Skip to contentSkip to conversation

drm/amdgpu: Fix module unload hang by KIQ IRQ set

PATCH In review Trigger Huang 3 participants 3 messages Feb 20, 2017 1 message not in the archive

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 drm/amdgpu: Fix module unload hang by KIQ IRQ set

In some cases, manually insmod/rmmod amdgpu is necessary. When unloading amdgpu, the KIQ IRQ enable/disable function will case system hang. The root cause is, in the sequence of function amdgpu_fini, the sw_fini of IP block AMD_IP_BLOCK_TYPE_GFX will be invoked earlier than that of AMD_IP_BLOCK_TYPE_IH. So continue to use the variable freed by AMD_IP_BLOCK_TYPE_GFX will cause system hang.

Changes in v2:
-- Refine according to Xiangliang Yu's suggestions

---
 drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 10 ++++------
 1 file changed, 4 insertions(+), 6 deletions(-)
drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +4 -6
index 5a484ef..c417324 100644
@@ -1395,7 +1395,6 @@ static int gfx_v8_0_kiq_init_ring(struct amdgpu_device *adev,
 		ring->pipe = 1;
 	}
 
-	irq->data = ring;
 	ring->queue = 0;
 	sprintf(ring->name, "kiq %d.%d.%d", ring->me, ring->pipe, ring->queue);
 	r = amdgpu_ring_init(adev, ring, 1024,
@@ -1410,7 +1409,6 @@ static void gfx_v8_0_kiq_free_ring(struct amdgpu_ring *ring,
 {
 	amdgpu_wb_free(ring->adev, ring->adev->virt.reg_val_offs);
 	amdgpu_ring_fini(ring);
-	irq->data = NULL;
 }
 
 #define MEC_HPD_SIZE 2048
@@ -6931,9 +6929,9 @@ static int gfx_v8_0_kiq_set_interrupt_state(struct amdgpu_device *adev,
 					    enum amdgpu_interrupt_state state)
 {
 	uint32_t tmp, target;
-	struct amdgpu_ring *ring = (struct amdgpu_ring *)src->data;
+	struct amdgpu_ring *ring = &(adev->gfx.kiq.ring);
 
-	BUG_ON(!ring || (ring->funcs->type != AMDGPU_RING_TYPE_KIQ));
+	BUG_ON(ring->funcs->type != AMDGPU_RING_TYPE_KIQ);
 
 	if (ring->me == 1)
 		target = mmCP_ME1_PIPE0_INT_CNTL;
@@ -6977,9 +6975,9 @@ static int gfx_v8_0_kiq_irq(struct amdgpu_device *adev,
 			    struct amdgpu_iv_entry *entry)
 {
 	u8 me_id, pipe_id, queue_id;
-	struct amdgpu_ring *ring = (struct amdgpu_ring *)source->data;
+	struct amdgpu_ring *ring = &(adev->gfx.kiq.ring);
 
-	BUG_ON(!ring || (ring->funcs->type != AMDGPU_RING_TYPE_KIQ));
+	BUG_ON(ring->funcs->type != AMDGPU_RING_TYPE_KIQ);
 
 	me_id = (entry->ring_id & 0x0c) >> 2;
 	pipe_id = (entry->ring_id & 0x03) >> 0;
Signature · 6 lines
2.7.4

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

A message that is not in the archive (sent off-list or not yet received): 1487559799-17644-1-git-send-email-trigger.huang-5C7GfCeVMHo@public.gmane.org

Reviewed-by Xiangliang Yu
#2 lore raw

Thanks!
Xiangliang Yu

Quoted 70 lines

-----Original Message-----
From: Trigger Huang [mailto:[email protected]]
Sent: Monday, February 20, 2017 11:03 AM
To: [email protected]
Cc: Liu, Monk <[email protected]>; Yu, Xiangliang <[email protected]>; Huang, Trigger <[email protected]>
Subject: [PATCH V2] drm/amdgpu: Fix module unload hang by KIQ IRQ set

In some cases, manually insmod/rmmod amdgpu is necessary. When unloading amdgpu, the KIQ IRQ enable/disable function will case system hang. The root cause is, in the sequence of function amdgpu_fini, the sw_fini of IP block AMD_IP_BLOCK_TYPE_GFX will be invoked earlier than that of AMD_IP_BLOCK_TYPE_IH. So continue to use the variable freed by AMD_IP_BLOCK_TYPE_GFX will cause system hang.

