Skip to contentSkip to conversation

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

PATCH Superseded Trigger Huang 3 participants 4 messages Feb 16, 2017 – Feb 17, 2017 2 messages not in the archive

Subsystems (2)

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

PATCH 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.

---
 drivers/gpu/drm/amd/amdgpu/amdgpu.h   |  2 ++
 drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 17 ++++++++---------
 2 files changed, 10 insertions(+), 9 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu.h +2 -0
index 1db2e7b..be43d09 100644
@@ -793,6 +793,8 @@ struct amdgpu_kiq {
 	struct amdgpu_bo	*eop_obj;
 	struct amdgpu_ring	ring;
 	struct amdgpu_irq_src	irq;
+	u32			me;
+	u32			pipe;
 };
 
 /*
drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +8 -9
index 772c42b..04e2a5b 100644
@@ -1390,12 +1390,15 @@ static int gfx_v8_0_kiq_init_ring(struct amdgpu_device *adev,
 	if (adev->gfx.mec2_fw) {
 		ring->me = 2;
 		ring->pipe = 0;
+		adev->gfx.kiq.me = 2;
+		adev->gfx.kiq.pipe = 0;
 	} else {
 		ring->me = 1;
 		ring->pipe = 1;
+		adev->gfx.kiq.me = 1;
+		adev->gfx.kiq.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 +1413,6 @@ 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
@@ -6927,15 +6929,12 @@ 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;
 
-	BUG_ON(!ring || (ring->funcs->type != AMDGPU_RING_TYPE_KIQ));
-
-	if (ring->me == 1)
+	if (adev->gfx.kiq.me == 1)
 		target = mmCP_ME1_PIPE0_INT_CNTL;
 	else
 		target = mmCP_ME2_PIPE0_INT_CNTL;
-	target += ring->pipe;
+	target += adev->gfx.kiq.pipe;
 
 	switch (type) {
 	case AMDGPU_CP_KIQ_IRQ_DRIVER0:
@@ -6973,7 +6972,7 @@ 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));
 
@@ -7380,4 +7379,4 @@ static void gfx_v8_0_compute_mqd_soft_fini(struct amdgpu_device *adev)
 
 	ring = &adev->gfx.kiq.ring;
 	amdgpu_bo_free_kernel(&ring->mqd_obj, &ring->mqd_gpu_addr, (void **)&ring->mqd_ptr);
-}
\ No newline at end of file
+}
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): 1487242099-16380-1-git-send-email-trigger.huang-5C7GfCeVMHo@public.gmane.org

Reviewed-by Monk Liu
#2 lore raw

________________________________
发件人: Trigger Huang <[email protected]>
发送时间: 2017年2月16日 18:48:19
收件人: [email protected]
抄送: Liu, Monk; Yu, Xiangliang; Huang, Trigger
主题: [PATCH] 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.

---
 drivers/gpu/drm/amd/amdgpu/amdgpu.h   |  2 ++
 drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 17 ++++++++---------
 2 files changed, 10 insertions(+), 9 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu.h +2 -0
index 1db2e7b..be43d09 100644
@@ -793,6 +793,8 @@ struct amdgpu_kiq {
         struct amdgpu_bo        *eop_obj;
         struct amdgpu_ring      ring;
         struct amdgpu_irq_src   irq;
+       u32                     me;
+       u32                     pipe;
 };
 
 /*
drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +8 -9
index 772c42b..04e2a5b 100644
@@ -1390,12 +1390,15 @@ static int gfx_v8_0_kiq_init_ring(struct amdgpu_device *adev,
         if (adev->gfx.mec2_fw) {
                 ring->me = 2;
                 ring->pipe = 0;
+               adev->gfx.kiq.me = 2;
+               adev->gfx.kiq.pipe = 0;
         } else {
                 ring->me = 1;
                 ring->pipe = 1;
+               adev->gfx.kiq.me = 1;
+               adev->gfx.kiq.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 +1413,6 @@ 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
@@ -6927,15 +6929,12 @@ 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;
 
