From 1fa8d710573f02ae9118bc5f53e7ede09d6920da Mon Sep 17 00:00:00 2001
From: Alan Liu <HaoPing.Liu@amd.com>
Date: Fri, 14 Apr 2023 18:39:52 +0800
Subject: drm/amdgpu: Fix desktop freezed after gpu-reset
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
[Why]
After gpu-reset, sometimes the driver fails to enable vblank irq,
causing flip_done timed out and the desktop freezed.
During gpu-reset, we disable and enable vblank irq in dm_suspend() and
dm_resume(). Later on in amdgpu_irq_gpu_reset_resume_helper(), we check
irqs' refcount and decide to enable or disable the irqs again.
However, we have 2 sets of API for controling vblank irq, one is
dm_vblank_get/put() and another is amdgpu_irq_get/put(). Each API has
its own refcount and flag to store the state of vblank irq, and they
are not synchronized.
In drm we use the first API to control vblank irq but in
amdgpu_irq_gpu_reset_resume_helper() we use the second set of API.
The failure happens when vblank irq was enabled by dm_vblank_get()
before gpu-reset, we have vblank->enabled true. However, during
gpu-reset, in amdgpu_irq_gpu_reset_resume_helper() vblank irq's state
checked from amdgpu_irq_update() is DISABLED. So finally it disables
vblank irq again. After gpu-reset, if there is a cursor plane commit,
the driver will try to enable vblank irq by calling drm_vblank_enable(),
but the vblank->enabled is still true, so it fails to turn on vblank
irq and causes flip_done can't be completed in vblank irq handler and
desktop become freezed.
[How]
Combining the 2 vblank control APIs by letting drm's API finally calls
amdgpu_irq's API, so the irq's refcount and state of both APIs can be
synchronized. Also add a check to prevent refcount from being less then
0 in amdgpu_irq_put().
v2:
- Add warning in amdgpu_irq_enable() if the irq is already disabled.
- Call dc_interrupt_set() in dm_set_vblank() to avoid refcount change
if it is in gpu-reset.
v3:
- Improve commit message and code comments.
Signed-off-by: Alan Liu <HaoPing.Liu@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c | 3 +++
1 file changed, 3 insertions(+)
(limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c')
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c
index d0a1cc88832c..fafebec5b7b6 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c
@@ -596,6 +596,9 @@ int amdgpu_irq_put(struct amdgpu_device *adev, struct amdgpu_irq_src *src,
if (!src->enabled_types || !src->funcs->set)
return -EINVAL;
+ if (WARN_ON(!amdgpu_irq_enabled(adev, src, type)))
+ return -EINVAL;
+
if (atomic_dec_and_test(&src->enabled_types[type]))
return amdgpu_irq_update(adev, src, type);
--
cgit v1.2.3
From 15091a6f4380a0c1a7202d52e82cdaaf80e2bb70 Mon Sep 17 00:00:00 2001
From: Le Ma <le.ma@amd.com>
Date: Fri, 19 Nov 2021 18:03:34 +0800
Subject: drm/amdgpu: add node_id to physical id conversion in EOP handler
A new field nodeid in interrupt cookie indicates the node ID.
