From 9d7a8bdb909e9b34bced6f56d4f2d320b0021697 Mon Sep 17 00:00:00 2001
From: "Dr. David Alan Gilbert" <linux@treblig.org>
Date: Mon, 23 Sep 2024 02:24:45 +0100
Subject: drm/amdgpu: Remove unused amdgpu_gfx_bit_to_me_queue

amdgpu_gfx_bit_to_me_queue has been unused since it was added in
commit 7470bfcf2014 ("drm/amdgpu: add helper function for gfx queue/bitmap
transition")

Remove it.

Signed-off-by: Dr. David Alan Gilbert <linux@treblig.org>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h | 2 --
 1 file changed, 2 deletions(-)

(limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h')

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h
index 5644e10a86a9..f710178a21bc 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h
@@ -540,8 +540,6 @@ bool amdgpu_gfx_is_high_priority_graphics_queue(struct amdgpu_device *adev,
 						struct amdgpu_ring *ring);
 int amdgpu_gfx_me_queue_to_bit(struct amdgpu_device *adev, int me,
 			       int pipe, int queue);
-void amdgpu_gfx_bit_to_me_queue(struct amdgpu_device *adev, int bit,
-				int *me, int *pipe, int *queue);
 bool amdgpu_gfx_is_me_queue_enabled(struct amdgpu_device *adev, int me,
 				    int pipe, int queue);
 void amdgpu_gfx_off_ctrl(struct amdgpu_device *adev, bool enable);
-- 
cgit 


From efe6a8774375ddcbdd46fb920be55cc2d0120836 Mon Sep 17 00:00:00 2001
From: Alex Deucher <alexander.deucher@amd.com>
Date: Mon, 14 Oct 2024 11:58:34 -0400
Subject: drm/amdgpu: fix fairness in enforce isolation handling

Make sure KFD gets a turn when serializing access to
the GC IP.  Currently non-KFD jobs can starve KFD if they
submit often enough.  This patch prevents that by stalling
non-KFD if its time period has elapsed.

v2: fix units
v3: check enablement properly

Acked-by: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu.h     |  2 +-
 drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c | 53 +++++++++++++++++++++++++++++++--
 drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h |  2 ++
 3 files changed, 54 insertions(+), 3 deletions(-)

(limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h')

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
index 607998d8b1d5..7645e498faa4 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
@@ -118,7 +118,7 @@
 
 #define MAX_GPU_INSTANCE		64
 
-#define GFX_SLICE_PERIOD		msecs_to_jiffies(250)
+#define GFX_SLICE_PERIOD_MS		250
 
 struct amdgpu_gpu_instance {
 	struct amdgpu_device		*adev;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
index e96984c53e72..b8cc4b146bdc 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
@@ -1752,7 +1752,7 @@ static void amdgpu_gfx_kfd_sch_ctrl(struct amdgpu_device *adev, u32 idx,
 		if (adev->gfx.kfd_sch_req_count[idx] == 0 &&
 		    adev->gfx.kfd_sch_inactive[idx]) {
 			schedule_delayed_work(&adev->gfx.enforce_isolation[idx].work,
-					      GFX_SLICE_PERIOD);
+					      msecs_to_jiffies(adev->gfx.enforce_isolation_time[idx]));
 		}
 	} else {
 		if (adev->gfx.kfd_sch_req_count[idx] == 0) {
@@ -1807,8 +1807,9 @@ void amdgpu_gfx_enforce_isolation_handler(struct work_struct *work)
 			fences += amdgpu_fence_count_emitted(&adev->gfx.compute_ring[i]);
 	}
 	if (fences) {
+		/* we've already had our timeslice, so let's wrap this up */
 		schedule_delayed_work(&adev->gfx.enforce_isolation[idx].work,
-				      GFX_SLICE_PERIOD);
+				      msecs_to_jiffies(1));
 	} else {
 		/* Tell KFD to resume the runqueue */
 		if (adev->kfd.init_complete) {
@@ -1821,6 +1822,51 @@ void amdgpu_gfx_enforce_isolation_handler(struct work_struct *work)
 	mutex_unlock(&adev->enforce_isolation_mutex);
 }
 
