interface}~\nameref{sec:Virtio Transport Options / Virtio Over MMIO / Legacy interface}) used 0x1. Devices that do not require drivers to poll for reset completion can use 0x2. See \ref{devicenormative:Virtio Transport Options / Virtio Over MMIO / MMIO Device Register Layout} for more details.
     \end{note}
   }
   \hline 

@@ -1916,7 +1916,7 @@ \subsection{MMIO Device Register Layout}\label{sec:Virtio Transport Options / Vi
 
 The device MUST return 0x74726976 in \field{MagicValue}.
 
-The device MUST return value 0x2 in \field{Version}.
+The device MUST return either value 0x3 or 0x2 in \field{Version} based on its reset behavior. Drivers trigger reset of a device by writing 0 to \field{Status}. The reset operation itself may or may not be completed by the time write operation is complete. Devices whose reset operation completes synchronously with the write operation are allowed to return value of 0x2 for \field{Version}. Other devices, whose reset operation can be incomplete by the time write operation completes MUST return value 0x3 as an indication for drivers to poll for reset completion.
 
 The device MUST present each event by setting the corresponding bit in \field{InterruptStatus} from the
 moment it takes place, until the driver acknowledges the interrupt

@@ -1947,12 +1947,15 @@ \subsection{MMIO Device Register Layout}\label{sec:Virtio Transport Options / Vi
 The driver MUST ignore a device with \field{MagicValue} which is not 0x74726976,
 although it MAY report an error.

content.tex +1 -1

+
 Before reading from \field{DeviceFeatures}, the driver MUST write a value to \field{DeviceFeaturesSel}.
 
 Before writing to the \field{DriverFeatures} register, the driver MUST write a value to the \field{DriverFeaturesSel} register.
 
 
---

Qualcomm Innovation Center, Inc. is submitting the attached "feedback" as a non-member to the virtio-dev mailing list for consideration and inclusion.

Quoted 2 lines

+After writing 0 to \field{Status}, which triggers a reset of device, the driver MUST wait for a read of +\field{Status} to return 0 before proceeding with the remaining steps of initializing the device.

I know you copied this from pci but looking at PCI now, I think this is not great there either.

1. This is IMHO too opaque in that we did not say driver should read.

E.g. after writing 0 to Status and before reading any fields, the driver MUST read Status and wait for a read etc.

also would not just a read be sufficient? Is there a case for device to return any value other than 0 to signal "reset in progress"?

2. what if device encounters an error and sets NEED_RESET meanwhile? waiting for a reset will then deadlock ... I know, it's problematic like this in PCI too.

Further what will device return after reset was written but before it completed? I think it is better to specify that rather than rely on any non-0 value meaning "wait for reset".

For pci maybe make it a SHOULD ...

Quoted 55 lines · Srivatsa Vaddagiri

Reset of a virtio-mmio device is initiated by writing 0 to its Status register. The reset operation itself may or may not be completed by the time write instruction completes. For devices that are emulated in software, writes to Status register are trapped and resumed only after the reset operation is complete. Thus a driver can be assured of reset completion as soon as its write completes. That may not be however true in other cases, such as when virtio-mmio devices implemented directly in hardware. Driver's write to Status register in such case could complete before the device reset operation is completed.

Update the specification to indicate when a driver may need to poll for reset completion before going ahead with the remaining initialization steps.

Suggested-by Jason Wang <[email protected]>

Signed-off-by Srivatsa Vaddagiri <[email protected]>

---
V2->V1:
- Use version 0x3 as an indication for drivers to poll for reset completion (rather than base it on a new feature bit)

