Re: [PATCH v2] drm/xe/ras: Validate sysctrl response

"Tauro, Riana" <[email protected]> Mon, 3 Aug 2026 15:21:29 +0530
Newsgroups org.freedesktop.lists.intel-xe
Message-ID <[email protected]>
On 23-07-2026 19:59, Tauro, Riana wrote:
>
> On 22-07-2026 20:01, Raag Jadav wrote:
>> Validate errors raised by sysctrl firmware instead of blindly trusting
>> them.
>>
>> Signed-off-by: Raag Jadav <[email protected]>
>
>
> Looks good to me
>
> Reviewed-by: Riana Tauro <[email protected]>

Thank you for the patch.

Pushed to drm-xe-next.

>
>
>> ---
>> v2: Continue invalid uncorrectable error case (Riana)
>>      s/error/counter (Riana)
>> ---
>>   drivers/gpu/drm/xe/xe_ras.c | 30 ++++++++++++++++++++++++++++++
>>   1 file changed, 30 insertions(+)
>>
>> diff --git a/drivers/gpu/drm/xe/xe_ras.c b/drivers/gpu/drm/xe/xe_ras.c
>> index 5e294e9f8f99..28123f08e513 100644
>> --- a/drivers/gpu/drm/xe/xe_ras.c
>> +++ b/drivers/gpu/drm/xe/xe_ras.c
>> @@ -156,6 +156,24 @@ static inline const char *comp_to_str(u8 component)
>>       return xe_ras_components[component];
>>   }
>>   +static bool ras_counter_is_valid(struct xe_device *xe, struct 
>> xe_ras_error_class *counter)
>> +{
>> +    u8 severity = counter->common.severity;
>> +    u8 component = counter->common.component;
>> +
>> +    if (!in_range(severity, XE_RAS_SEV_NOT_SUPPORTED + 1, 
>> XE_RAS_SEV_MAX - 1)) {
>> +        xe_err(xe, "sysctrl: unexpected severity %u\n", severity);
>> +        return false;
>> +    }
>> +
>> +    if (!in_range(component, XE_RAS_COMP_NOT_SUPPORTED + 1, 
>> XE_RAS_COMP_MAX - 1)) {
>> +        xe_err(xe, "sysctrl: unexpected component %u\n", component);
>> +        return false;
>> +    }
>> +
>> +    return true;
>> +}
>> +
>>   static struct pci_dev *find_usp_dev(struct pci_dev *pdev)
>>   {
>>       struct pci_dev *vsp;
>> @@ -327,6 +345,9 @@ void xe_ras_counter_threshold_crossed(struct 
>> xe_device *xe,
>>           severity = errors[id].common.severity;
>>           component = errors[id].common.component;
>>   +        if (!ras_counter_is_valid(xe, &errors[id]))
>> +            continue;
>> +
>>           xe_warn(xe, "[RAS]: %s %s detected\n",
>>               comp_to_str(component), sev_to_str(severity));
>>       }
>> @@ -358,6 +379,9 @@ static int get_counter(struct xe_device *xe, 
>> struct xe_ras_error_class *counter,
>>           return -EIO;
>>       }
>>   +    if (!ras_counter_is_valid(xe, &response.counter))
>> +        return -EBADMSG;
>> +
>>       common = &response.counter.common;
>>       *value = response.value;
>>   @@ -422,6 +446,9 @@ enum xe_ras_recovery_action 
>> xe_ras_process_errors(struct xe_device *xe)
>>               component = arr->counter.common.component;
>>               severity = arr->counter.common.severity;
>>   +            if (!ras_counter_is_valid(xe, &arr->counter))
>> +                continue;
>> +
>>               xe_info(xe, "[RAS]: %s %s detected\n", 
>> comp_to_str(component),
>>                   sev_to_str(severity));
>>   @@ -532,6 +559,9 @@ int xe_ras_clear_counter(struct xe_device *xe, 
>> u8 severity, u8 component)
>>         counter = &response.counter;
>>   +    if (!ras_counter_is_valid(xe, counter))
>> +        return -EBADMSG;
>> +
>>       xe_dbg(xe, "[RAS]: clear counter for %s %s\n", 
>> comp_to_str(counter->common.component),
>>              sev_to_str(counter->common.severity));