mbox series

[v5,0/2] Allow parameter in smc/hvc calls

Message ID 20230506182428.25343-1-quic_nkela@quicinc.com
Headers
Series Allow parameter in smc/hvc calls |

Message

Nikunj Kela May 6, 2023, 6:24 p.m. UTC 
  Currently, smc/hvc calls are made with parameters set
to zeros. We are using multiple scmi instances within
a VM. We are sharing the same smc-id(func_id) with all
scmi instance. The hypervisor needs a way to distinguish
among hvc calls made from different instances.

This patch series introduces new compatible string which
can be used to pass shmem channel address as parameters
to smc/hvc calls.

---
v5 -> avoid computing page and offset in send function
Link: https://lore.kernel.org/all/20230418185659.29745-1-quic_nkela@quicinc.com/

v4 -> split shmem address into 4KB-pages and offset

v3 -> pass shmem channel address as parameter

v2 -> fix the compilation erros on 32bit platform(see below)
Reported-by: kernel test robot <lkp@intel.com>
Closes: https://lore.kernel.org/oe-kbuild-all/202304100606.kUjhsRYf-lkp@intel.com/

v1 -> original patches

Nikunj Kela (2):
  dt-bindings: firmware: arm,scmi: support for parameter in smc/hvc call
  firmware: arm_scmi: Augment SMC/HVC to allow optional parameters

 .../bindings/firmware/arm,scmi.yaml           |  8 ++++-
 drivers/firmware/arm_scmi/driver.c            |  1 +
 drivers/firmware/arm_scmi/smc.c               | 30 ++++++++++++++++++-
 3 files changed, 37 insertions(+), 2 deletions(-)

Comments

Sudeep Holla May 9, 2023, 4:01 p.m. UTC | #1 
On Sat, May 06, 2023 at 11:24:26AM -0700, Nikunj Kela wrote:
> Currently, smc/hvc calls are made with parameters set
> to zeros. We are using multiple scmi instances within
> a VM. We are sharing the same smc-id(func_id) with all
> scmi instance. The hypervisor needs a way to distinguish
> among hvc calls made from different instances.
> 
> This patch series introduces new compatible string which
> can be used to pass shmem channel address as parameters
> to smc/hvc calls.
> 
> ---
> v5 -> avoid computing page and offset in send function
> Link: https://lore.kernel.org/all/20230418185659.29745-1-quic_nkela@quicinc.com/
>

Thanks for the updated patch. I will pick this up for v6.5 and update
once queued.
Sudeep Holla May 31, 2023, 12:26 p.m. UTC | #2 
On Sat, 06 May 2023 11:24:26 -0700, Nikunj Kela wrote:
> Currently, smc/hvc calls are made with parameters set
> to zeros. We are using multiple scmi instances within
> a VM. We are sharing the same smc-id(func_id) with all
> scmi instance. The hypervisor needs a way to distinguish
> among hvc calls made from different instances.
> 
> This patch series introduces new compatible string which
> can be used to pass shmem channel address as parameters
> to smc/hvc calls.
> 
> [...]

Applied to sudeep.holla/linux (for-next/scmi/updates), thanks!

[1/2] dt-bindings: firmware: arm,scmi: support for parameter in smc/hvc call
      https://git.kernel.org/sudeep.holla/c/8f9d530cffc1
[2/2] firmware: arm_scmi: Augment SMC/HVC to allow optional parameters
      https://git.kernel.org/sudeep.holla/c/5f2ea10a808a

--
Regards,
Sudeep
Nikunj Kela June 30, 2023, 3:14 p.m. UTC | #3 
On 6/30/2023 2:44 AM, Sudeep Holla wrote:
> On Wed, Jun 28, 2023 at 01:38:32PM -0700, Nikunj Kela wrote:
>> On 5/31/2023 5:26 AM, Sudeep Holla wrote:
>>> On Sat, 06 May 2023 11:24:26 -0700, Nikunj Kela wrote:
>>>> Currently, smc/hvc calls are made with parameters set
>>>> to zeros. We are using multiple scmi instances within
>>>> a VM. We are sharing the same smc-id(func_id) with all
>>>> scmi instance. The hypervisor needs a way to distinguish
>>>> among hvc calls made from different instances.
>>>>
>>>> This patch series introduces new compatible string which
>>>> can be used to pass shmem channel address as parameters
>>>> to smc/hvc calls.
>>>>
>>>> [...]
>>> Applied to sudeep.holla/linux (for-next/scmi/updates), thanks!
>> Hi Sudeep, our hypervisor team is evaluating other scmi transport
>> options(including new qcom specific transport) along with this one so there
>> is a possibility that we might not use this solution. If you think this
>> patch is not useful to others, you can hold off its merge. Sorry about the
>> last minute inconvenience.
> Firstly, not sure why you took this in private. I am going to reply on the
> list and from now on I will never trust or rush to take any Qualcomm changes.
> Nothing personal but I have to follow that to keep it simple. It is in
> Linux tree yesterday or today and yes it is too late.
I understand your frustration and trust me I am even more disappointed 
by the decision by our hypervisor team and that too this late.  My 
efforts on this series are all wasted. If you like me to post a revert 
of this patch series, I will be happy to do that. Please don't let this 
one example reflect whole Qualcomm culture. Again, extremely sorry about 
all the efforts you put in.
>