drm/amdgpu: remove old lockup detection infrastructure

It didn't worked to well anyway.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Chunming Zhou <david1.zhou@amd.com>
Reviewed-by: Junwei Zhang <Jerry.Zhang@amd.com>

drivers/gpu/drm/amd/amdgpu/amdgpu.h

@@ -345,7 +345,6 @@ struct amdgpu_ring_funcs {
 	/* testing functions */
 	int (*test_ring)(struct amdgpu_ring *ring);
 	int (*test_ib)(struct amdgpu_ring *ring);
-	bool (*is_lockup)(struct amdgpu_ring *ring);
 	/* insert NOP packets */
 	void (*insert_nop)(struct amdgpu_ring *ring, uint32_t count);
 };
@@ -907,8 +906,6 @@ struct amdgpu_ring {
 	unsigned		ring_size;
 	unsigned		ring_free_dw;
 	int			count_dw;
-	atomic_t		last_rptr;
-	atomic64_t		last_activity;
 	uint64_t		gpu_addr;
 	uint32_t		align_mask;
 	uint32_t		ptr_mask;
@@ -1230,8 +1227,6 @@ void amdgpu_ring_commit(struct amdgpu_ring *ring);
 void amdgpu_ring_unlock_commit(struct amdgpu_ring *ring);
 void amdgpu_ring_undo(struct amdgpu_ring *ring);
 void amdgpu_ring_unlock_undo(struct amdgpu_ring *ring);
-void amdgpu_ring_lockup_update(struct amdgpu_ring *ring);
-bool amdgpu_ring_test_lockup(struct amdgpu_ring *ring);
 unsigned amdgpu_ring_backup(struct amdgpu_ring *ring,
 			    uint32_t **data);
 int amdgpu_ring_restore(struct amdgpu_ring *ring,
@@ -1974,7 +1969,6 @@ struct amdgpu_device {
 	bool				suspend;
 	bool				need_dma32;
 	bool				accel_working;
-	bool				needs_reset;
 	struct work_struct 		reset_work;
 	struct notifier_block		acpi_nb;
 	struct amdgpu_i2c_chan		*i2c_bus[AMDGPU_MAX_I2C_BUS];
@@ -2253,7 +2247,6 @@ amdgpu_get_sdma_instance(struct amdgpu_ring *ring)
 #define amdgpu_ring_parse_cs(r, p, ib) ((r)->funcs->parse_cs((p), (ib)))
 #define amdgpu_ring_test_ring(r) (r)->funcs->test_ring((r))
 #define amdgpu_ring_test_ib(r) (r)->funcs->test_ib((r))
-#define amdgpu_ring_is_lockup(r) (r)->funcs->is_lockup((r))
 #define amdgpu_ring_get_rptr(r) (r)->funcs->get_rptr((r))
 #define amdgpu_ring_get_wptr(r) (r)->funcs->get_wptr((r))
 #define amdgpu_ring_set_wptr(r) (r)->funcs->set_wptr((r))

drivers/gpu/drm/amd/amdgpu/amdgpu_device.c

@@ -1816,12 +1816,6 @@ int amdgpu_gpu_reset(struct amdgpu_device *adev)
 
 	down_write(&adev->exclusive_lock);
 
-	if (!adev->needs_reset) {
-		up_write(&adev->exclusive_lock);
-		return 0;
-	}
-
-	adev->needs_reset = false;
 	atomic_inc(&adev->gpu_reset_counter);
 