+static void
+amdgpu_gfx_enforce_isolation_wait_for_kfd(struct amdgpu_device *adev,
+					  u32 idx)
+{
+	unsigned long cjiffies;
+	bool wait = false;
+
+	mutex_lock(&adev->enforce_isolation_mutex);
+	if (adev->enforce_isolation[idx]) {
+		/* set the initial values if nothing is set */
+		if (!adev->gfx.enforce_isolation_jiffies[idx]) {
+			adev->gfx.enforce_isolation_jiffies[idx] = jiffies;
+			adev->gfx.enforce_isolation_time[idx] =	GFX_SLICE_PERIOD_MS;
+		}
+		/* Make sure KFD gets a chance to run */
+		if (amdgpu_amdkfd_compute_active(adev, idx)) {
+			cjiffies = jiffies;
+			if (time_after(cjiffies, adev->gfx.enforce_isolation_jiffies[idx])) {
+				cjiffies -= adev->gfx.enforce_isolation_jiffies[idx];
+				if ((jiffies_to_msecs(cjiffies) >= GFX_SLICE_PERIOD_MS)) {
+					/* if our time is up, let KGD work drain before scheduling more */
+					wait = true;
+					/* reset the timer period */
+					adev->gfx.enforce_isolation_time[idx] =	GFX_SLICE_PERIOD_MS;
+				} else {
+					/* set the timer period to what's left in our time slice */
+					adev->gfx.enforce_isolation_time[idx] =
+						GFX_SLICE_PERIOD_MS - jiffies_to_msecs(cjiffies);
+				}
+			} else {
+				/* if jiffies wrap around we will just wait a little longer */
+				adev->gfx.enforce_isolation_jiffies[idx] = jiffies;
+			}
+		} else {
+			/* if there is no KFD work, then set the full slice period */
+			adev->gfx.enforce_isolation_jiffies[idx] = jiffies;
+			adev->gfx.enforce_isolation_time[idx] = GFX_SLICE_PERIOD_MS;
+		}
+	}
+	mutex_unlock(&adev->enforce_isolation_mutex);
+
+	if (wait)
+		msleep(GFX_SLICE_PERIOD_MS);
+}
+
 void amdgpu_gfx_enforce_isolation_ring_begin_use(struct amdgpu_ring *ring)
 {
 	struct amdgpu_device *adev = ring->adev;
@@ -1837,6 +1883,9 @@ void amdgpu_gfx_enforce_isolation_ring_begin_use(struct amdgpu_ring *ring)
 	if (idx >= MAX_XCP)
 		return;
 
+	/* Don't submit more work until KFD has had some time */
+	amdgpu_gfx_enforce_isolation_wait_for_kfd(adev, idx);
+
 	mutex_lock(&adev->enforce_isolation_mutex);
 	if (adev->enforce_isolation[idx]) {
 		if (adev->kfd.init_complete)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h
index f710178a21bc..af9dbd760fee 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h
@@ -472,6 +472,8 @@ struct amdgpu_gfx {
 	struct mutex                    kfd_sch_mutex;
 	u64				kfd_sch_req_count[MAX_XCP];
 	bool				kfd_sch_inactive[MAX_XCP];
+	unsigned long			enforce_isolation_jiffies[MAX_XCP];
+	unsigned long			enforce_isolation_time[MAX_XCP];
 };
 
 struct amdgpu_gfx_ras_reg_entry {
-- 
cgit 


From c5c63d9cb5d3bbb2fc5973757616b17629795829 Mon Sep 17 00:00:00 2001
From: Jesse Zhang <jesse.zhang@amd.com>
Date: Tue, 29 Oct 2024 10:11:05 +0800
Subject: drm/amdgpu: add amdgpu_gfx_sched_mask and amdgpu_compute_sched_mask
 debugfs

compute/gfx may have multiple rings on some hardware.
In some cases, userspace wants to run jobs on a specific ring for validation purposes.
This debugfs entry helps to disable or enable submitting jobs to a specific ring.
This entry is populated only if there are at least two or more cores in the gfx/compute ip.