Signed-off-by: Le Ma <le.ma@amd.com>
Reviewed-by: Shiwu Zhang <shiwu.zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c | 11 +++++++++++
drivers/gpu/drm/amd/amdgpu/amdgpu_irq.h | 14 ++++++++++++++
drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c | 6 ++++--
3 files changed, 29 insertions(+), 2 deletions(-)
(limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c')
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c
index fafebec5b7b6..031610c1340a 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c
@@ -99,6 +99,17 @@ const char *soc15_ih_clientid_name[] = {
"MP1"
};
+const int node_id_to_phys_map[NODEID_MAX] = {
+ [XCD0_NODEID] = 0,
+ [XCD1_NODEID] = 1,
+ [XCD2_NODEID] = 2,
+ [XCD3_NODEID] = 3,
+ [XCD4_NODEID] = 4,
+ [XCD5_NODEID] = 5,
+ [XCD6_NODEID] = 6,
+ [XCD7_NODEID] = 7,
+};
+
/**
* amdgpu_irq_disable_all - disable *all* interrupts
*
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.h
index 1c747ac4129a..efe8a278cbdf 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.h
@@ -102,6 +102,20 @@ struct amdgpu_irq {
bool retry_cam_enabled;
};
+enum interrupt_node_id_per_xcp {
+ XCD0_NODEID = 1,
+ XCD1_NODEID = 2,
+ XCD2_NODEID = 5,
+ XCD3_NODEID = 6,
+ XCD4_NODEID = 9,
+ XCD5_NODEID = 10,
+ XCD6_NODEID = 13,
+ XCD7_NODEID = 14,
+ NODEID_MAX,
+};
+
+extern const int node_id_to_phys_map[NODEID_MAX];
+
void amdgpu_irq_disable_all(struct amdgpu_device *adev);
int amdgpu_irq_init(struct amdgpu_device *adev);
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
index ec273a217666..ce64c4fc5f1a 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
@@ -2799,7 +2799,7 @@ static int gfx_v9_4_3_eop_irq(struct amdgpu_device *adev,
struct amdgpu_irq_src *source,
struct amdgpu_iv_entry *entry)
{
- int i;
+ int i, phys_id;
u8 me_id, pipe_id, queue_id;
struct amdgpu_ring *ring;
@@ -2808,12 +2808,14 @@ static int gfx_v9_4_3_eop_irq(struct amdgpu_device *adev,
pipe_id = (entry->ring_id & 0x03) >> 0;
queue_id = (entry->ring_id & 0x70) >> 4;
+ phys_id = node_id_to_phys_map[entry->node_id];
+
switch (me_id) {
case 0:
case 1:
case 2:
for (i = 0; i < adev->gfx.num_compute_rings; i++) {
- ring = &adev->gfx.compute_ring[i];
+ ring = &adev->gfx.compute_ring[i + phys_id * adev->gfx.num_compute_rings];
/* Per-queue interrupt is supported for MEC starting from VI.
* The interrupt can only be enabled/disabled per pipe instead of per queue.
*/
--
cgit v1.2.3
From 0d81101c190d1835f1bfba85dc3d65b9f9cacd68 Mon Sep 17 00:00:00 2001
From: Le Ma <le.ma@amd.com>
Date: Fri, 25 Feb 2022 15:14:19 +0800
Subject: drm/amdgpu: complement the IH node_id table for multiple AIDs
With different node_id, the SDMA interrupt from multiple AIDs can be
distinguished by sw driver.