 	/* block TTM */

drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c

@@ -269,17 +269,6 @@ static void amdgpu_fence_check_lockup(struct work_struct *work)
 	if (amdgpu_fence_activity(ring)) {
 		wake_up_all(&ring->fence_drv.fence_queue);
 	}
-	else if (amdgpu_ring_is_lockup(ring)) {
-		/* good news we believe it's a lockup */
-		dev_warn(ring->adev->dev, "GPU lockup (current fence id "
-			"0x%016llx last fence id 0x%016llx on ring %d)\n",
-			(uint64_t)atomic64_read(&fence_drv->last_seq),
-			fence_drv->sync_seq[ring->idx], ring->idx);
-
-		/* remember that we need an reset */
-		ring->adev->needs_reset = true;
-		wake_up_all(&ring->fence_drv.fence_queue);
-	}
 	up_read(&ring->adev->exclusive_lock);
 }
 
@@ -380,7 +369,6 @@ static bool amdgpu_fence_enable_signaling(struct fence *f)
  */
 static int amdgpu_fence_ring_wait_seq(struct amdgpu_ring *ring, uint64_t seq)
 {
-	struct amdgpu_device *adev = ring->adev;
 	bool signaled = false;
 
 	BUG_ON(!ring);
@@ -391,8 +379,7 @@ static int amdgpu_fence_ring_wait_seq(struct amdgpu_ring *ring, uint64_t seq)
 		return 0;
 
 	wait_event(ring->fence_drv.fence_queue, (
-		   (signaled = amdgpu_fence_seq_signaled(ring, seq))
-		   || adev->needs_reset));
+		   (signaled = amdgpu_fence_seq_signaled(ring, seq))));
 
 	if (signaled)
 		return 0;
@@ -939,11 +926,6 @@ signed long amdgpu_fence_wait_any(struct amdgpu_device *adev,
 		if (amdgpu_test_signaled_any(array, count))
 			break;
 
-		if (adev->needs_reset) {
-			t = -EDEADLK;
-			break;
-		}
-
 		t = schedule_timeout(t);
 
 		if (t > 0 && intr && signal_pending(current))

drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c

@@ -298,7 +298,6 @@ int amdgpu_ib_ring_tests(struct amdgpu_device *adev)
 		r = amdgpu_ring_test_ib(ring);
 		if (r) {
 			ring->ready = false;
-			adev->needs_reset = false;
 
 			if (ring == &adev->gfx.gfx_ring[0]) {
 				/* oh, oh, that's really bad */

drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c

@@ -67,8 +67,6 @@ void amdgpu_ring_free_size(struct amdgpu_ring *ring)
 	if (!ring->ring_free_dw) {
 		/* this is an empty ring */
 		ring->ring_free_dw = ring->ring_size / 4;
-		/*  update lockup info to avoid false positive */
-		amdgpu_ring_lockup_update(ring);
 	}
 }
 
@@ -208,46 +206,6 @@ void amdgpu_ring_unlock_undo(struct amdgpu_ring *ring)
 	mutex_unlock(ring->ring_lock);
 }
 
-/**
- * amdgpu_ring_lockup_update - update lockup variables
- *
- * @ring: amdgpu_ring structure holding ring information
- *
- * Update the last rptr value and timestamp (all asics).
- */
-void amdgpu_ring_lockup_update(struct amdgpu_ring *ring)
-{
-	atomic_set(&ring->last_rptr, amdgpu_ring_get_rptr(ring));
-	atomic64_set(&ring->last_activity, jiffies_64);
-}
-
-/**
- * amdgpu_ring_test_lockup() - check if ring is lockedup by recording information
- * @ring:       amdgpu_ring structure holding ring information
- *
- */
-bool amdgpu_ring_test_lockup(struct amdgpu_ring *ring)
-{
-	uint32_t rptr = amdgpu_ring_get_rptr(ring);
-	uint64_t last = atomic64_read(&ring->last_activity);
-	uint64_t elapsed;
-
-	if (rptr != atomic_read(&ring->last_rptr)) {
-		/* ring is still working, no lockup */
-		amdgpu_ring_lockup_update(ring);
-		return false;
-	}
-
-	elapsed = jiffies_to_msecs(jiffies_64 - last);
-	if (amdgpu_lockup_timeout && elapsed >= amdgpu_lockup_timeout) {
-		dev_err(ring->adev->dev, "ring %d stalled for more than %llumsec\n",
-			ring->idx, elapsed);
-		return true;
-	}
-	/* give a chance to the GPU ... */
-	return false;
-}
-
 /**
  * amdgpu_ring_backup - Back up the content of a ring
  *
@@ -436,7 +394,6 @@ int amdgpu_ring_init(struct amdgpu_device *adev, struct amdgpu_ring *ring,
 	if (amdgpu_debugfs_ring_init(adev, ring)) {
 		DRM_ERROR("Failed to register debugfs file for rings !\n");
 	}
-	amdgpu_ring_lockup_update(ring);
 	return 0;
 }
 