Signed-off-by: Jesse Zhang <jesse.zhang@amd.com>
Suggested-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Tim Huang <tim.huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c |   2 +
 drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c     | 141 ++++++++++++++++++++++++++++
 drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h     |   2 +
 3 files changed, 145 insertions(+)

(limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h')

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c
index 37d8657f0776..6e3f657cab9c 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c
@@ -2096,6 +2096,8 @@ int amdgpu_debugfs_init(struct amdgpu_device *adev)
 		amdgpu_debugfs_umsch_fwlog_init(adev, &adev->umsch_mm);
 
 	amdgpu_debugfs_jpeg_sched_mask_init(adev);
+	amdgpu_debugfs_gfx_sched_mask_init(adev);
+	amdgpu_debugfs_compute_sched_mask_init(adev);
 
 	amdgpu_ras_debugfs_create_all(adev);
 	amdgpu_rap_debugfs_init(adev);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
index b8cc4b146bdc..63f5fa736f32 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
@@ -1917,3 +1917,144 @@ void amdgpu_gfx_enforce_isolation_ring_end_use(struct amdgpu_ring *ring)
 	}
 	mutex_unlock(&adev->enforce_isolation_mutex);
 }
+
+/*
+ * debugfs for to enable/disable gfx job submission to specific core.
+ */
+#if defined(CONFIG_DEBUG_FS)
+static int amdgpu_debugfs_gfx_sched_mask_set(void *data, u64 val)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)data;
+	u32 i;
+	u64 mask = 0;
+	struct amdgpu_ring *ring;
+
+	if (!adev)
+		return -ENODEV;
+
+	mask = (1 << adev->gfx.num_gfx_rings) - 1;
+	if ((val & mask) == 0)
+		return -EINVAL;
+
+	for (i = 0; i < adev->gfx.num_gfx_rings; ++i) {
+		ring = &adev->gfx.gfx_ring[i];
+		if (val & (1 << i))
+			ring->sched.ready = true;
+		else
+			ring->sched.ready = false;
+	}
+	/* publish sched.ready flag update effective immediately across smp */
+	smp_rmb();
+	return 0;
+}
+
+static int amdgpu_debugfs_gfx_sched_mask_get(void *data, u64 *val)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)data;
+	u32 i;
+	u64 mask = 0;
+	struct amdgpu_ring *ring;
+
+	if (!adev)
+		return -ENODEV;
+	for (i = 0; i < adev->gfx.num_gfx_rings; ++i) {
+		ring = &adev->gfx.gfx_ring[i];
+		if (ring->sched.ready)
+			mask |= 1 << i;
+	}
+
+	*val = mask;
+	return 0;
+}
+
+DEFINE_DEBUGFS_ATTRIBUTE(amdgpu_debugfs_gfx_sched_mask_fops,
+			 amdgpu_debugfs_gfx_sched_mask_get,
+			 amdgpu_debugfs_gfx_sched_mask_set, "%llx\n");
+
+#endif
+
+void amdgpu_debugfs_gfx_sched_mask_init(struct amdgpu_device *adev)
+{
+#if defined(CONFIG_DEBUG_FS)
+	struct drm_minor *minor = adev_to_drm(adev)->primary;
+	struct dentry *root = minor->debugfs_root;
+	char name[32];
+
+	if (!(adev->gfx.num_gfx_rings > 1))
+		return;
+	sprintf(name, "amdgpu_gfx_sched_mask");
+	debugfs_create_file(name, 0600, root, adev,
+			    &amdgpu_debugfs_gfx_sched_mask_fops);
+#endif
+}
+
+/*
+ * debugfs for to enable/disable compute job submission to specific core.
+ */
+#if defined(CONFIG_DEBUG_FS)
+static int amdgpu_debugfs_compute_sched_mask_set(void *data, u64 val)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)data;
+	u32 i;
+	u64 mask = 0;
+	struct amdgpu_ring *ring;
+
+	if (!adev)
+		return -ENODEV;
+
+	mask = (1 << adev->gfx.num_compute_rings) - 1;
+	if ((val & mask) == 0)
+		return -EINVAL;
+
+	for (i = 0; i < adev->gfx.num_compute_rings; ++i) {
+		ring = &adev->gfx.compute_ring[i];
+		if (val & (1 << i))
+			ring->sched.ready = true;
+		else
+			ring->sched.ready = false;
+	}
+
+	/* publish sched.ready flag update effective immediately across smp */
+	smp_rmb();
+	return 0;
+}
+
+static int amdgpu_debugfs_compute_sched_mask_get(void *data, u64 *val)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)data;
+	u32 i;
+	u64 mask = 0;
+	struct amdgpu_ring *ring;
+
+	if (!adev)
+		return -ENODEV;
+	for (i = 0; i < adev->gfx.num_compute_rings; ++i) {
+		ring = &adev->gfx.compute_ring[i];
+		if (ring->sched.ready)
+			mask |= 1 << i;
+	}
+
+	*val = mask;
+	return 0;
+}
+
+DEFINE_DEBUGFS_ATTRIBUTE(amdgpu_debugfs_compute_sched_mask_fops,
+			 amdgpu_debugfs_compute_sched_mask_get,
+			 amdgpu_debugfs_compute_sched_mask_set, "%llx\n");
+
+#endif
+
+void amdgpu_debugfs_compute_sched_mask_init(struct amdgpu_device *adev)
+{
+#if defined(CONFIG_DEBUG_FS)
+	struct drm_minor *minor = adev_to_drm(adev)->primary;
+	struct dentry *root = minor->debugfs_root;
+	char name[32];
+
+	if (!(adev->gfx.num_compute_rings > 1))
+		return;
+	sprintf(name, "amdgpu_compute_sched_mask");
+	debugfs_create_file(name, 0600, root, adev,
+			    &amdgpu_debugfs_compute_sched_mask_fops);
+#endif
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h
index af9dbd760fee..7183831df0c3 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h
@@ -584,6 +584,8 @@ void amdgpu_gfx_sysfs_isolation_shader_fini(struct amdgpu_device *adev);
 void amdgpu_gfx_enforce_isolation_handler(struct work_struct *work);
 void amdgpu_gfx_enforce_isolation_ring_begin_use(struct amdgpu_ring *ring);
 void amdgpu_gfx_enforce_isolation_ring_end_use(struct amdgpu_ring *ring);
+void amdgpu_debugfs_gfx_sched_mask_init(struct amdgpu_device *adev);
+void amdgpu_debugfs_compute_sched_mask_init(struct amdgpu_device *adev);
 