Previous version can be found at:

https://lists.oasis-open.org/archives/virtio-dev/202107/msg00168.html

 content.tex | 11 +++++++----
 1 file changed, 7 insertions(+), 4 deletions(-)

content.tex +3 -3

index 7cec1c3..b6b30a0 100644
@@ -1730,9 +1730,9 @@ \subsection{MMIO Device Register Layout}\label{sec:Virtio Transport Options / Vi
   }
   \hline
   \mmioreg{Version}{Device version number}{0x004}{R}{%
-    0x2.
+    0x3.
     \begin{note}
-      Legacy devices (see \ref{sec:Virtio Transport Options / Virtio Over MMIO / Legacy interface}~\nameref{sec:Virtio Transport Options / Virtio Over MMIO / Legacy interface}) used 0x1.
+      Legacy devices (see \ref{sec:Virtio Transport Options / Virtio Over MMIO / Legacy interface}~\nameref{sec:Virtio Transport Options / Virtio Over MMIO / Legacy interface}) used 0x1. Devices that do not require drivers to poll for reset completion can use 0x2. See \ref{devicenormative:Virtio Transport Options / Virtio Over MMIO / MMIO Device Register Layout} for more details.
     \end{note}
   }
   \hline

@@ -1916,7 +1916,7 @@ \subsection{MMIO Device Register Layout}\label{sec:Virtio Transport Options / Vi
 
 The device MUST return 0x74726976 in \field{MagicValue}.
 
-The device MUST return value 0x2 in \field{Version}.
+The device MUST return either value 0x3 or 0x2 in \field{Version} based on its reset behavior. Drivers trigger reset of a device by writing 0 to \field{Status}. The reset operation itself may or may not be completed by the time write operation is complete. Devices whose reset operation completes synchronously with the write operation are allowed to return value of 0x2 for \field{Version}. Other devices, whose reset operation can be incomplete by the time write operation completes MUST return value 0x3 as an indication for drivers to poll for reset completion.
 
 The device MUST present each event by setting the corresponding bit in \field{InterruptStatus} from the
 moment it takes place, until the driver acknowledges the interrupt

@@ -1947,12 +1947,15 @@ \subsection{MMIO Device Register Layout}\label{sec:Virtio Transport Options / Vi
 The driver MUST ignore a device with \field{MagicValue} which is not 0x74726976,
 although it MAY report an error.

content.tex \subsection{MMIO Device Register Layout}\label{sec:Virtio Transport Options / Vi · line 1957

Quoted 10 lines

content.tex +3 -1

 
-The driver MUST ignore a device with \field{Version} which is not 0x2,
+The driver MUST ignore a device with \field{Version} which is neither 0x2 nor 0x3,
 although it MAY report an error.
 
 The driver MUST ignore a device with \field{DeviceID} 0x0,
 but MUST NOT report any error.
 
+After writing 0 to \field{Status}, which triggers a reset of device, the driver MUST wait for a read of
+\field{Status} to return 0 before proceeding with the remaining steps of initializing the device.

I know you copied this from pci but looking at PCI now, I think this is not great there either.

1. This is IMHO too opaque in that we did not say driver should read.

E.g. after writing 0 to Status and before reading any fields, the driver MUST read Status and wait for a read etc.

also would not just a read be sufficient? Is there a case for device to return any value other than 0 to signal "reset in progress"?

2. what if device encounters an error and sets NEED_RESET meanwhile? waiting for a reset will then deadlock ... I know, it's problematic like this in PCI too.

Further what will device return after reset was written but before it completed? I think it is better to specify that rather than rely on any non-0 value meaning "wait for reset". For pci maybe make it a SHOULD ...

Quoted 11 lines

content.tex +1 -1

+
 Before reading from \field{DeviceFeatures}, the driver MUST write a value to \field{DeviceFeaturesSel}.
 
 Before writing to the \field{DriverFeatures} register, the driver MUST write a value to the \field{DriverFeaturesSel} register.
 
 
---

Qualcomm Innovation Center, Inc. is submitting the attached "feedback" as a non-member to the virtio-dev mailing list for consideration and inclusion.

Reset of a virtio-mmio device is initiated by writing 0 to its Status register. The reset operation itself may or may not be completed by the time write instruction completes. For devices that are emulated in software, writes to Status register are trapped and resumed only after the reset operation is complete. Thus a driver can be assured of reset completion as soon as its write completes. That may not be however true in other cases, such as when virtio-mmio devices implemented directly in hardware. Driver's write to Status register in such case could complete before the device reset operation is completed.

Update the specification to indicate when a driver may need to poll for reset completion before going ahead with the remaining initialization steps.

