Re: [PATCH v3 4/4] drm/xe/pf: Handle MERT catastrophic errors

["Laguna, Lukasz" <lukasz.laguna@intel.com>]

On 11/27/2025 11:46, K V P, Satyanarayana wrote:
>> -----Original Message-----
>> From: Intel-xe <intel-xe-bounces@lists.freedesktop.org> On Behalf Of Lukasz
>> Laguna
>> Sent: Tuesday, November 25, 2025 12:33 AM
>> To: intel-xe@lists.freedesktop.org
>> Cc: Wajdeczko, Michal <Michal.Wajdeczko@intel.com>; Piorkowski, Piotr
>> <piotr.piorkowski@intel.com>; Laguna, Lukasz <lukasz.laguna@intel.com>
>> Subject: [PATCH v3 4/4] drm/xe/pf: Handle MERT catastrophic errors
>>
>> The MERT block triggers an interrupt when a catastrophic error occurs.
>> Update the interrupt handler to read the MERT catastrophic error type
>> and log appropriate debug message.
>>
>> Signed-off-by: Lukasz Laguna <lukasz.laguna@intel.com>
>> Reviewed-by: Piotr Piórkowski <piotr.piorkowski@intel.com>
>> ---
>> v3:
>>   - use FIELD_GET() for consistency
>> ---
>>   drivers/gpu/drm/xe/regs/xe_mert_regs.h |  5 +++++
>>   drivers/gpu/drm/xe/xe_mert.c           | 11 +++++++++++
>>   2 files changed, 16 insertions(+)
>>
>> diff --git a/drivers/gpu/drm/xe/regs/xe_mert_regs.h
>> b/drivers/gpu/drm/xe/regs/xe_mert_regs.h
>> index aef66c04901d..c345e11ceea8 100644
>> --- a/drivers/gpu/drm/xe/regs/xe_mert_regs.h
>> +++ b/drivers/gpu/drm/xe/regs/xe_mert_regs.h
>> @@ -10,6 +10,11 @@
>>
>>   #define MERT_LMEM_CFG				XE_REG(0x1448b0)
>>
>> +#define MERT_TLB_CT_INTR_ERR_ID_PORT		XE_REG(0x145190)
>> +#define   MERT_TLB_CT_VFID_MASK			REG_GENMASK(16,
>> 9)
>> +#define   MERT_TLB_CT_ERROR_MASK		REG_GENMASK(5, 0)
>> +#define     MERT_TLB_CT_LMTT_FAULT		0x05
>> +
>>   #define MERT_TLB_INV_DESC_A			XE_REG(0x14cf7c)
>>   #define   MERT_TLB_INV_DESC_A_VALID		REG_BIT(0)
>>
>> diff --git a/drivers/gpu/drm/xe/xe_mert.c b/drivers/gpu/drm/xe/xe_mert.c
>> index 304cc8421999..f7689e922953 100644
>> --- a/drivers/gpu/drm/xe/xe_mert.c
>> +++ b/drivers/gpu/drm/xe/xe_mert.c
>> @@ -55,10 +55,21 @@ void xe_mert_irq_handler(struct xe_device *xe, u32
>> master_ctl)
>>   	struct xe_tile *tile = xe_device_get_root_tile(xe);
>>   	unsigned long flags;
>>   	u32 reg_val;
>> +	u8 err;
>>
>>   	if (!(master_ctl & SOC_H2DMEMINT_IRQ))
>>   		return;
>>
>> +	reg_val = xe_mmio_read32(&tile->mmio,
>> MERT_TLB_CT_INTR_ERR_ID_PORT);
>> +	xe_mmio_write32(&tile->mmio, MERT_TLB_CT_INTR_ERR_ID_PORT,
>> 0);
>> +
>> +	err = FIELD_GET(MERT_TLB_CT_ERROR_MASK, reg_val);
>> +	if (err == MERT_TLB_CT_LMTT_FAULT)
>> +		drm_dbg(&xe->drm, "MERT catastrophic error: LMTT fault
>> (VF%u)\n",
>> +			FIELD_GET(MERT_TLB_CT_VFID_MASK, reg_val));
>> +	else if (err)
>> +		drm_dbg(&xe->drm, "MERT catastrophic error: Unexpected
>> fault (0x%x)\n", err);
>> +
> Shouldn't we log with drm_err when catastrophic error is received?
> -Satya.

In case of an LMTT fault, it's not a PF driver error, so we don't want 
to flood PF logs with errors for VF misbehavior.
For the "Unexpected fault" case, using drm_err() makes sense in my 
opinion, but it doesn't seem critical to change, especially since the 
patch is already merged.

Thanks,
Lukasz

>>   	spin_lock_irqsave(&tile->mert.lock, flags);
>>   	if (tile->mert.tlb_inv_triggered) {
>>   		reg_val = xe_mmio_read32(&tile->mmio,
>> MERT_TLB_INV_DESC_A);
>> --
>> 2.40.0