[PATCH] content: Reserve ID 52 for Snapdragon device

[PATCH] content: Reserve ID 52 for Snapdragon device

[Peter Hilber]

Reserve a device ID for the Snapdragon device.  A Virtio Snapdragon
device exposes functionality across Snapdragon SoC subsystems through a
common Virtio-based protocol.

Device ID 51 is already reserved for the Virtio HAB device:

https://lore.kernel.org/virtio-comment/20260702140939.3535968-1-zeloren@qti.qualcomm.com/

Signed-off-by: Peter Hilber <peter.hilber@oss.qualcomm.com>
---
 content.tex | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/content.tex b/content.tex
index 243ce2d..ac61ec6 100644
--- a/content.tex
+++ b/content.tex
@@ -805,6 +805,8 @@ \chapter{Device Types}\label{sec:Device Types}
 \hline
 50         &   Video encoder/decoder device \\
 \hline
+52         &   Snapdragon device \\
+\hline
 \end{longtable}
 
 Some of the devices above are unspecified by this document,

base-commit: bb1dd2e1fe89b862f38f15873d835a698b196f89
-- 
2.43.0

Re: [PATCH] content: Reserve ID 52 for Snapdragon device

[Jan Kiszka]

On 26.08.26 13:55, Peter Hilber wrote:
> Reserve a device ID for the Snapdragon device.  A Virtio Snapdragon
> device exposes functionality across Snapdragon SoC subsystems through a
> common Virtio-based protocol.
> 

Virtio is supposed to provide generic hardware models, not so much a
channel to expose concrete, physical chip IP over it. This looks very
weird to me.

Also, isn't Snapdragon a protected brand by Qualcomm? Would it even
legally be fine use it in the spec?

> Device ID 51 is already reserved for the Virtio HAB device:
> 
> https://lore.kernel.org/virtio-comment/20260702140939.3535968-1-zeloren@qti.qualcomm.com/
> 

I think I even voted for this, but likely without checking back what
this may become. I was expecting (hoping?) for a generic, portable inter
VM channel using shared memory. I hope HAB can be that as well - besides
enabling some very Qualcomm-specific and unportable pass-through of
accelerators or so.

Does a "Snapdragon device" have any relation to / dependency on HAB, or
is this just the ID collision why it is mentioned here?

Jan

-- 
Siemens AG, Foundational Technologies
Linux Expert Center

Re: [PATCH] content: Reserve ID 52 for Snapdragon device

[Stefan Hajnoczi]

[-- Attachment #1: Type: text/plain, Size: 1811 bytes --]

On Wed, Aug 26, 2026 at 05:31:54PM +0200, Jan Kiszka wrote:
> On 26.08.26 13:55, Peter Hilber wrote:
> > Reserve a device ID for the Snapdragon device.  A Virtio Snapdragon
> > device exposes functionality across Snapdragon SoC subsystems through a
> > common Virtio-based protocol.
> > 
> 
> Virtio is supposed to provide generic hardware models, not so much a
> channel to expose concrete, physical chip IP over it. This looks very
> weird to me.

Device IDs assignments don't have to result in standardized devices. See
"5 Device Types" in VIRTIO 1.4 where there is a table of currently
reserved Device IDs.

While I'm in favor of defining generic device interfaces with many
possible implementations like virtio-net, virtio-blk, etc, I think a
SoC-specific control device is unlikely to fit into that category and
it's acceptable to allocate the Device ID with the understanding that
there will never be standardization of the device.

Stefan

> 
> Also, isn't Snapdragon a protected brand by Qualcomm? Would it even
> legally be fine use it in the spec?
> 
> > Device ID 51 is already reserved for the Virtio HAB device:
> > 
> > https://lore.kernel.org/virtio-comment/20260702140939.3535968-1-zeloren@qti.qualcomm.com/
> > 
> 
> I think I even voted for this, but likely without checking back what
> this may become. I was expecting (hoping?) for a generic, portable inter
> VM channel using shared memory. I hope HAB can be that as well - besides
> enabling some very Qualcomm-specific and unportable pass-through of
> accelerators or so.
> 
> Does a "Snapdragon device" have any relation to / dependency on HAB, or
> is this just the ID collision why it is mentioned here?
> 
> Jan
> 
> -- 
> Siemens AG, Foundational Technologies
> Linux Expert Center
> 

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: [PATCH] content: Reserve ID 52 for Snapdragon device

[Jan Kiszka]

On 26.08.26 20:44, Stefan Hajnoczi wrote:
> On Wed, Aug 26, 2026 at 05:31:54PM +0200, Jan Kiszka wrote:
>> On 26.08.26 13:55, Peter Hilber wrote:
>>> Reserve a device ID for the Snapdragon device.  A Virtio Snapdragon
>>> device exposes functionality across Snapdragon SoC subsystems through a
>>> common Virtio-based protocol.
>>>
>>
>> Virtio is supposed to provide generic hardware models, not so much a
>> channel to expose concrete, physical chip IP over it. This looks very
>> weird to me.
> 
> Device IDs assignments don't have to result in standardized devices. See
> "5 Device Types" in VIRTIO 1.4 where there is a table of currently
> reserved Device IDs.
> 
> While I'm in favor of defining generic device interfaces with many
> possible implementations like virtio-net, virtio-blk, etc, I think a
> SoC-specific control device is unlikely to fit into that category and
> it's acceptable to allocate the Device ID with the understanding that
> there will never be standardization of the device.

Would still list this ID by that vendor branding in the spec.