Suggested-by Jason Wang <[email protected]>

Signed-off-by Srivatsa Vaddagiri <[email protected]>

---
V2->V1:
- Use version 0x3 as an indication for drivers to poll for reset completion (rather than base it on a new feature bit)

Previous version can be found at:

https://lists.oasis-open.org/archives/virtio-dev/202107/msg00168.html

 content.tex | 11 +++++++----
 1 file changed, 7 insertions(+), 4 deletions(-)

content.tex +2 -2

index 7cec1c3..b6b30a0 100644
@@ -1730,9 +1730,9 @@ \subsection{MMIO Device Register Layout}\label{sec:Virtio Transport Options / Vi
   } 
   \hline
   \mmioreg{Version}{Device version number}{0x004}{R}{%
-    0x2.
+    0x3.
     \begin{note}
-      Legacy devices (see \ref{sec:Virtio Transport Options / Virtio Over MMIO / Legacy interface}~\nameref{sec:Virtio Transport Options / Virtio Over MMIO / Legacy interface}) used 0x1.
+      Legacy devices (see \ref{sec:Virtio Transport Options / Virtio Over MMIO / Legacy interface}~\nameref{sec:Virtio Transport Options / Virtio Over MMIO / Legacy interface}) used 0x1. Devices that do not require drivers to poll for reset completion can use 0x2. See \ref{devicenormative:Virtio Transport Options / Virtio Over MMIO / MMIO Device Register Layout} for more details.
     \end{note}
   }
   \hline 

Quoted 47 lines · Srivatsa Vaddagiri

Reset of a virtio-mmio device is initiated by writing 0 to its Status register. The reset operation itself may or may not be completed by the time write instruction completes. For devices that are emulated in software, writes to Status register are trapped and resumed only after the reset operation is complete. Thus a driver can be assured of reset completion as soon as its write completes. That may not be however true in other cases, such as when virtio-mmio devices implemented directly in hardware. Driver's write to Status register in such case could complete before the device reset operation is completed.

Update the specification to indicate when a driver may need to poll for reset completion before going ahead with the remaining initialization steps.

Suggested-by Jason Wang <[email protected]>

Signed-off-by Srivatsa Vaddagiri <[email protected]>

---
V2->V1:
- Use version 0x3 as an indication for drivers to poll for reset completion (rather than base it on a new feature bit)

Previous version can be found at:

https://lists.oasis-open.org/archives/virtio-dev/202107/msg00168.html

  content.tex | 11 +++++++----
  1 file changed, 7 insertions(+), 4 deletions(-)

content.tex +2 -2

index 7cec1c3..b6b30a0 100644
@@ -1730,9 +1730,9 @@ \subsection{MMIO Device Register Layout}\label{sec:Virtio Transport Options / Vi
    }
    \hline
    \mmioreg{Version}{Device version number}{0x004}{R}{%
-    0x2.
+    0x3.
      \begin{note}
-      Legacy devices (see \ref{sec:Virtio Transport Options / Virtio Over MMIO / Legacy interface}~\nameref{sec:Virtio Transport Options / Virtio Over MMIO / Legacy interface}) used 0x1.
+      Legacy devices (see \ref{sec:Virtio Transport Options / Virtio Over MMIO / Legacy interface}~\nameref{sec:Virtio Transport Options / Virtio Over MMIO / Legacy interface}) used 0x1. Devices that do not require drivers to poll for reset completion can use 0x2. See \ref{devicenormative:Virtio Transport Options / Virtio Over MMIO / MMIO Device Register Layout} for more details.
      \end{note}
    }
    \hline

@@ -1916,7 +1916,7 @@ \subsection{MMIO Device Register Layout}\label{sec:Virtio Transport Options / Vi
 
  The device MUST return 0x74726976 in \field{MagicValue}.
 

content.tex \subsection{MMIO Device Register Layout}\label{sec:Virtio Transport Options / Vi · line 1919

Quoted 2 lines

content.tex +1 -1

-The device MUST return value 0x2 in \field{Version}.
+The device MUST return either value 0x3 or 0x2 in \field{Version} based on its reset behavior. Drivers trigger reset of a device by writing 0 to \field{Status}. The reset operation itself may or may not be completed by the time write operation is complete. Devices whose reset operation completes synchronously with the write operation are allowed to return value of 0x2 for \field{Version}. Other devices, whose reset operation can be incomplete by the time write operation completes MUST return value 0x3 as an indication for drivers to poll for reset completion.