drivers/gpu/drm/amd/amdgpu/cik_sdma.c

@@ -1290,24 +1290,6 @@ const struct amd_ip_funcs cik_sdma_ip_funcs = {
 	.set_powergating_state = cik_sdma_set_powergating_state,
 };
 
-/**
- * cik_sdma_ring_is_lockup - Check if the DMA engine is locked up
- *
- * @ring: amdgpu_ring structure holding ring information
- *
- * Check if the async DMA engine is locked up (CIK).
- * Returns true if the engine appears to be locked up, false if not.
- */
-static bool cik_sdma_ring_is_lockup(struct amdgpu_ring *ring)
-{
-
-	if (cik_sdma_is_idle(ring->adev)) {
-		amdgpu_ring_lockup_update(ring);
-		return false;
-	}
-	return amdgpu_ring_test_lockup(ring);
-}
-
 static const struct amdgpu_ring_funcs cik_sdma_ring_funcs = {
 	.get_rptr = cik_sdma_ring_get_rptr,
 	.get_wptr = cik_sdma_ring_get_wptr,
@@ -1320,7 +1302,6 @@ static const struct amdgpu_ring_funcs cik_sdma_ring_funcs = {
 	.emit_hdp_flush = cik_sdma_ring_emit_hdp_flush,
 	.test_ring = cik_sdma_ring_test_ring,
 	.test_ib = cik_sdma_ring_test_ib,
-	.is_lockup = cik_sdma_ring_is_lockup,
 	.insert_nop = cik_sdma_ring_insert_nop,
 };
 

drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c

@@ -5542,24 +5542,6 @@ const struct amd_ip_funcs gfx_v7_0_ip_funcs = {
 	.set_powergating_state = gfx_v7_0_set_powergating_state,
 };
 
-/**
- * gfx_v7_0_ring_is_lockup - check if the 3D engine is locked up
- *
- * @adev: amdgpu_device pointer
- * @ring: amdgpu_ring structure holding ring information
- *
- * Check if the 3D engine is locked up (CIK).
- * Returns true if the engine is locked, false if not.
- */
-static bool gfx_v7_0_ring_is_lockup(struct amdgpu_ring *ring)
-{
-	if (gfx_v7_0_is_idle(ring->adev)) {
-		amdgpu_ring_lockup_update(ring);
-		return false;
-	}
-	return amdgpu_ring_test_lockup(ring);
-}
-
 static const struct amdgpu_ring_funcs gfx_v7_0_ring_funcs_gfx = {
 	.get_rptr = gfx_v7_0_ring_get_rptr_gfx,
 	.get_wptr = gfx_v7_0_ring_get_wptr_gfx,
@@ -5573,7 +5555,6 @@ static const struct amdgpu_ring_funcs gfx_v7_0_ring_funcs_gfx = {
 	.emit_hdp_flush = gfx_v7_0_ring_emit_hdp_flush,
 	.test_ring = gfx_v7_0_ring_test_ring,
 	.test_ib = gfx_v7_0_ring_test_ib,
-	.is_lockup = gfx_v7_0_ring_is_lockup,
 	.insert_nop = amdgpu_ring_insert_nop,
 };
 
