Re: [PATCH v5 2/2] remoteproc: xlnx: add crash detection mechanism

["Shah, Tanmay" <tanmays@amd.com>]

On 6/22/2026 12:46 PM, Mathieu Poirier wrote:
> On Tue, Jun 16, 2026 at 02:52:50PM -0500, Shah, Tanmay wrote:
>> Hello,
>>
>> Please find my comments below:
>>
>> On 6/16/2026 11:01 AM, Mathieu Poirier wrote:
>>> On Wed, Jun 10, 2026 at 11:27:11AM -0700, Tanmay Shah wrote:
>>>> Remote processor will report the crash reason via the resource table
>>>> and notify the host via mailbox notification. The host checks this
>>>> crash reason on every mailbox notification from the remote and report
>>>> to the rproc core framework. Then the rproc core framework will start
>>>> the recovery process.
>>>>
>>>> Signed-off-by: Tanmay Shah <tanmay.shah@amd.com>
>>>> ---
>>>>
>>>> changes in v5
>>>>   - use local variable to access crash report pointer
>>>>   - End crash report string with '\0' without relying on fw
>>>>
>>>>  drivers/remoteproc/xlnx_r5_remoteproc.c | 73 ++++++++++++++++++++++++-
>>>>  1 file changed, 72 insertions(+), 1 deletion(-)
>>>>
>>>> diff --git a/drivers/remoteproc/xlnx_r5_remoteproc.c b/drivers/remoteproc/xlnx_r5_remoteproc.c
>>>> index 3349d1877751..86afff9f3b40 100644
>>>> --- a/drivers/remoteproc/xlnx_r5_remoteproc.c
>>>> +++ b/drivers/remoteproc/xlnx_r5_remoteproc.c
>>>> @@ -112,6 +112,10 @@ struct rsc_tbl_data {
>>>>  	const uintptr_t rsc_tbl;
>>>>  } __packed;
>>>>  
>>>> +enum xlnx_rproc_fw_rsc {
>>>> +	XLNX_RPROC_FW_CRASH_REASON = RSC_VENDOR_START,
>>>
>>> I would call this XLNX_RPROC_FW_CRASH_REPORT
>>>
>>
>> Ack.
>>
>>>> +};
>>>> +
>>>>  /*
>>>>   * Hardcoded TCM bank values. This will stay in driver to maintain backward
>>>>   * compatibility with device-tree that does not have TCM information.
>>>> @@ -131,9 +135,27 @@ static const struct mem_bank_data zynqmp_tcm_banks_lockstep[] = {
>>>>  	{0xffe30000UL, 0x30000, 0x10000UL, PD_R5_1_BTCM, "btcm1"},
>>>>  };
>>>>  
>>>> +#define CRASH_REASON_STR_LEN 16
>>>> +
>>>> +/**
>>>> + * struct xlnx_rproc_crash_report - resource to know crash status and reason
>>>> + *
>>>> + * @version: version of this resource
>>>> + * @crash_state: if true, the rproc is notifying crash, time to recover
>>>> + * @crash_reason: number to describe reason of crash
>>>> + * @crash_reason_str: short string description of crash reason
>>>> + */
>>>> +struct xlnx_rproc_crash_report {
>>>> +	u8 version;
>>>> +	u8 crash_state;
>>>> +	u8 crash_reason;
>>>
>>> Do you need 2 variables?  Would it be possible for @crash_reason != 0 to
>>> indicate a crash, making @cash_state irrelevant?
>>>
>>
>> Actually initially I had defined crash_reason only, but when I looked at
>> how crash reason/type is defined in the kernel, I got idea to keep
>> crash_state separate than crash_reason:
>>
>> https://git.kernel.org/pub/scm/linux/kernel/git/remoteproc/linux.git/tree/include/linux/remoteproc.h?h=for-next#n183
>>
>> enum rproc_crash_type {
>> 	RPROC_MMUFAULT,
>> 	RPROC_WATCHDOG,
>> 	RPROC_FATAL_ERROR,
>> };
>>
>> So, if crash_reason is defined like above, then it's easy to make
>> mistake and not define no_crash = 0. Different firmware projects can
>> treat crash_reason differently. I would like to keep crash_state and
>> crash_reason separate and not impose policy on firmware projects on how
>> to define crash_reason.
>>
> 
> I agree with your assessment.  Please move "crash_state" to "crashed".
>  

Ack.

