Re: [PATCH v15 10/16] PCI/CXL: Add CXL Device Reset sequencing

[Srirangan Madhavan <smadhavan@nvidia.com>]

On 10/27/26 8:01 PM, Li Ming wrote:
>> +     if (PCI_POSSIBLE_ERROR(status2))
>> +             return -ENODEV;
>> +     if (status2 & PCI_DVSEC_CXL_CACHE_INV)
>> +             return 0;
> I think checking again after deadline is not needed, if it usually
> happens, it means we should adjust CXL_CACHE_WBI_TIMEOUT_US.
>> +
>> +     return -ETIMEDOUT;
> It is better to output a warning before returning -ETIMEDOUT.
>> +}
Ack. Fixes this in v16 and added warining.

>> +
>> +     idx = FIELD_GET(PCI_DVSEC_CXL_RST_TIMEOUT, cap);
>> +     if (idx >= ARRAY_SIZE(cxl_reset_timeout_ms)) {
>> +             pci_warn(pdev,
>> +                      "unknown CXL reset timeout encoding %d; using 
>> %u ms\n",
>> +                      idx, CXL_RESET_MAX_WAIT_MS);
>> +             idx = ARRAY_SIZE(cxl_reset_timeout_ms) - 1;
>> +     }
>> +
> I noticed that the worst situation is that we have to wait cxl reset
> done for 100s. If my understanding is correct, userspace would hang on
> "echo 1 > reset" until reset done. Is it better to add a pci_info() here
> to output the reset timeout value?

Done. Added a log print.

-- 
Regards,
Srirangan