Signed-off-by: Le Ma <le.ma@amd.com>
Acked-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c | 4 ++++
drivers/gpu/drm/amd/amdgpu/amdgpu_irq.h | 6 +++++-
drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c | 3 +++
3 files changed, 12 insertions(+), 1 deletion(-)
(limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c')
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c
index 031610c1340a..3dafaf70c987 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c
@@ -100,12 +100,16 @@ const char *soc15_ih_clientid_name[] = {
};
const int node_id_to_phys_map[NODEID_MAX] = {
+ [AID0_NODEID] = 0,
[XCD0_NODEID] = 0,
[XCD1_NODEID] = 1,
+ [AID1_NODEID] = 1,
[XCD2_NODEID] = 2,
[XCD3_NODEID] = 3,
+ [AID2_NODEID] = 2,
[XCD4_NODEID] = 4,
[XCD5_NODEID] = 5,
+ [AID3_NODEID] = 3,
[XCD6_NODEID] = 6,
[XCD7_NODEID] = 7,
};
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.h
index efe8a278cbdf..04c0b4fa17a4 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.h
@@ -102,13 +102,17 @@ struct amdgpu_irq {
bool retry_cam_enabled;
};
-enum interrupt_node_id_per_xcp {
+enum interrupt_node_id_per_aid {
+ AID0_NODEID = 0,
XCD0_NODEID = 1,
XCD1_NODEID = 2,
+ AID1_NODEID = 4,
XCD2_NODEID = 5,
XCD3_NODEID = 6,
+ AID2_NODEID = 8,
XCD4_NODEID = 9,
XCD5_NODEID = 10,
+ AID3_NODEID = 12,
XCD6_NODEID = 13,
XCD7_NODEID = 14,
NODEID_MAX,
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
index 6935a24d1e89..d3c7f9a43ef1 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
@@ -1492,6 +1492,9 @@ static int sdma_v4_4_2_process_trap_irq(struct amdgpu_device *adev,
DRM_DEBUG("IH: SDMA trap\n");
instance = sdma_v4_4_2_irq_id_to_seq(entry->client_id);
+ instance += node_id_to_phys_map[entry->node_id] *
+ adev->sdma.num_inst_per_aid;
+
switch (entry->ring_id) {
case 0:
amdgpu_fence_process(&adev->sdma.instance[instance].ring);
--
cgit v1.2.3
From 413521a4c9f4bbb4637b9ff3427070325890b08f Mon Sep 17 00:00:00 2001
From: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com>
Date: Wed, 24 May 2023 01:17:48 +0530
Subject: drm/amd/amdgpu: Fix warnings in amdgpu_irq.c
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
checkpatch reports:
WARNING: Prefer 'unsigned int' to bare use of 'unsigned'
WARNING: braces {} are not necessary for any arm of this statement
+ if (nvec <= 0) {
[...]
+ } else {
[...]
WARNING: Block comments use a trailing */ on a separate line
Cc: Christian König <christian.koenig@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c | 27 ++++++++++++++-------------
1 file changed, 14 insertions(+), 13 deletions(-)
(limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c')
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c
index 3dafaf70c987..5667053c295c 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c
@@ -124,7 +124,7 @@ const int node_id_to_phys_map[NODEID_MAX] = {
void amdgpu_irq_disable_all(struct amdgpu_device *adev)
{
unsigned long irqflags;
- unsigned i, j, k;
+ unsigned int i, j, k;
int r;
spin_lock_irqsave(&adev->irq.lock, irqflags);
@@ -283,11 +283,11 @@ int amdgpu_irq_init(struct amdgpu_device *adev)
int nvec = pci_msix_vec_count(adev->pdev);
unsigned int flags;
- if (nvec <= 0) {
+ if (nvec <= 0)
flags = PCI_IRQ_MSI;
- } else {
+ else
flags = PCI_IRQ_MSI | PCI_IRQ_MSIX;
- }
+
/* we only need one vector */
nvec = pci_alloc_irq_vectors(adev->pdev, 1, 1, flags);
if (nvec > 0) {
@@ -346,7 +346,7 @@ void amdgpu_irq_fini_hw(struct amdgpu_device *adev)
*/
void amdgpu_irq_fini_sw(struct amdgpu_device *adev)
{
- unsigned i, j;
+ unsigned int i, j;
for (i = 0; i < AMDGPU_IRQ_CLIENTID_MAX; ++i) {
if (!adev->irq.client[i].sources)
@@ -380,7 +380,7 @@ void amdgpu_irq_fini_sw(struct amdgpu_device *adev)
* 0 on success or error code otherwise
*/
int amdgpu_irq_add_id(struct amdgpu_device *adev,
- unsigned client_id, unsigned src_id,
+ unsigned int client_id, unsigned int src_id,
struct amdgpu_irq_src *source)
{
if (client_id >= AMDGPU_IRQ_CLIENTID_MAX)
@@ -432,7 +432,7 @@ void amdgpu_irq_dispatch(struct amdgpu_device *adev,
{
u32 ring_index = ih->rptr >> 2;
struct amdgpu_iv_entry entry;
- unsigned client_id, src_id;
+ unsigned int client_id, src_id;
struct amdgpu_irq_src *src;
bool handled = false;
int r;
@@ -507,7 +507,7 @@ void amdgpu_irq_delegate(struct amdgpu_device *adev,
* Updates interrupt state for the specific source (all ASICs).