-       BUG_ON(!ring || (ring->funcs->type != AMDGPU_RING_TYPE_KIQ));
-
-       if (ring->me == 1)
+       if (adev->gfx.kiq.me == 1)
                 target = mmCP_ME1_PIPE0_INT_CNTL;
         else
                 target = mmCP_ME2_PIPE0_INT_CNTL;
-       target += ring->pipe;
+       target += adev->gfx.kiq.pipe;
 
         switch (type) {
         case AMDGPU_CP_KIQ_IRQ_DRIVER0:
@@ -6973,7 +6972,7 @@ 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));
 
@@ -7380,4 +7379,4 @@ static void gfx_v8_0_compute_mqd_soft_fini(struct amdgpu_device *adev)
 
         ring = &adev->gfx.kiq.ring;
         amdgpu_bo_free_kernel(&ring->mqd_obj, &ring->mqd_gpu_addr, (void **)&ring->mqd_ptr);
-}
\ No newline at end of file
+}

--
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

Hi Trigger,

Can you try to move src->data= NULL into gfx_v8_0_kiq_set_interrupt_state function? I think it is more simple and clear.

Thanks!
Xiangliang Yu

Quoted 100 lines

-----Original Message-----
From: Trigger Huang [mailto:[email protected]]
Sent: Thursday, February 16, 2017 6:48 PM
To: [email protected]
Cc: Liu, Monk <[email protected]>; Yu, Xiangliang <[email protected]>; Huang, Trigger <[email protected]>
Subject: [PATCH] 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.

---
 drivers/gpu/drm/amd/amdgpu/amdgpu.h   |  2 ++
 drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 17 ++++++++---------
 2 files changed, 10 insertions(+), 9 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu.h +2 -0
index 1db2e7b..be43d09 100644
@@ -793,6 +793,8 @@ struct amdgpu_kiq {
 	struct amdgpu_bo	*eop_obj;
 	struct amdgpu_ring	ring;
 	struct amdgpu_irq_src	irq;
+	u32			me;
+	u32			pipe;
 };
 
 /*
drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +4 -1
index 772c42b..04e2a5b 100644
@@ -1390,12 +1390,15 @@ static int gfx_v8_0_kiq_init_ring(struct
 amdgpu_device *adev,
 	if (adev->gfx.mec2_fw) {
 		ring->me = 2;
 		ring->pipe = 0;
+		adev->gfx.kiq.me = 2;
+		adev->gfx.kiq.pipe = 0;
 	} else {
 		ring->me = 1;
 		ring->pipe = 1;
+		adev->gfx.kiq.me = 1;
+		adev->gfx.kiq.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 +1413,6 @@ static
void gfx_v8_0_kiq_free_ring(struct amdgpu_ring *ring, {

drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +2 -6
 	amdgpu_wb_free(ring->adev, ring->adev->virt.reg_val_offs);
 	amdgpu_ring_fini(ring);
-	irq->data = NULL;
 }
 
 #define MEC_HPD_SIZE 2048
@@ -6927,15 +6929,12 @@ 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;
 
-	BUG_ON(!ring || (ring->funcs->type != AMDGPU_RING_TYPE_KIQ));
-
-	if (ring->me == 1)
+	if (adev->gfx.kiq.me == 1)
 		target = mmCP_ME1_PIPE0_INT_CNTL;
 	else
 		target = mmCP_ME2_PIPE0_INT_CNTL;
-	target += ring->pipe;
+	target += adev->gfx.kiq.pipe;
 
 	switch (type) {
drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +1 -1
 	case AMDGPU_CP_KIQ_IRQ_DRIVER0:
@@ -6973,7 +6972,7 @@ 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));
@@ -7380,4 +7379,4 @@ static void
 gfx_v8_0_compute_mqd_soft_fini(struct amdgpu_device *adev)
 
 	ring = &adev->gfx.kiq.ring;
 	amdgpu_bo_free_kernel(&ring->mqd_obj, &ring->mqd_gpu_addr,
Quoted 4 lines

