[PATCH] mailbox: pcc: Notify clients on polled completion

[PATCH] mailbox: pcc: Notify clients on polled completion

[Sudeep Holla]

PCC channels without a platform interrupt rely on the mailbox
polling path to detect command completion.

That path currently only reports transmit completion to the mailbox
core, so clients that wait for their receive callback do not get
notified when the command completes.

Call mbox_chan_received_data() when polling observes completion on a
channel without a platform IRQ, matching the interrupt-driven
completion path.

Reported-by: Cristian Marussi <cristian.marussi@arm.com>
Signed-off-by: Sudeep Holla <sudeep.holla@kernel.org>
---
 drivers/mailbox/pcc.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/drivers/mailbox/pcc.c b/drivers/mailbox/pcc.c
index 636879ae1db7..7c9451ab4527 100644
--- a/drivers/mailbox/pcc.c
+++ b/drivers/mailbox/pcc.c
@@ -447,9 +447,14 @@ static int pcc_send_data(struct mbox_chan *chan, void *data)
 
 static bool pcc_last_tx_done(struct mbox_chan *chan)
 {
+	bool ret;
 	struct pcc_chan_info *pchan = chan->con_priv;
 
-	return pcc_mbox_cmd_complete_check(pchan);
+	ret = pcc_mbox_cmd_complete_check(pchan);
+	if (ret && !pchan->plat_irq)
+		mbox_chan_received_data(chan, NULL);
+
+	return ret;
 }
 
 /**
-- 
2.43.0

Re: [PATCH] mailbox: pcc: Notify clients on polled completion

[lihuisong (C)]

Hi Sudeep,

On 6/13/2026 1:02 AM, Sudeep Holla wrote:
> PCC channels without a platform interrupt rely on the mailbox
> polling path to detect command completion.
>
> That path currently only reports transmit completion to the mailbox
> core, so clients that wait for their receive callback do not get
> notified when the command completes.
>
> Call mbox_chan_received_data() when polling observes completion on a
> channel without a platform IRQ, matching the interrupt-driven
> completion path.
>
> Reported-by: Cristian Marussi <cristian.marussi@arm.com>
> Signed-off-by: Sudeep Holla <sudeep.holla@kernel.org>
> ---
>   drivers/mailbox/pcc.c | 7 ++++++-
>   1 file changed, 6 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/mailbox/pcc.c b/drivers/mailbox/pcc.c
> index 636879ae1db7..7c9451ab4527 100644
> --- a/drivers/mailbox/pcc.c
> +++ b/drivers/mailbox/pcc.c
> @@ -447,9 +447,14 @@ static int pcc_send_data(struct mbox_chan *chan, void *data)
>   
>   static bool pcc_last_tx_done(struct mbox_chan *chan)
>   {
> +	bool ret;
>   	struct pcc_chan_info *pchan = chan->con_priv;
>   
> -	return pcc_mbox_cmd_complete_check(pchan);
> +	ret = pcc_mbox_cmd_complete_check(pchan);
> +	if (ret && !pchan->plat_irq)
> +		mbox_chan_received_data(chan, NULL);
> +
> +	return ret;
>   }
The mailbox_controller.h said that .last_tx_done() is used only if 
txdone_poll:=true && txdone_irq:=false.
How about add a verification at the begining of this function? like "if 
(chan->txdone_method != MBOX_TXDONE_BY_POLL) return false;"
And then call mbox_chan_received_data() directly if command completed.

This patch is ok to me if has above changes.
Acked-by: lihuisong@huawei.com


But pcc mbox controller can no longer work in MBOX_TXDONE_BY_ACK after 
your commit:
3349f800609e (mailbox: pcc: Set txdone_irq/txdone_poll based on PCCT flags).
This may lead to above client drivers fail to send mailbox command, like 
cppc_acpi,hisi_uncore_freq and kunpeng_hccs.
Because these client driver knows tx_done(knows_txdone: true), work on 
this mode and need to call mbox_client_txdone.
Do you have some idea for this?
>   
>   /**

Re: [PATCH] mailbox: pcc: Notify clients on polled completion

[Sudeep Holla]