>>>> +	char crash_reason_str[CRASH_REASON_STR_LEN];
>>>> +} __packed;
>>>> +
>>>>  /**
>>>>   * struct zynqmp_r5_core - remoteproc core's internal data
>>>>   *
>>>> + * @crash_report: rproc crash state and reason
>>>>   * @rsc_tbl_va: resource table virtual address
>>>>   * @sram: Array of sram memories assigned to this core
>>>>   * @num_sram: number of sram for this core
>>>> @@ -147,6 +169,7 @@ static const struct mem_bank_data zynqmp_tcm_banks_lockstep[] = {
>>>>   * @ipi: pointer to mailbox information
>>>>   */
>>>>  struct zynqmp_r5_core {
>>>> +	struct xlnx_rproc_crash_report *crash_report;
>>>>  	void __iomem *rsc_tbl_va;
>>>>  	struct zynqmp_sram_bank *sram;
>>>>  	int num_sram;
>>>> @@ -204,11 +227,27 @@ static int event_notified_idr_cb(int id, void *ptr, void *data)
>>>>   */
>>>>  static void handle_event_notified(struct work_struct *work)
>>>>  {
>>>> +	struct xlnx_rproc_crash_report *report;
>>>> +	struct zynqmp_r5_core *r5_core;
>>>>  	struct mbox_info *ipi;
>>>>  	struct rproc *rproc;
>>>>  
>>>>  	ipi = container_of(work, struct mbox_info, mbox_work);
>>>>  	rproc = ipi->r5_core->rproc;
>>>> +	r5_core = ipi->r5_core;
>>>> +	report = r5_core->crash_report;
>>>> +
>>>> +	/* report crash only if expected */
>>>> +	if (report && report->crash_state) {
>>>> +		if (rproc->state == RPROC_ATTACHED || rproc->state == RPROC_RUNNING) {
>>>> +			report->crash_reason_str[CRASH_REASON_STR_LEN - 1] = '\0';
>>>> +			dev_warn(&rproc->dev, "crash reason id: %d %s\n",
>>>> +				 report->crash_reason, report->crash_reason_str);
>>>> +			rproc_report_crash(rproc, RPROC_FATAL_ERROR);
>>>> +			report->crash_state = false;
>>>> +			return;
>>>> +		}
>>>> +	}
>>>>  
>>>>  	/*
>>>>  	 * We only use IPI for interrupt. The RPU firmware side may or may
>>>> @@ -448,6 +487,13 @@ static int zynqmp_r5_rproc_stop(struct rproc *rproc)
>>>>  	if (ret)
>>>>  		dev_err(r5_core->dev, "core force power down failed\n");
>>>>  
>>>> +	/*
>>>> +	 * Clear attach on recovery flag during stop operation. The next state
>>>> +	 * of the remote processor is expected to be "Running" state. In this
>>>> +	 * state boot recovery method must take place over attach on recovery.
>>>> +	 */
>>>> +	test_and_clear_bit(RPROC_FEAT_ATTACH_ON_RECOVERY, rproc->features);
>>>
>>> Why is there a need to play set/clear the RPROC_FEAT_ATTACH_ON_RECOVERY flag
>>> here and in zynqmp_r5_attach() and zynqmp_r5_detach()?  I think we talked about
>>> that before but can't remember the outcome of that conversation.
>>>
>>
>> The remote processor can go through following life cycle:
>>
>> attach() -> stop() -> start(). When this happens, ATTACH_ON_RECOVERY is
>> not valid anymore. At this point, it should be BOOT_RECOVERY which is
>> indicated by clearing ATTACH_RECOVERY flag. That is why it is important
>> to clear this bit on stop().
> 
> I'm good with that part...
> 
>>
>> Now in detach() it is not needed. I am just clearing the flag as part of
>> a cleanup sequence.
>>
>>  >> +
>>>>  	return ret;
>>>>  }
>>>>  
>>>> @@ -869,6 +915,9 @@ static int zynqmp_r5_get_rsc_table_va(struct zynqmp_r5_core *r5_core)
>>>>  
>>>>  static int zynqmp_r5_attach(struct rproc *rproc)
>>>>  {
>>>> +	/* Enable attach on recovery method. Clear it during rproc stop. */
>>>> +	rproc_set_feature(rproc, RPROC_FEAT_ATTACH_ON_RECOVERY);
>>>> +
>>>>  	dev_dbg(&rproc->dev, "rproc %d attached\n", rproc->index);
>>>>  
>>>>  	return 0;
>>>> @@ -883,9 +932,30 @@ static int zynqmp_r5_detach(struct rproc *rproc)
>>>>  	 */
>>>>  	zynqmp_r5_rproc_kick(rproc, 0);
>>>>  
>>>> +	clear_bit(RPROC_FEAT_ATTACH_ON_RECOVERY, rproc->features);
>>>> +
> 
> ... but why is this needed?  Why can't it just be set once in
> zynqmp_r5_add_rproc_core() ?  The only time the bit should be cleared is in
> zynqmp_r5_rproc_stop().
> 