(void **)&ring->mqd_ptr); -} \ No newline at end of file +}
--
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): DM5PR12MB1705C5254EB2867EF39417B8EB5D0-2J9CzHegvk9755J/[email protected]

#5 lore raw

Hi Xiangliang,

Thanks for your suggestions, I will make a new patch according to the discussion

Thanks & Best Wishes,
Trigger Huang

-----Original Message-----
From: Yu, Xiangliang
Sent: Friday, February 17, 2017 2:15 PM
To: Huang, Trigger <[email protected]>; [email protected]
Cc: Liu, Monk <[email protected]>; Huang, Trigger <[email protected]>
Subject: RE: [PATCH] drm/amdgpu: Fix module unload hang by KIQ IRQ set

Hi Trigger,

Can you try to move src->data= NULL into gfx_v8_0_kiq_set_interrupt_state function? I think it is more simple and clear.

Thanks!
Xiangliang Yu

Quoted 99 lines

-----Original Message-----
From: Trigger Huang [mailto:[email protected]]
Sent: Thursday, February 16, 2017 6:48 PM
To: [email protected]
Cc: Liu, Monk <[email protected]>; Yu, Xiangliang <[email protected]>; Huang, Trigger <[email protected]>
Subject: [PATCH] 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.

---
 drivers/gpu/drm/amd/amdgpu/amdgpu.h   |  2 ++
 drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 17 ++++++++---------
 2 files changed, 10 insertions(+), 9 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu.h +2 -0
index 1db2e7b..be43d09 100644
@@ -793,6 +793,8 @@ struct amdgpu_kiq {
 	struct amdgpu_bo	*eop_obj;
 	struct amdgpu_ring	ring;
 	struct amdgpu_irq_src	irq;
+	u32			me;
+	u32			pipe;
 };
 
 /*
drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +4 -1
index 772c42b..04e2a5b 100644
@@ -1390,12 +1390,15 @@ static int gfx_v8_0_kiq_init_ring(struct 
 amdgpu_device *adev,
 	if (adev->gfx.mec2_fw) {
 		ring->me = 2;
 		ring->pipe = 0;
+		adev->gfx.kiq.me = 2;
+		adev->gfx.kiq.pipe = 0;
 	} else {
 		ring->me = 1;
 		ring->pipe = 1;
+		adev->gfx.kiq.me = 1;
+		adev->gfx.kiq.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 +1413,6 @@ static
void gfx_v8_0_kiq_free_ring(struct amdgpu_ring *ring, {

drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +2 -6
 	amdgpu_wb_free(ring->adev, ring->adev->virt.reg_val_offs);
 	amdgpu_ring_fini(ring);
-	irq->data = NULL;
 }
 
 #define MEC_HPD_SIZE 2048
@@ -6927,15 +6929,12 @@ 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;
 
-	BUG_ON(!ring || (ring->funcs->type != AMDGPU_RING_TYPE_KIQ));
-
-	if (ring->me == 1)
+	if (adev->gfx.kiq.me == 1)
 		target = mmCP_ME1_PIPE0_INT_CNTL;
 	else
 		target = mmCP_ME2_PIPE0_INT_CNTL;
-	target += ring->pipe;
+	target += adev->gfx.kiq.pipe;
 
 	switch (type) {
drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +1 -1
 	case AMDGPU_CP_KIQ_IRQ_DRIVER0:
@@ -6973,7 +6972,7 @@ 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));
 
@@ -7380,4 +7379,4 @@ static void
 gfx_v8_0_compute_mqd_soft_fini(struct amdgpu_device *adev)
 
 	ring = &adev->gfx.kiq.ring;
 	amdgpu_bo_free_kernel(&ring->mqd_obj, &ring->mqd_gpu_addr, (void 
Quoted 4 lines

**)&ring->mqd_ptr); -} \ No newline at end of file +}
--
2.7.4

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