Maybe I'm overthining this, but might that suddenly render some existing device non-compliant? IIUC, we never mandated reset to be completed when the write was completed, only kind of implied it. An existing device will have version 2, but it might not be quite finished with reset after the status write yet.

Quoted 49 lines · Cornelia Huck

Quoted 47 lines · Srivatsa Vaddagiri

Reset of a virtio-mmio device is initiated by writing 0 to its Status register. The reset operation itself may or may not be completed by the time write instruction completes. For devices that are emulated in software, writes to Status register are trapped and resumed only after the reset operation is complete. Thus a driver can be assured of reset completion as soon as its write completes. That may not be however true in other cases, such as when virtio-mmio devices implemented directly in hardware. Driver's write to Status register in such case could complete before the device reset operation is completed.

Update the specification to indicate when a driver may need to poll for reset completion before going ahead with the remaining initialization steps.

Suggested-by Jason Wang <[email protected]>

Signed-off-by Srivatsa Vaddagiri <[email protected]>

---
V2->V1:
- Use version 0x3 as an indication for drivers to poll for reset completion (rather than base it on a new feature bit)

Previous version can be found at:

https://lists.oasis-open.org/archives/virtio-dev/202107/msg00168.html

  content.tex | 11 +++++++----
  1 file changed, 7 insertions(+), 4 deletions(-)

content.tex +2 -2

index 7cec1c3..b6b30a0 100644
@@ -1730,9 +1730,9 @@ \subsection{MMIO Device Register Layout}\label{sec:Virtio Transport Options / Vi
    }
    \hline
    \mmioreg{Version}{Device version number}{0x004}{R}{%
-    0x2.
+    0x3.
      \begin{note}
-      Legacy devices (see \ref{sec:Virtio Transport Options / Virtio Over MMIO / Legacy interface}~\nameref{sec:Virtio Transport Options / Virtio Over MMIO / Legacy interface}) used 0x1.
+      Legacy devices (see \ref{sec:Virtio Transport Options / Virtio Over MMIO / Legacy interface}~\nameref{sec:Virtio Transport Options / Virtio Over MMIO / Legacy interface}) used 0x1. Devices that do not require drivers to poll for reset completion can use 0x2. See \ref{devicenormative:Virtio Transport Options / Virtio Over MMIO / MMIO Device Register Layout} for more details.
      \end{note}
    }
    \hline

@@ -1916,7 +1916,7 @@ \subsection{MMIO Device Register Layout}\label{sec:Virtio Transport Options / Vi
 
  The device MUST return 0x74726976 in \field{MagicValue}.
 

Quoted 7 lines

content.tex \subsection{MMIO Device Register Layout}\label{sec:Virtio Transport Options / Vi · line 1919

Quoted 2 lines

content.tex +1 -1

-The device MUST return value 0x2 in \field{Version}.
+The device MUST return either value 0x3 or 0x2 in \field{Version} based on its reset behavior. Drivers trigger reset of a device by writing 0 to \field{Status}. The reset operation itself may or may not be completed by the time write operation is complete. Devices whose reset operation completes synchronously with the write operation are allowed to return value of 0x2 for \field{Version}. Other devices, whose reset operation can be incomplete by the time write operation completes MUST return value 0x3 as an indication for drivers to poll for reset completion.

Maybe I'm overthining this, but might that suddenly render some existing device non-compliant? IIUC, we never mandated reset to be completed when the write was completed, only kind of implied it. An existing device will have version 2, but it might not be quite finished with reset after the status write yet.

Won't this break the driver probe/registering?

int register_virtio_device(struct virtio_device *dev)

{

...