 static inline const char *amdgpu_gfx_compute_mode_desc(int mode)
 {
-- 
cgit 


From 047767ddc93666704026c79c01554597375beb50 Mon Sep 17 00:00:00 2001
From: Lijo Lazar <lijo.lazar@amd.com>
Date: Tue, 29 Oct 2024 10:47:26 +0530
Subject: drm/amdgpu: Group gfx sysfs functions

Make amdgpu_gfx_sysfs_init/fini functions as common entry points for all
gfx related sysfs nodes.

Signed-off-by: Lijo Lazar <lijo.lazar@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_gfx.c | 40 +++++++++++++++++++++++++--------
 drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h |  2 --
 drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c  |  5 +++--
 drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c  |  4 ++--
 drivers/gpu/drm/amd/amdgpu/gfx_v12_0.c  |  4 ++--
 drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c   |  4 ++--
 drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c |  5 -----
 7 files changed, 40 insertions(+), 24 deletions(-)

(limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h')

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
index 63f5fa736f32..2f3f09dfb1fd 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
@@ -1602,7 +1602,7 @@ static DEVICE_ATTR(current_compute_partition, 0644,
 static DEVICE_ATTR(available_compute_partition, 0444,
 		   amdgpu_gfx_get_available_compute_partition, NULL);
 
-int amdgpu_gfx_sysfs_init(struct amdgpu_device *adev)
+static int amdgpu_gfx_sysfs_xcp_init(struct amdgpu_device *adev)
 {
 	struct amdgpu_xcp_mgr *xcp_mgr = adev->xcp_mgr;
 	bool xcp_switch_supported;
@@ -1629,7 +1629,7 @@ int amdgpu_gfx_sysfs_init(struct amdgpu_device *adev)
 	return r;
 }
 
