Re: [PATCH v4 6/7] remoteproc: qcom: pas: Add late attach support for subsystems

[Stephan Gerhold <stephan.gerhold@linaro.org>]

On Tue, Mar 10, 2026 at 03:03:22AM -0700, Jingyi Wang wrote:
> From: Gokul Krishna Krishnakumar <gokul.krishnakumar@oss.qualcomm.com>
> 
> Subsystems can be brought out of reset by entities such as bootloaders.
> As the irq enablement could be later than subsystem bring up, the state
> of subsystem should be checked by reading SMP2P bits and performing ping
> test.
> 
> A new qcom_pas_attach() function is introduced. if a crash state is
> detected for the subsystem, rproc_report_crash() is called. If the
> subsystem is ready either at the first check or within a 5-second timeout
> and the ping is successful, it will be marked as "attached". The ready
> state could be set by either ready interrupt or handover interrupt.
> 
> If "early_boot" is set by kernel but "subsys_booted" is not completed
> within the timeout, It could be the early boot feature is not supported
> by other entities. In this case, the state will be marked as RPROC_OFFLINE
> so that the PAS driver can load the firmware and start the remoteproc. As
> the running state is set once attach function is called, the watchdog or
> fatal interrupt received can be handled correctly.
> 
> Signed-off-by: Gokul Krishna Krishnakumar <gokul.krishnakumar@oss.qualcomm.com>
> Co-developed-by: Jingyi Wang <jingyi.wang@oss.qualcomm.com>
> Signed-off-by: Jingyi Wang <jingyi.wang@oss.qualcomm.com>
> ---
>  drivers/remoteproc/qcom_q6v5.c      |  88 +++++++++++++++++++++++++++++-
>  drivers/remoteproc/qcom_q6v5.h      |  17 +++++-
>  drivers/remoteproc/qcom_q6v5_adsp.c |   2 +-
>  drivers/remoteproc/qcom_q6v5_mss.c  |   2 +-
>  drivers/remoteproc/qcom_q6v5_pas.c  | 103 ++++++++++++++++++++++++++++++++++--
>  drivers/remoteproc/qcom_q6v5_wcss.c |   2 +-
>  6 files changed, 204 insertions(+), 10 deletions(-)
> 
> [...]
> diff --git a/drivers/remoteproc/qcom_q6v5_pas.c b/drivers/remoteproc/qcom_q6v5_pas.c
> index 46204da046fa..4700d111e058 100644
> --- a/drivers/remoteproc/qcom_q6v5_pas.c
> +++ b/drivers/remoteproc/qcom_q6v5_pas.c
> @@ -36,6 +36,8 @@
>  
>  #define MAX_ASSIGN_COUNT 3
>  
> +#define EARLY_ATTACH_TIMEOUT_MS 5000
> +
>  struct qcom_pas_data {
>  	int crash_reason_smem;
>  	const char *firmware_name;
> [...]
> @@ -510,6 +521,80 @@ static unsigned long qcom_pas_panic(struct rproc *rproc)
>  	return qcom_q6v5_panic(&pas->q6v5);
>  }
>  
> +static int qcom_pas_attach(struct rproc *rproc)
> +{
> +	int ret;
> +	struct qcom_pas *pas = rproc->priv;
> +	bool ready_state;
> +	bool crash_state;
> +
> +	pas->q6v5.running = true;
> +	ret = irq_get_irqchip_state(pas->q6v5.fatal_irq,
> +				    IRQCHIP_STATE_LINE_LEVEL, &crash_state);
> +
> +	if (!ret && crash_state) {
> +		dev_err(pas->dev, "Sub system has crashed before driver probe\n");
> +		rproc_report_crash(rproc, RPROC_FATAL_ERROR);
> +		ret = -EINVAL;
> +		goto disable_running;
> +	}
> +
> +	if (!ret)
> +		ret = irq_get_irqchip_state(pas->q6v5.ready_irq,
> +					    IRQCHIP_STATE_LINE_LEVEL, &ready_state);
> +
> +	/*
> +	 * smp2p allocate irq entry can be delayed, irq_get_irqchip_state will get -ENODEV,
> +	 * the 5 seconds timeout is set to wait for this, after the entry is allocated, smp2p
> +	 * will call the qcom_smp2p_intr and complete the timeout in the ISR.
> +	 */
> +	if (unlikely(ret == -ENODEV) || unlikely(!ready_state)) {
> +		ret = wait_for_completion_timeout(&pas->q6v5.subsys_booted,
> +						  msecs_to_jiffies(EARLY_ATTACH_TIMEOUT_MS));

I have asked this back in October for v2 [1] and again in December for
v3 [2], but you still haven't really answered it. Please answer all
of the following questions:

 1. What is the use case for this timeout?
 2. In which situations will the start of the remoteproc be delayed?
 3. Why does the boot firmware not wait until the remoteproc is fully
    started before it continues booting?
 4. If the boot firmware gives up control before the remoteproc is fully
    started, how do you ensure that the handover resources are
    maintained until the remoteproc signals handover?

v4 looks a bit less dangerous now since you don't enable the handover
IRQ anymore. Still, I don't understand how this would work in practice.
Removing this timeout would be preferable because then we could actually
support firmware versions that do not automatically start the remoteproc
without having to delay the boot process for 5s.

Thanks,
Stephan

[1]: https://lore.kernel.org/r/aQHmanEiWmEac7aV@linaro.org/
[2]: https://lore.kernel.org/r/aUsUhX8Km275qonq@linaro.org/