Re: [PATCH RESEND] PCI/DPC: Fix print AER status in DPC event handling

[Dongdong Liu <liudongdong3@huawei.com>]

Hi Keith

Many thanks for your review.

在 2019/2/11 23:46, Keith Busch 写道:
> On Mon, Feb 11, 2019 at 03:02:59PM +0800, Dongdong Liu wrote:
>> +static int dpc_get_aer_uncorrect_severity(struct pci_dev *dev,
>> +					  struct aer_err_info *info)
>> +{
>> +	int pos = dev->aer_cap;
>> +	u32 status, mask, sev;
>> +
>> +	pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_STATUS, &status);
>> +	pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_MASK, &mask);
>> +	status &= ~mask;
>> +	if (!status)
>> +		return 0;
>> +
>> +	pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_SEVER, &sev);
>> +	status &= sev;
>> +	if (status)
>> +		info->severity = AER_FATAL;
>> +	else
>> +		info->severity = AER_NONFATAL;
>> +
>> +	return 1;
>> +}
>> +
>
> You can set info->severity to AER_FATAL since that's the only type we
> enable DPC triggering.

DPC Trigger Enable
01b-DPC is enabled and is triggered when the Downstream Port detects
an unmasked uncorrectable error or when the Downstream Port receives an
ERR_FATAL Message.

DPC Trigger Reason
00b-DPC was triggered due to an unmasked uncorrectable error
reason == 0, due to detect an unmasked uncorrectable error, include non-fatal
and fatal error, so need to get the severity.

>
>>  static irqreturn_t dpc_handler(int irq, void *context)
>>  {
>>  	struct aer_err_info info;
>> @@ -229,9 +251,12 @@ static irqreturn_t dpc_handler(int irq, void *context)
>>  	/* show RP PIO error detail information */
>>  	if (dpc->rp_extensions && reason == 3 && ext_reason == 0)
>>  		dpc_process_rp_pio_error(dpc);
>> -	else if (reason == 0 && aer_get_device_error_info(pdev, &info)) {
>> +	else if (reason == 0 &&
>> +		 dpc_get_aer_uncorrect_severity(pdev, &info) &&
>> +		 aer_get_device_error_info(pdev, &info)) {
>>  		aer_print_error(pdev, &info);
>>  		pci_cleanup_aer_uncorrect_error_status(pdev);
>> +		pci_aer_clear_fatal_status(pdev);
>
> Good catch here, but let's clear the pending bits with a single call
> to pci_cleanup_aer_error_status_regs() rather than NONFATAL and
> FATAL separately.

pci_cleanup_aer_error_status_regs() also clear correctable error status.
seems not good enough as reason == 0 means detect an unmasked uncorrectable error.

Thanks,
Dongdong
>
> .
>