        /* We always start by resetting the device, in case a previous
         * driver messed it up.  This also tests that code path a
little. */
        dev->config->reset(dev);

        /* Acknowledge that we've seen the device. */ virtio_add_status(dev, VIRTIO_CONFIG_S_ACKNOWLEDGE);

}

Thanks

Quoted 1 line

Quoted 59 lines · Jason Wang

Quoted 56 lines · Cornelia Huck

Quoted 44 lines · Srivatsa Vaddagiri

Reset of a virtio-mmio device is initiated by writing 0 to its Status register. The reset operation itself may or may not be completed by the time write instruction completes. For devices that are emulated in software, writes to Status register are trapped and resumed only after the reset operation is complete. Thus a driver can be assured of reset completion as soon as its write completes. That may not be however true in other cases, such as when virtio-mmio devices implemented directly in hardware. Driver's write to Status register in such case could complete before the device reset operation is completed.

Update the specification to indicate when a driver may need to poll for reset completion before going ahead with the remaining initialization steps.

Suggested-by Jason Wang <[email protected]>

Signed-off-by Srivatsa Vaddagiri <[email protected]>

---
V2->V1:
- Use version 0x3 as an indication for drivers to poll for reset completion (rather than base it on a new feature bit)

Previous version can be found at:

https://lists.oasis-open.org/archives/virtio-dev/202107/msg00168.html

  content.tex | 11 +++++++----
  1 file changed, 7 insertions(+), 4 deletions(-)

content.tex +2 -2

index 7cec1c3..b6b30a0 100644
@@ -1730,9 +1730,9 @@ \subsection{MMIO Device Register Layout}\label{sec:Virtio Transport Options / Vi
    }
    \hline
    \mmioreg{Version}{Device version number}{0x004}{R}{%
-    0x2.
+    0x3.
      \begin{note}
-      Legacy devices (see \ref{sec:Virtio Transport Options / Virtio Over MMIO / Legacy interface}~\nameref{sec:Virtio Transport Options / Virtio Over MMIO / Legacy interface}) used 0x1.
+      Legacy devices (see \ref{sec:Virtio Transport Options / Virtio Over MMIO / Legacy interface}~\nameref{sec:Virtio Transport Options / Virtio Over MMIO / Legacy interface}) used 0x1. Devices that do not require drivers to poll for reset completion can use 0x2. See \ref{devicenormative:Virtio Transport Options / Virtio Over MMIO / MMIO Device Register Layout} for more details.
      \end{note}
    }
    \hline

content.tex \subsection{MMIO Device Register Layout}\label{sec:Virtio Transport Options / Vi · line 1919

Quoted 5 lines

content.tex +1 -1

@@ -1916,7 +1916,7 @@ \subsection{MMIO Device Register Layout}\label{sec:Virtio Transport Options / Vi
 
  The device MUST return 0x74726976 in \field{MagicValue}.
 
-The device MUST return value 0x2 in \field{Version}.
+The device MUST return either value 0x3 or 0x2 in \field{Version} based on its reset behavior. Drivers trigger reset of a device by writing 0 to \field{Status}. The reset operation itself may or may not be completed by the time write operation is complete. Devices whose reset operation completes synchronously with the write operation are allowed to return value of 0x2 for \field{Version}. Other devices, whose reset operation can be incomplete by the time write operation completes MUST return value 0x3 as an indication for drivers to poll for reset completion.

Maybe I'm overthining this, but might that suddenly render some existing device non-compliant? IIUC, we never mandated reset to be completed when the write was completed, only kind of implied it. An existing device will have version 2, but it might not be quite finished with reset after the status write yet.

Quoted 1 line

Won't this break the driver probe/registering?

It might be racy, I guess; I'm not sure how other device or driver implementations work.

Quoted 22 lines

int register_virtio_device(struct virtio_device *dev)

{

...

/* We always start by resetting the device, in case a previous
* driver messed it up. This also tests that code path a
little. */

         dev->config->reset(dev);

         /* Acknowledge that we've seen the device. */
         virtio_add_status(dev, VIRTIO_CONFIG_S_ACKNOWLEDGE);

}

Thanks

Quoted 1 line

Quoted 62 lines · Cornelia Huck

Quoted 59 lines · Jason Wang

Quoted 56 lines · Cornelia Huck

Quoted 44 lines · Srivatsa Vaddagiri

Reset of a virtio-mmio device is initiated by writing 0 to its Status register. The reset operation itself may or may not be completed by the time write instruction completes. For devices that are emulated in software, writes to Status register are trapped and resumed only after the reset operation is complete. Thus a driver can be assured of reset completion as soon as its write completes. That may not be however true in other cases, such as when virtio-mmio devices implemented directly in hardware. Driver's write to Status register in such case could complete before the device reset operation is completed.