Changes in v2:
-- Refine according to Xiangliang Yu's suggestions

---
 drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 10 ++++------
 1 file changed, 4 insertions(+), 6 deletions(-)
drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +0 -1
index 5a484ef..c417324 100644
@@ -1395,7 +1395,6 @@ static int gfx_v8_0_kiq_init_ring(struct
 amdgpu_device *adev,
 		ring->pipe = 1;
 	}
 
-	irq->data = ring;
 	ring->queue = 0;
 	sprintf(ring->name, "kiq %d.%d.%d", ring->me, ring->pipe, ring-
Quoted 1 line

queue);

r = amdgpu_ring_init(adev, ring, 1024, @@ -1410,7 +1409,6 @@ static
void gfx_v8_0_kiq_free_ring(struct amdgpu_ring *ring, {

drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +2 -3
 	amdgpu_wb_free(ring->adev, ring->adev->virt.reg_val_offs);
 	amdgpu_ring_fini(ring);
-	irq->data = NULL;
 }
 
 #define MEC_HPD_SIZE 2048
@@ -6931,9 +6929,9 @@ static int gfx_v8_0_kiq_set_interrupt_state(struct
 amdgpu_device *adev,
 					    enum amdgpu_interrupt_state
 state)  {
 	uint32_t tmp, target;
-	struct amdgpu_ring *ring = (struct amdgpu_ring *)src->data;
+	struct amdgpu_ring *ring = &(adev->gfx.kiq.ring);
 
-	BUG_ON(!ring || (ring->funcs->type != AMDGPU_RING_TYPE_KIQ));
+	BUG_ON(ring->funcs->type != AMDGPU_RING_TYPE_KIQ);
 
 	if (ring->me == 1)
drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +2 -2
 		target = mmCP_ME1_PIPE0_INT_CNTL;
@@ -6977,9 +6975,9 @@ static int gfx_v8_0_kiq_irq(struct amdgpu_device
 *adev,
 			    struct amdgpu_iv_entry *entry)
 {
 	u8 me_id, pipe_id, queue_id;
-	struct amdgpu_ring *ring = (struct amdgpu_ring *)source->data;
+	struct amdgpu_ring *ring = &(adev->gfx.kiq.ring);
 
-	BUG_ON(!ring || (ring->funcs->type != AMDGPU_RING_TYPE_KIQ));
+	BUG_ON(ring->funcs->type != AMDGPU_RING_TYPE_KIQ);
 
 	me_id = (entry->ring_id & 0x0c) >> 2;
Quoted 3 lines

pipe_id = (entry->ring_id & 0x03) >> 0;
--
2.7.4

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

#3 lore raw
In reply to a message not in the archive

Am 20.02.2017 um 04:03 schrieb Trigger Huang:

Quoted 11 lines

In some cases, manually insmod/rmmod amdgpu is necessary. When unloading amdgpu, the KIQ IRQ enable/disable function will case system hang. The root cause is, in the sequence of function amdgpu_fini, the sw_fini of IP block AMD_IP_BLOCK_TYPE_GFX will be invoked earlier than that of AMD_IP_BLOCK_TYPE_IH. So continue to use the variable freed by AMD_IP_BLOCK_TYPE_GFX will cause system hang.

Changes in v2:
-- Refine according to Xiangliang Yu's suggestions

Quoted 38 lines
---
  drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 10 ++++------
  1 file changed, 4 insertions(+), 6 deletions(-)
drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +2 -4
index 5a484ef..c417324 100644
@@ -1395,7 +1395,6 @@ static int gfx_v8_0_kiq_init_ring(struct amdgpu_device *adev,
  		ring->pipe = 1;
  	}
 
-	irq->data = ring;
  	ring->queue = 0;
  	sprintf(ring->name, "kiq %d.%d.%d", ring->me, ring->pipe, ring->queue);
  	r = amdgpu_ring_init(adev, ring, 1024,
@@ -1410,7 +1409,6 @@ static void gfx_v8_0_kiq_free_ring(struct amdgpu_ring *ring,
  {
  	amdgpu_wb_free(ring->adev, ring->adev->virt.reg_val_offs);
  	amdgpu_ring_fini(ring);
-	irq->data = NULL;
  }
 
  #define MEC_HPD_SIZE 2048
@@ -6931,9 +6929,9 @@ static int gfx_v8_0_kiq_set_interrupt_state(struct amdgpu_device *adev,
  					    enum amdgpu_interrupt_state state)
  {
  	uint32_t tmp, target;
-	struct amdgpu_ring *ring = (struct amdgpu_ring *)src->data;
+	struct amdgpu_ring *ring = &(adev->gfx.kiq.ring);
 
-	BUG_ON(!ring || (ring->funcs->type != AMDGPU_RING_TYPE_KIQ));
+	BUG_ON(ring->funcs->type != AMDGPU_RING_TYPE_KIQ);
 
  	if (ring->me == 1)
  		target = mmCP_ME1_PIPE0_INT_CNTL;
@@ -6977,9 +6975,9 @@ static int gfx_v8_0_kiq_irq(struct amdgpu_device *adev,
  			    struct amdgpu_iv_entry *entry)

drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c static int gfx_v8_0_kiq_irq(struct amdgpu_device *adev, · line 6983

Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +2 -2
  {
  	u8 me_id, pipe_id, queue_id;
-	struct amdgpu_ring *ring = (struct amdgpu_ring *)source->data;
+	struct amdgpu_ring *ring = &(adev->gfx.kiq.ring);
 
-	BUG_ON(!ring || (ring->funcs->type != AMDGPU_RING_TYPE_KIQ));
+	BUG_ON(ring->funcs->type != AMDGPU_RING_TYPE_KIQ);
 
  	me_id = (entry->ring_id & 0x0c) >> 2;
  	pipe_id = (entry->ring_id & 0x03) >> 0;

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx