Re: [PATCH v3 23/23] drm/xe/pci: Report 'cannot re-enable' error using SIGID

[Michal Wajdeczko <michal.wajdeczko@intel.com>]

On 8/12/2026 8:58 AM, Mallesh, Koujalagi wrote:
> 
> On 30-07-2026 08:51 pm, Michal Wajdeczko wrote:
>> Report 'cannot re-enable PCI device' error using xe_log() helper.
>>
>> Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
>> Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
>> Cc: Riana Tauro <riana.tauro@intel.com>
>> Cc: Aravind Iddamsetty <aravind.iddamsetty@intel.com>
>> Cc: Mallesh Koujalagi <mallesh.koujalagi@intel.com>
>> ---
>>   drivers/gpu/drm/xe/xe_pci_error.c | 7 +++++--
>>   1 file changed, 5 insertions(+), 2 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/xe/xe_pci_error.c b/drivers/gpu/drm/xe/xe_pci_error.c
>> index e41af2ac7f23..fb965f52be5e 100644
>> --- a/drivers/gpu/drm/xe/xe_pci_error.c
>> +++ b/drivers/gpu/drm/xe/xe_pci_error.c
>> @@ -7,6 +7,7 @@
>>     #include "xe_device.h"
>>   #include "xe_gt.h"
>> +#include "xe_log.h"
>>   #include "xe_pci.h"
>>   #include "xe_pm.h"
>>   #include "xe_printk.h"
>> @@ -90,13 +91,15 @@ static pci_ers_result_t xe_pci_error_slot_reset(struct pci_dev *pdev)
>>   {
>>       const struct pci_device_id *ent = pci_match_id(pdev->driver->id_table, pdev);
>>       struct xe_device *xe = pdev_to_xe_device(pdev);
>> +    int err;
>>         xe_info(xe, "PCI error: slot reset\n");
> Please use xe_log_info.

but do we care now?

it's info only and I guess we should prioritize to convert
real errors into SIGID logs first

>>         pci_restore_state(pdev);
>>   -    if (pci_enable_device(pdev)) {
>> -        xe_err(xe, "Cannot re-enable PCI device after reset\n");
>> +    err = pci_enable_device(pdev);
>> +    if (err) {
>> +        xe_log_err_fatal(xe, PCI, err, "Cannot re-enable PCI device after reset\n");
> 
> PCI component used XE_SIGID_SW looks semantic mismatch, we can add DRIVER_HARDWARE with PCI 

but introduced in patch 3/23 component PCI represents the component
that is beyond our Xe driver (part of the SYSTEM) as we don't manage
that directly

and since our SIGID choices are limited, the generic SIGID_SW is
the most applicable IMO

if you disagree, please comment patch 3/23 as it is already
reviewed

> and make
> 
> a PCIe link failure that required active recovery right?

this code is already part of the HW error handling
maybe we should (or there is) declare wedged

but I don't want to make here any changes beyond simple
conversion from xe_err() to xe_log_err()

> 
> Thanks,
> 
> -/Mallesh
> 
>>           return PCI_ERS_RESULT_DISCONNECT;
>>       }
>>