-void amdgpu_gfx_sysfs_fini(struct amdgpu_device *adev)
+static void amdgpu_gfx_sysfs_xcp_fini(struct amdgpu_device *adev)
 {
 	struct amdgpu_xcp_mgr *xcp_mgr = adev->xcp_mgr;
 	bool xcp_switch_supported;
@@ -1646,25 +1646,47 @@ void amdgpu_gfx_sysfs_fini(struct amdgpu_device *adev)
 				   &dev_attr_available_compute_partition);
 }
 
-int amdgpu_gfx_sysfs_isolation_shader_init(struct amdgpu_device *adev)
+static int amdgpu_gfx_sysfs_isolation_shader_init(struct amdgpu_device *adev)
 {
 	int r;
 
 	r = device_create_file(adev->dev, &dev_attr_enforce_isolation);
 	if (r)
 		return r;
+	if (adev->gfx.enable_cleaner_shader)
+		r = device_create_file(adev->dev, &dev_attr_run_cleaner_shader);
 
-	r = device_create_file(adev->dev, &dev_attr_run_cleaner_shader);
-	if (r)
+	return r;
+}
+
+static void amdgpu_gfx_sysfs_isolation_shader_fini(struct amdgpu_device *adev)
+{
+	device_remove_file(adev->dev, &dev_attr_enforce_isolation);
+	if (adev->gfx.enable_cleaner_shader)
+		device_remove_file(adev->dev, &dev_attr_run_cleaner_shader);
+}
+
+int amdgpu_gfx_sysfs_init(struct amdgpu_device *adev)
+{
+	int r;
+
+	r = amdgpu_gfx_sysfs_xcp_init(adev);
+	if (r) {
+		dev_err(adev->dev, "failed to create xcp sysfs files");
 		return r;
+	}
 
-	return 0;
+	r = amdgpu_gfx_sysfs_isolation_shader_init(adev);
+	if (r)
+		dev_err(adev->dev, "failed to create isolation sysfs files");
+
+	return r;
 }
 
-void amdgpu_gfx_sysfs_isolation_shader_fini(struct amdgpu_device *adev)
+void amdgpu_gfx_sysfs_fini(struct amdgpu_device *adev)
 {
-	device_remove_file(adev->dev, &dev_attr_enforce_isolation);
-	device_remove_file(adev->dev, &dev_attr_run_cleaner_shader);
+	amdgpu_gfx_sysfs_xcp_fini(adev);
+	amdgpu_gfx_sysfs_isolation_shader_fini(adev);
 }
 
 int amdgpu_gfx_cleaner_shader_sw_init(struct amdgpu_device *adev,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h
index 7183831df0c3..fd73e527f446 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h
@@ -579,8 +579,6 @@ void amdgpu_gfx_cleaner_shader_sw_fini(struct amdgpu_device *adev);
 void amdgpu_gfx_cleaner_shader_init(struct amdgpu_device *adev,
 				    unsigned int cleaner_shader_size,
 				    const void *cleaner_shader_ptr);
-int amdgpu_gfx_sysfs_isolation_shader_init(struct amdgpu_device *adev);
-void amdgpu_gfx_sysfs_isolation_shader_fini(struct amdgpu_device *adev);
 void amdgpu_gfx_enforce_isolation_handler(struct work_struct *work);
 void amdgpu_gfx_enforce_isolation_ring_begin_use(struct amdgpu_ring *ring);
 void amdgpu_gfx_enforce_isolation_ring_end_use(struct amdgpu_ring *ring);
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c
index 9da95b25e158..d1a18ca584dd 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c
@@ -4853,9 +4853,10 @@ static int gfx_v10_0_sw_init(struct amdgpu_ip_block *ip_block)
 
 	gfx_v10_0_alloc_ip_dump(adev);
 
-	r = amdgpu_gfx_sysfs_isolation_shader_init(adev);
+	r = amdgpu_gfx_sysfs_init(adev);
 	if (r)
 		return r;
+
 	return 0;
 }
 