*/
int amdgpu_irq_update(struct amdgpu_device *adev,
- struct amdgpu_irq_src *src, unsigned type)
+ struct amdgpu_irq_src *src, unsigned int type)
{
unsigned long irqflags;
enum amdgpu_interrupt_state state;
@@ -516,7 +516,8 @@ int amdgpu_irq_update(struct amdgpu_device *adev,
spin_lock_irqsave(&adev->irq.lock, irqflags);
/* We need to determine after taking the lock, otherwise
- we might disable just enabled interrupts again */
+ * we might disable just enabled interrupts again
+ */
if (amdgpu_irq_enabled(adev, src, type))
state = AMDGPU_IRQ_STATE_ENABLE;
else
@@ -570,7 +571,7 @@ void amdgpu_irq_gpu_reset_resume_helper(struct amdgpu_device *adev)
* 0 on success or error code otherwise
*/
int amdgpu_irq_get(struct amdgpu_device *adev, struct amdgpu_irq_src *src,
- unsigned type)
+ unsigned int type)
{
if (!adev->irq.installed)
return -ENOENT;
@@ -600,7 +601,7 @@ int amdgpu_irq_get(struct amdgpu_device *adev, struct amdgpu_irq_src *src,
* 0 on success or error code otherwise
*/
int amdgpu_irq_put(struct amdgpu_device *adev, struct amdgpu_irq_src *src,
- unsigned type)
+ unsigned int type)
{
if (!adev->irq.installed)
return -ENOENT;
@@ -634,7 +635,7 @@ int amdgpu_irq_put(struct amdgpu_device *adev, struct amdgpu_irq_src *src,
* invalid parameters
*/
bool amdgpu_irq_enabled(struct amdgpu_device *adev, struct amdgpu_irq_src *src,
- unsigned type)
+ unsigned int type)
{
if (!adev->irq.installed)
return false;
@@ -747,7 +748,7 @@ void amdgpu_irq_remove_domain(struct amdgpu_device *adev)
* Returns:
* Linux IRQ
*/
-unsigned amdgpu_irq_create_mapping(struct amdgpu_device *adev, unsigned src_id)
+unsigned int amdgpu_irq_create_mapping(struct amdgpu_device *adev, unsigned int src_id)
{
adev->irq.virq[src_id] = irq_create_mapping(adev->irq.domain, src_id);
--
cgit v1.2.3
From 8ffd6f0442674f32c048ec8dffdbc5ec67829beb Mon Sep 17 00:00:00 2001
From: Guchun Chen <guchun.chen@amd.com>
Date: Thu, 25 May 2023 17:24:31 +0800
Subject: drm/amdgpu: keep irq count in amdgpu_irq_disable_all
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
This can clean up all irq warnings because of unbalanced
amdgpu_irq_get/put when unplugging/unbinding device, and leave
irq count decrease in each ip fini function.
Signed-off-by: Guchun Chen <guchun.chen@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c | 1 -
1 file changed, 1 deletion(-)
(limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c')
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c
index 5667053c295c..3481d2808ce5 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c
@@ -139,7 +139,6 @@ void amdgpu_irq_disable_all(struct amdgpu_device *adev)
continue;
for (k = 0; k < src->num_types; ++k) {
- atomic_set(&src->enabled_types[k], 0);
r = src->funcs->set(adev, src, k,
AMDGPU_IRQ_STATE_DISABLE);
if (r)
--
cgit v1.2.3