@@ -5590,7 +5571,6 @@ static const struct amdgpu_ring_funcs gfx_v7_0_ring_funcs_compute = {
 	.emit_hdp_flush = gfx_v7_0_ring_emit_hdp_flush,
 	.test_ring = gfx_v7_0_ring_test_ring,
 	.test_ib = gfx_v7_0_ring_test_ib,
-	.is_lockup = gfx_v7_0_ring_is_lockup,
 	.insert_nop = amdgpu_ring_insert_nop,
 };
 

drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c

@@ -4075,15 +4075,6 @@ static void gfx_v8_0_ring_emit_vm_flush(struct amdgpu_ring *ring,
 	}
 }
 
-static bool gfx_v8_0_ring_is_lockup(struct amdgpu_ring *ring)
-{
-	if (gfx_v8_0_is_idle(ring->adev)) {
-		amdgpu_ring_lockup_update(ring);
-		return false;
-	}
-	return amdgpu_ring_test_lockup(ring);
-}
-
 static u32 gfx_v8_0_ring_get_rptr_compute(struct amdgpu_ring *ring)
 {
 	return ring->adev->wb.wb[ring->rptr_offs];
@@ -4365,7 +4356,6 @@ static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_gfx = {
 	.emit_hdp_flush = gfx_v8_0_ring_emit_hdp_flush,
 	.test_ring = gfx_v8_0_ring_test_ring,
 	.test_ib = gfx_v8_0_ring_test_ib,
-	.is_lockup = gfx_v8_0_ring_is_lockup,
 	.insert_nop = amdgpu_ring_insert_nop,
 };
 
@@ -4382,7 +4372,6 @@ static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_compute = {
 	.emit_hdp_flush = gfx_v8_0_ring_emit_hdp_flush,
 	.test_ring = gfx_v8_0_ring_test_ring,
 	.test_ib = gfx_v8_0_ring_test_ib,
-	.is_lockup = gfx_v8_0_ring_is_lockup,
 	.insert_nop = amdgpu_ring_insert_nop,
 };
 

drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c

@@ -1295,24 +1295,6 @@ const struct amd_ip_funcs sdma_v2_4_ip_funcs = {
 	.set_powergating_state = sdma_v2_4_set_powergating_state,
 };
 
-/**
- * sdma_v2_4_ring_is_lockup - Check if the DMA engine is locked up
- *
- * @ring: amdgpu_ring structure holding ring information
- *
- * Check if the async DMA engine is locked up (VI).
- * Returns true if the engine appears to be locked up, false if not.
- */
-static bool sdma_v2_4_ring_is_lockup(struct amdgpu_ring *ring)
-{
-
-	if (sdma_v2_4_is_idle(ring->adev)) {
-		amdgpu_ring_lockup_update(ring);
-		return false;
-	}
-	return amdgpu_ring_test_lockup(ring);
-}
-
 static const struct amdgpu_ring_funcs sdma_v2_4_ring_funcs = {
 	.get_rptr = sdma_v2_4_ring_get_rptr,
 	.get_wptr = sdma_v2_4_ring_get_wptr,
@@ -1325,7 +1307,6 @@ static const struct amdgpu_ring_funcs sdma_v2_4_ring_funcs = {
 	.emit_hdp_flush = sdma_v2_4_ring_emit_hdp_flush,
 	.test_ring = sdma_v2_4_ring_test_ring,
 	.test_ib = sdma_v2_4_ring_test_ib,
-	.is_lockup = sdma_v2_4_ring_is_lockup,
 	.insert_nop = sdma_v2_4_ring_insert_nop,
 };
 

drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c

@@ -1428,24 +1428,6 @@ const struct amd_ip_funcs sdma_v3_0_ip_funcs = {
 	.set_powergating_state = sdma_v3_0_set_powergating_state,
 };
 