On Thu, Jun 18, 2026 at 10:04:12AM +0800, lihuisong (C) wrote:
> Hi Sudeep,
> 
> On 6/13/2026 1:02 AM, Sudeep Holla wrote:
> > PCC channels without a platform interrupt rely on the mailbox
> > polling path to detect command completion.
> > 
> > That path currently only reports transmit completion to the mailbox
> > core, so clients that wait for their receive callback do not get
> > notified when the command completes.
> > 
> > Call mbox_chan_received_data() when polling observes completion on a
> > channel without a platform IRQ, matching the interrupt-driven
> > completion path.
> > 
> > Reported-by: Cristian Marussi <cristian.marussi@arm.com>
> > Signed-off-by: Sudeep Holla <sudeep.holla@kernel.org>
> > ---
> >   drivers/mailbox/pcc.c | 7 ++++++-
> >   1 file changed, 6 insertions(+), 1 deletion(-)
> > 
> > diff --git a/drivers/mailbox/pcc.c b/drivers/mailbox/pcc.c
> > index 636879ae1db7..7c9451ab4527 100644
> > --- a/drivers/mailbox/pcc.c
> > +++ b/drivers/mailbox/pcc.c
> > @@ -447,9 +447,14 @@ static int pcc_send_data(struct mbox_chan *chan, void *data)
> >   static bool pcc_last_tx_done(struct mbox_chan *chan)
> >   {
> > +	bool ret;
> >   	struct pcc_chan_info *pchan = chan->con_priv;
> > -	return pcc_mbox_cmd_complete_check(pchan);
> > +	ret = pcc_mbox_cmd_complete_check(pchan);
> > +	if (ret && !pchan->plat_irq)
> > +		mbox_chan_received_data(chan, NULL);
> > +
> > +	return ret;
> >   }
> The mailbox_controller.h said that .last_tx_done() is used only if
> txdone_poll:=true && txdone_irq:=false.
> How about add a verification at the begining of this function? like "if
> (chan->txdone_method != MBOX_TXDONE_BY_POLL) return false;"
> And then call mbox_chan_received_data() directly if command completed.
> 

Thanks, it does makes sense to me. I will have look.

> This patch is ok to me if has above changes.
> Acked-by: lihuisong@huawei.com
> 
> 
> But pcc mbox controller can no longer work in MBOX_TXDONE_BY_ACK after your
> commit:
> 3349f800609e (mailbox: pcc: Set txdone_irq/txdone_poll based on PCCT flags).
> This may lead to above client drivers fail to send mailbox command, like
> cppc_acpi,hisi_uncore_freq and kunpeng_hccs.
> Because these client driver knows tx_done(knows_txdone: true), work on this
> mode and need to call mbox_client_txdone.
> Do you have some idea for this?

I wonder if we can ask the platform to avoid generating interrupts in those
case by setting/clearing "Notify on completion" in shmem. Do you think that
would work ?

-- 
Regards,
Sudeep

Re: [PATCH] mailbox: pcc: Notify clients on polled completion

[lihuisong (C)]

On 6/18/2026 8:06 PM, Sudeep Holla wrote:
> On Thu, Jun 18, 2026 at 10:04:12AM +0800, lihuisong (C) wrote:
>> Hi Sudeep,
>>
>> On 6/13/2026 1:02 AM, Sudeep Holla wrote:
>>> PCC channels without a platform interrupt rely on the mailbox
>>> polling path to detect command completion.
>>>
>>> That path currently only reports transmit completion to the mailbox
>>> core, so clients that wait for their receive callback do not get
>>> notified when the command completes.
>>>
>>> Call mbox_chan_received_data() when polling observes completion on a
>>> channel without a platform IRQ, matching the interrupt-driven
>>> completion path.
>>>
>>> Reported-by: Cristian Marussi <cristian.marussi@arm.com>
>>> Signed-off-by: Sudeep Holla <sudeep.holla@kernel.org>
>>> ---
>>>    drivers/mailbox/pcc.c | 7 ++++++-
>>>    1 file changed, 6 insertions(+), 1 deletion(-)
>>>
>>> diff --git a/drivers/mailbox/pcc.c b/drivers/mailbox/pcc.c
>>> index 636879ae1db7..7c9451ab4527 100644
>>> --- a/drivers/mailbox/pcc.c
>>> +++ b/drivers/mailbox/pcc.c
>>> @@ -447,9 +447,14 @@ static int pcc_send_data(struct mbox_chan *chan, void *data)
>>>    static bool pcc_last_tx_done(struct mbox_chan *chan)
>>>    {
>>> +	bool ret;
>>>    	struct pcc_chan_info *pchan = chan->con_priv;
>>> -	return pcc_mbox_cmd_complete_check(pchan);
>>> +	ret = pcc_mbox_cmd_complete_check(pchan);
>>> +	if (ret && !pchan->plat_irq)
>>> +		mbox_chan_received_data(chan, NULL);
>>> +
>>> +	return ret;
>>>    }
>> The mailbox_controller.h said that .last_tx_done() is used only if
>> txdone_poll:=true && txdone_irq:=false.
>> How about add a verification at the begining of this function? like "if
>> (chan->txdone_method != MBOX_TXDONE_BY_POLL) return false;"
>> And then call mbox_chan_received_data() directly if command completed.
>>
> Thanks, it does makes sense to me. I will have look.
>
>> This patch is ok to me if has above changes.
>> Acked-by: lihuisong@huawei.com
>>
>>
>> But pcc mbox controller can no longer work in MBOX_TXDONE_BY_ACK after your
>> commit:
>> 3349f800609e (mailbox: pcc: Set txdone_irq/txdone_poll based on PCCT flags).
>> This may lead to above client drivers fail to send mailbox command, like
>> cppc_acpi,hisi_uncore_freq and kunpeng_hccs.
>> Because these client driver knows tx_done(knows_txdone: true), work on this
>> mode and need to call mbox_client_txdone.
>> Do you have some idea for this?
> I wonder if we can ask the platform to avoid generating interrupts in those
> case by setting/clearing "Notify on completion" in shmem. Do you think that
> would work ?
These drivers didn't work on interrupt mode.
They poll the complete status of command.
>