Update the specification to indicate when a driver may need to poll for reset completion before going ahead with the remaining initialization steps.

Suggested-by Jason Wang <[email protected]>

Signed-off-by Srivatsa Vaddagiri <[email protected]>

---
V2->V1:
- Use version 0x3 as an indication for drivers to poll for reset completion (rather than base it on a new feature bit)

Previous version can be found at:

https://lists.oasis-open.org/archives/virtio-dev/202107/msg00168.html

  content.tex | 11 +++++++----
  1 file changed, 7 insertions(+), 4 deletions(-)

content.tex +2 -2

index 7cec1c3..b6b30a0 100644
@@ -1730,9 +1730,9 @@ \subsection{MMIO Device Register Layout}\label{sec:Virtio Transport Options / Vi
    }
    \hline
    \mmioreg{Version}{Device version number}{0x004}{R}{%
-    0x2.
+    0x3.
      \begin{note}
-      Legacy devices (see \ref{sec:Virtio Transport Options / Virtio Over MMIO / Legacy interface}~\nameref{sec:Virtio Transport Options / Virtio Over MMIO / Legacy interface}) used 0x1.
+      Legacy devices (see \ref{sec:Virtio Transport Options / Virtio Over MMIO / Legacy interface}~\nameref{sec:Virtio Transport Options / Virtio Over MMIO / Legacy interface}) used 0x1. Devices that do not require drivers to poll for reset completion can use 0x2. See \ref{devicenormative:Virtio Transport Options / Virtio Over MMIO / MMIO Device Register Layout} for more details.
      \end{note}
    }
    \hline

content.tex \subsection{MMIO Device Register Layout}\label{sec:Virtio Transport Options / Vi · line 1919

Quoted 5 lines

content.tex +1 -1

@@ -1916,7 +1916,7 @@ \subsection{MMIO Device Register Layout}\label{sec:Virtio Transport Options / Vi
 
  The device MUST return 0x74726976 in \field{MagicValue}.
 
-The device MUST return value 0x2 in \field{Version}.
+The device MUST return either value 0x3 or 0x2 in \field{Version} based on its reset behavior. Drivers trigger reset of a device by writing 0 to \field{Status}. The reset operation itself may or may not be completed by the time write operation is complete. Devices whose reset operation completes synchronously with the write operation are allowed to return value of 0x2 for \field{Version}. Other devices, whose reset operation can be incomplete by the time write operation completes MUST return value 0x3 as an indication for drivers to poll for reset completion.

Maybe I'm overthining this, but might that suddenly render some existing device non-compliant? IIUC, we never mandated reset to be completed when the write was completed, only kind of implied it. An existing device will have version 2, but it might not be quite finished with reset after the status write yet.

Quoted 4 lines

Quoted 1 line

Won't this break the driver probe/registering?

It might be racy, I guess; I'm not sure how other device or driver implementations work.

PCI forces a re-read like:

        /* 0 status means a reset. */
        vp_modern_set_status(mdev, 0);
        /* After writing 0 to device_status, the driver MUST wait for a read of
         * device_status to return 0 before reinitializing the device.
         * This will flush out the status write, and flush in device writes,
         * including MSI-X interrupts, if any.
         */
        while (vp_modern_get_status(mdev))
                msleep(1);
        /* Flush pending VQ/configuration callbacks. */
        vp_synchronize_vectors(vdev);

So it's fine.

Thanks

Quoted 23 lines

Quoted 22 lines

int register_virtio_device(struct virtio_device *dev)

{

...

/* We always start by resetting the device, in case a previous
* driver messed it up. This also tests that code path a
little. */

         dev->config->reset(dev);

         /* Acknowledge that we've seen the device. */
         virtio_add_status(dev, VIRTIO_CONFIG_S_ACKNOWLEDGE);

}

Thanks

Quoted 1 line