-/**
- * sdma_v3_0_ring_is_lockup - Check if the DMA engine is locked up
- *
- * @ring: amdgpu_ring structure holding ring information
- *
- * Check if the async DMA engine is locked up (VI).
- * Returns true if the engine appears to be locked up, false if not.
- */
-static bool sdma_v3_0_ring_is_lockup(struct amdgpu_ring *ring)
-{
-
-	if (sdma_v3_0_is_idle(ring->adev)) {
-		amdgpu_ring_lockup_update(ring);
-		return false;
-	}
-	return amdgpu_ring_test_lockup(ring);
-}
-
 static const struct amdgpu_ring_funcs sdma_v3_0_ring_funcs = {
 	.get_rptr = sdma_v3_0_ring_get_rptr,
 	.get_wptr = sdma_v3_0_ring_get_wptr,
@@ -1458,7 +1440,6 @@ static const struct amdgpu_ring_funcs sdma_v3_0_ring_funcs = {
 	.emit_hdp_flush = sdma_v3_0_ring_emit_hdp_flush,
 	.test_ring = sdma_v3_0_ring_test_ring,
 	.test_ib = sdma_v3_0_ring_test_ib,
-	.is_lockup = sdma_v3_0_ring_is_lockup,
 	.insert_nop = sdma_v3_0_ring_insert_nop,
 };
 

drivers/gpu/drm/amd/amdgpu/uvd_v4_2.c

@@ -885,7 +885,6 @@ static const struct amdgpu_ring_funcs uvd_v4_2_ring_funcs = {
 	.emit_semaphore = uvd_v4_2_ring_emit_semaphore,
 	.test_ring = uvd_v4_2_ring_test_ring,
 	.test_ib = uvd_v4_2_ring_test_ib,
-	.is_lockup = amdgpu_ring_test_lockup,
 	.insert_nop = amdgpu_ring_insert_nop,
 };
 

drivers/gpu/drm/amd/amdgpu/uvd_v5_0.c

@@ -824,7 +824,6 @@ static const struct amdgpu_ring_funcs uvd_v5_0_ring_funcs = {
 	.emit_semaphore = uvd_v5_0_ring_emit_semaphore,
 	.test_ring = uvd_v5_0_ring_test_ring,
 	.test_ib = uvd_v5_0_ring_test_ib,
-	.is_lockup = amdgpu_ring_test_lockup,
 	.insert_nop = amdgpu_ring_insert_nop,
 };
 

drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c

@@ -808,7 +808,6 @@ static const struct amdgpu_ring_funcs uvd_v6_0_ring_funcs = {
 	.emit_semaphore = uvd_v6_0_ring_emit_semaphore,
 	.test_ring = uvd_v6_0_ring_test_ring,
 	.test_ib = uvd_v6_0_ring_test_ib,
-	.is_lockup = amdgpu_ring_test_lockup,
 	.insert_nop = amdgpu_ring_insert_nop,
 };
 

drivers/gpu/drm/amd/amdgpu/vce_v2_0.c

@@ -642,7 +642,6 @@ static const struct amdgpu_ring_funcs vce_v2_0_ring_funcs = {
 	.emit_semaphore = amdgpu_vce_ring_emit_semaphore,
 	.test_ring = amdgpu_vce_ring_test_ring,
 	.test_ib = amdgpu_vce_ring_test_ib,
-	.is_lockup = amdgpu_ring_test_lockup,
 	.insert_nop = amdgpu_ring_insert_nop,
 };
 

drivers/gpu/drm/amd/amdgpu/vce_v3_0.c

@@ -643,7 +643,6 @@ static const struct amdgpu_ring_funcs vce_v3_0_ring_funcs = {
 	.emit_semaphore = amdgpu_vce_ring_emit_semaphore,
 	.test_ring = amdgpu_vce_ring_test_ring,
 	.test_ib = amdgpu_vce_ring_test_ib,
-	.is_lockup = amdgpu_ring_test_lockup,
 	.insert_nop = amdgpu_ring_insert_nop,
 };