Donate to e Foundation | Murena handsets with /e/OS | Own a part of Murena! Learn more

Commit a411409e authored by George Shen's avatar George Shen
Browse files

msm: kgsl: Improve GPU snapshot design 



If GPU recovers from fault, KGSL shall continue doing GPU snapshot
in case GPU/GMU faults again. If GPU fault cannot be recovered,
stop taking new GPU snapshot. Keep the snapshot of last
unrecoverable snapshot for post-mortem debugging.

Change-Id: I988a7ee8f8b01dbf57bc4751708c5cf33bea835e
Signed-off-by: default avatarGeorge Shen <sqiao@codeaurora.org>
parent 1f7554fb

drivers/gpu/msm/kgsl_device.h

+2 −0
Original line number Diff line number Diff line
@@ -498,6 +498,7 @@ struct kgsl_device_private { 
 * @sysfs_read: Count of current reads via sysfs

 * @first_read: True until the snapshot read is started

 * @gmu_fault: Snapshot collected when GMU fault happened

 * @recovered: True if GPU was recovered after previous snapshot

 */

struct kgsl_snapshot {

	uint64_t ib1base;
@@ -521,6 +522,7 @@ struct kgsl_snapshot { 
	unsigned int sysfs_read;

	bool first_read;

	bool gmu_fault;

	bool recovered;

};



/**

drivers/gpu/msm/kgsl_gmu.c

+21 −23
Original line number Diff line number Diff line
@@ -1343,7 +1343,6 @@ void gmu_snapshot(struct kgsl_device *device) 
	struct adreno_device *adreno_dev = ADRENO_DEVICE(device);

	struct gmu_device *gmu = &device->gmu;



	if (!gmu->fault_count) {

	/* Mask so there's no interrupt caused by NMI */

	adreno_write_gmureg(adreno_dev,

			ADRENO_REG_GMU_GMU2HOST_INTR_MASK, 0xFFFFFFFF);
@@ -1365,7 +1364,6 @@ void gmu_snapshot(struct kgsl_device *device) 
	adreno_write_gmureg(adreno_dev,

			ADRENO_REG_GMU_GMU2HOST_INTR_MASK,

			(unsigned int) ~HFI_IRQ_MASK);

	}



	gmu->fault_count++;

}

drivers/gpu/msm/kgsl_pwrctrl.c

+3 −0
Original line number Diff line number Diff line
@@ -2691,6 +2691,9 @@ _aware(struct kgsl_device *device) 
				 * GPU will not be powered on

				 */

				WARN_ONCE(1, "Failed to recover GMU\n");

				device->snapshot->recovered = false;

			} else {

				device->snapshot->recovered = true;

			}



			clear_bit(GMU_FAULT, &gmu->flags);

drivers/gpu/msm/kgsl_snapshot.c

+2 −1
Original line number Diff line number Diff line
@@ -663,7 +663,7 @@ void kgsl_device_snapshot(struct kgsl_device *device, 
	 * Overwrite a non-GMU fault snapshot if a GMU fault occurs.

	 */

	if (device->snapshot != NULL) {

		if (!gmu_fault || device->snapshot->gmu_fault)

		if (!gmu_fault || !device->snapshot->recovered)

			return;



		/*
@@ -689,6 +689,7 @@ void kgsl_device_snapshot(struct kgsl_device *device, 
	snapshot->ptr = device->snapshot_memory.ptr;

	snapshot->remain = device->snapshot_memory.size;

	snapshot->gmu_fault = gmu_fault;

	snapshot->recovered = false;

	snapshot->first_read = true;

	snapshot->sysfs_read = 0;