For me the question is if reservations for devices that will never be
acceptable should be in scope for virtio. IIRC, the currently available
ID range for virtio over PCI is 63, of which 50 (including invalid 0)
are officially reserved already. If we see an inflation of such
vendor-specific reservations, we might be force to hunt for another PCI
device ID range earlier.

Jan

-- 
Siemens AG, Foundational Technologies
Linux Expert Center

Re: [PATCH] content: Reserve ID 52 for Snapdragon device

[Stefan Hajnoczi]

[-- Attachment #1: Type: text/plain, Size: 1715 bytes --]

On Thu, Aug 27, 2026 at 03:24:18PM +0200, Jan Kiszka wrote:
> On 26.08.26 20:44, Stefan Hajnoczi wrote:
> > On Wed, Aug 26, 2026 at 05:31:54PM +0200, Jan Kiszka wrote:
> >> On 26.08.26 13:55, Peter Hilber wrote:
> >>> Reserve a device ID for the Snapdragon device.  A Virtio Snapdragon
> >>> device exposes functionality across Snapdragon SoC subsystems through a
> >>> common Virtio-based protocol.
> >>>
> >>
> >> Virtio is supposed to provide generic hardware models, not so much a
> >> channel to expose concrete, physical chip IP over it. This looks very
> >> weird to me.
> > 
> > Device IDs assignments don't have to result in standardized devices. See
> > "5 Device Types" in VIRTIO 1.4 where there is a table of currently
> > reserved Device IDs.
> > 
> > While I'm in favor of defining generic device interfaces with many
> > possible implementations like virtio-net, virtio-blk, etc, I think a
> > SoC-specific control device is unlikely to fit into that category and
> > it's acceptable to allocate the Device ID with the understanding that
> > there will never be standardization of the device.
> 
> Would still list this ID by that vendor branding in the spec.
> 
> For me the question is if reservations for devices that will never be
> acceptable should be in scope for virtio. IIRC, the currently available
> ID range for virtio over PCI is 63, of which 50 (including invalid 0)
> are officially reserved already. If we see an inflation of such
> vendor-specific reservations, we might be force to hunt for another PCI
> device ID range earlier.

Interesting, maybe Michael Tsirkin knows what options are available for
extending the PCI Device ID range.

Stefan

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: [PATCH] content: Reserve ID 52 for Snapdragon device

[Peter Hilber]

On Wed, Aug 26, 2026 at 05:31:54PM +0200, Jan Kiszka wrote:
> On 26.08.26 13:55, Peter Hilber wrote:
> > Reserve a device ID for the Snapdragon device.  A Virtio Snapdragon
> > device exposes functionality across Snapdragon SoC subsystems through a
> > common Virtio-based protocol.
> > 
> 
> Virtio is supposed to provide generic hardware models, not so much a
> channel to expose concrete, physical chip IP over it. This looks very
> weird to me.

The Snapdragon device is not planned to undergo standardization, so it
should be acceptable as argued by Stefan.  Qualcomm has some Linux
kernel Virtio drivers which are intended for upstreaming, and as such
need a reserved device ID.

> 
> Also, isn't Snapdragon a protected brand by Qualcomm? Would it even
> legally be fine use it in the spec?

This use of the Snapdragon name in the Virtio specification has been
approved on the Qualcomm side.

> 
> > Device ID 51 is already reserved for the Virtio HAB device:
> > 
> > https://lore.kernel.org/virtio-comment/20260702140939.3535968-1-zeloren@qti.qualcomm.com/
> > 
> 
> I think I even voted for this, but likely without checking back what
> this may become. I was expecting (hoping?) for a generic, portable inter
> VM channel using shared memory. I hope HAB can be that as well - besides
> enabling some very Qualcomm-specific and unportable pass-through of
> accelerators or so.
> 
> Does a "Snapdragon device" have any relation to / dependency on HAB, or
> is this just the ID collision why it is mentioned here?

I mentioned it only because the HAB device ID patch has not been merged
yet.  The Snapdragon device is not related to the HAB device, which aims
for standardization.

Thank you for your feedback, and thank you to Stefan as well! 

Peter

> 
> Jan
> 
> -- 
> Siemens AG, Foundational Technologies
> Linux Expert Center

Re: [PATCH] content: Reserve ID 52 for Snapdragon device

[Jan Kiszka]

On 27.08.26 18:17, Peter Hilber wrote:
> On Wed, Aug 26, 2026 at 05:31:54PM +0200, Jan Kiszka wrote:
>> On 26.08.26 13:55, Peter Hilber wrote:
>>> Reserve a device ID for the Snapdragon device.  A Virtio Snapdragon
>>> device exposes functionality across Snapdragon SoC subsystems through a
>>> common Virtio-based protocol.
>>>
>>
>> Virtio is supposed to provide generic hardware models, not so much a
>> channel to expose concrete, physical chip IP over it. This looks very
>> weird to me.
> 
> The Snapdragon device is not planned to undergo standardization, so it
> should be acceptable as argued by Stefan.  Qualcomm has some Linux
> kernel Virtio drivers which are intended for upstreaming, and as such
> need a reserved device ID.
> 

Thanks for confirming.

But I'm personally not convinced that this is a pattern the virtio
specification should support, even if we had plenty of free space. It is
one thing to reserve an ID for a conceptually hardware-agnostic device
type where there is at least the possibility to standardize this one
day. But doing that for a single-vendor hardware which is likely not to
survive very long (seen that way too often...), this just seems wrong.

Jan

-- 
Siemens AG, Foundational Technologies
Linux Expert Center