Re: [PATCH] mailbox: pcc: Notify clients on polled completion

[lihuisong (C)]

On 6/18/2026 8:06 PM, Sudeep Holla wrote:
> On Thu, Jun 18, 2026 at 10:04:12AM +0800, lihuisong (C) wrote:
>> Hi Sudeep,
>>
>> On 6/13/2026 1:02 AM, Sudeep Holla wrote:
>>> PCC channels without a platform interrupt rely on the mailbox
>>> polling path to detect command completion.
>>>
>>> That path currently only reports transmit completion to the mailbox
>>> core, so clients that wait for their receive callback do not get
>>> notified when the command completes.
>>>
>>> Call mbox_chan_received_data() when polling observes completion on a
>>> channel without a platform IRQ, matching the interrupt-driven
>>> completion path.
>>>
>>> Reported-by: Cristian Marussi <cristian.marussi@arm.com>
>>> Signed-off-by: Sudeep Holla <sudeep.holla@kernel.org>
>>> ---
>>>    drivers/mailbox/pcc.c | 7 ++++++-
>>>    1 file changed, 6 insertions(+), 1 deletion(-)
>>>
>>> diff --git a/drivers/mailbox/pcc.c b/drivers/mailbox/pcc.c
>>> index 636879ae1db7..7c9451ab4527 100644
>>> --- a/drivers/mailbox/pcc.c
>>> +++ b/drivers/mailbox/pcc.c
>>> @@ -447,9 +447,14 @@ static int pcc_send_data(struct mbox_chan *chan, void *data)
>>>    static bool pcc_last_tx_done(struct mbox_chan *chan)
>>>    {
>>> +	bool ret;
>>>    	struct pcc_chan_info *pchan = chan->con_priv;
>>> -	return pcc_mbox_cmd_complete_check(pchan);
>>> +	ret = pcc_mbox_cmd_complete_check(pchan);
>>> +	if (ret && !pchan->plat_irq)
>>> +		mbox_chan_received_data(chan, NULL);
>>> +
>>> +	return ret;
>>>    }
>> The mailbox_controller.h said that .last_tx_done() is used only if
>> txdone_poll:=true && txdone_irq:=false.
>> How about add a verification at the begining of this function? like "if
>> (chan->txdone_method != MBOX_TXDONE_BY_POLL) return false;"
>> And then call mbox_chan_received_data() directly if command completed.
>>
> Thanks, it does makes sense to me. I will have look.
>
>> This patch is ok to me if has above changes.
>> Acked-by: lihuisong@huawei.com
>>
>>
>> But pcc mbox controller can no longer work in MBOX_TXDONE_BY_ACK after your
>> commit:
>> 3349f800609e (mailbox: pcc: Set txdone_irq/txdone_poll based on PCCT flags).
>> This may lead to above client drivers fail to send mailbox command, like
>> cppc_acpi,hisi_uncore_freq and kunpeng_hccs.
>> Because these client driver knows tx_done(knows_txdone: true), work on this
>> mode and need to call mbox_client_txdone.
>> Do you have some idea for this?
> I wonder if we can ask the platform to avoid generating interrupts in those
> case by setting/clearing "Notify on completion" in shmem. Do you think that
> would work ?
Hi Sudeep，

Review this logic again.
The mbox_bind_client is called on requesting PCC channel.
And mbox_bind_client could modify chan->txdone_method to 
MBOX_TXDONE_BY_ACK if client knows txdone and 'txdone_method' is 
MBOX_TXDONE_BY_POLL.
So I think your above commit is ok for these client drivers.

/Huisong