Yes we don't need in detach(). I was just clearing it as a reverse
operation of what's done in attach(). You are right if we move the _set
part during probe, then we don't need to clean it in the detach(). I
will send next revision with this refactor.

Thanks,
Tanmay

> 
>>>>  	return 0;
>>>>  }
>>>>  
>>>> +static int zynqmp_r5_handle_rsc(struct rproc *rproc, u32 rsc_type, void *rsc,
>>>> +				int offset, int avail)
>>>> +{
>>>> +	struct zynqmp_r5_core *r5_core = rproc->priv;
>>>> +	void *rsc_offset = (r5_core->rsc_tbl_va + offset);
>>>> +
>>>
>>>         if (rsc_type != XLNX_RPROC_FW_CRASH_REASON)
>>>                 return RSC_IGNORED;
>>>
>>
>> Ack.
>>
>>>> +	if (rsc_type == XLNX_RPROC_FW_CRASH_REASON) {
>>>> +		r5_core->crash_report = rsc_offset;
>>>> +		/* reset all values */
>>>> +		r5_core->crash_report->crash_state = false;
>>>> +		r5_core->crash_report->crash_reason = 0;
>>>> +		r5_core->crash_report->crash_reason_str[0] = '\0';
>>>> +	} else {
>>>> +		return RSC_IGNORED;
>>>> +	}
>>>> +
>>>> +	return RSC_HANDLED;
>>>> +}
>>>> +
>>>>  static const struct rproc_ops zynqmp_r5_rproc_ops = {
>>>>  	.prepare	= zynqmp_r5_rproc_prepare,
>>>>  	.unprepare	= zynqmp_r5_rproc_unprepare,
>>>> @@ -900,6 +970,7 @@ static const struct rproc_ops zynqmp_r5_rproc_ops = {
>>>>  	.get_loaded_rsc_table = zynqmp_r5_get_loaded_rsc_table,
>>>>  	.attach		= zynqmp_r5_attach,
>>>>  	.detach		= zynqmp_r5_detach,
>>>> +	.handle_rsc	= zynqmp_r5_handle_rsc,
>>>>  };
>>>>  
>>>>  /**
>>>> @@ -939,7 +1010,7 @@ static struct zynqmp_r5_core *zynqmp_r5_alloc_rproc_core(struct device *cdev)
>>>>  
>>>>  	rproc_coredump_set_elf_info(r5_rproc, ELFCLASS32, EM_ARM);
>>>>  
>>>> -	r5_rproc->recovery_disabled = true;
>>>> +	r5_rproc->recovery_disabled = false;
>>>
>>> I'm good with the overall architecture of this set.
>>>
>>
>> Okay. If above comments looks good, I will send v6 soon.
>>
>>> Regards,
>>> Mathieu
>>>
>>>>  	r5_rproc->has_iommu = false;
>>>>  	r5_rproc->auto_boot = false;
>>>>  
>>>> -- 
>>>> 2.34.1
>>>>
>>