@@ -4907,7 +4908,7 @@ static int gfx_v10_0_sw_fini(struct amdgpu_ip_block *ip_block)
 		gfx_v10_0_rlc_backdoor_autoload_buffer_fini(adev);
 
 	gfx_v10_0_free_microcode(adev);
-	amdgpu_gfx_sysfs_isolation_shader_fini(adev);
+	amdgpu_gfx_sysfs_fini(adev);
 
 	kfree(adev->gfx.ip_dump_core);
 	kfree(adev->gfx.ip_dump_compute_queues);
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
index 894fc04201c3..cfe17a36b8ee 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
@@ -1708,7 +1708,7 @@ static int gfx_v11_0_sw_init(struct amdgpu_ip_block *ip_block)
 
 	gfx_v11_0_alloc_ip_dump(adev);
 
-	r = amdgpu_gfx_sysfs_isolation_shader_init(adev);
+	r = amdgpu_gfx_sysfs_init(adev);
 	if (r)
 		return r;
 
@@ -1773,7 +1773,7 @@ static int gfx_v11_0_sw_fini(struct amdgpu_ip_block *ip_block)
 
 	gfx_v11_0_free_microcode(adev);
 
-	amdgpu_gfx_sysfs_isolation_shader_fini(adev);
+	amdgpu_gfx_sysfs_fini(adev);
 
 	kfree(adev->gfx.ip_dump_core);
 	kfree(adev->gfx.ip_dump_compute_queues);
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v12_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v12_0.c
index 9fec28d8a5fc..1b99f90cd193 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v12_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v12_0.c
@@ -1466,7 +1466,7 @@ static int gfx_v12_0_sw_init(struct amdgpu_ip_block *ip_block)
 
 	gfx_v12_0_alloc_ip_dump(adev);
 
-	r = amdgpu_gfx_sysfs_isolation_shader_init(adev);
+	r = amdgpu_gfx_sysfs_init(adev);
 	if (r)
 		return r;
 
@@ -1529,7 +1529,7 @@ static int gfx_v12_0_sw_fini(struct amdgpu_ip_block *ip_block)
 
 	gfx_v12_0_free_microcode(adev);
 
-	amdgpu_gfx_sysfs_isolation_shader_fini(adev);
+	amdgpu_gfx_sysfs_fini(adev);
 
 	kfree(adev->gfx.ip_dump_core);
 	kfree(adev->gfx.ip_dump_compute_queues);
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
index e9248a855ba7..a880dce16ae2 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
@@ -2402,7 +2402,7 @@ static int gfx_v9_0_sw_init(struct amdgpu_ip_block *ip_block)
 
 	gfx_v9_0_alloc_ip_dump(adev);
 
-	r = amdgpu_gfx_sysfs_isolation_shader_init(adev);
+	r = amdgpu_gfx_sysfs_init(adev);
 	if (r)
 		return r;
 
@@ -2443,7 +2443,7 @@ static int gfx_v9_0_sw_fini(struct amdgpu_ip_block *ip_block)
 	}
 	gfx_v9_0_free_microcode(adev);
 
-	amdgpu_gfx_sysfs_isolation_shader_fini(adev);
+	amdgpu_gfx_sysfs_fini(adev);
 
 	kfree(adev->gfx.ip_dump_core);
 	kfree(adev->gfx.ip_dump_compute_queues);
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 016290f00592..983088805c3a 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
@@ -1171,10 +1171,6 @@ static int gfx_v9_4_3_sw_init(struct amdgpu_ip_block *ip_block)
 
 	gfx_v9_4_3_alloc_ip_dump(adev);
 
-	r = amdgpu_gfx_sysfs_isolation_shader_init(adev);
-	if (r)
-		return r;
-
 	return 0;
 }
 
@@ -1199,7 +1195,6 @@ static int gfx_v9_4_3_sw_fini(struct amdgpu_ip_block *ip_block)
 	amdgpu_bo_unref(&adev->gfx.rlc.clear_state_obj);
 	gfx_v9_4_3_free_microcode(adev);
 	amdgpu_gfx_sysfs_fini(adev);
-	amdgpu_gfx_sysfs_isolation_shader_fini(adev);
 
 	kfree(adev->gfx.ip_dump_core);
 	kfree(adev->gfx.ip_dump_compute_queues);
-- 
cgit 


From 6c8d1f4b042e706ccd7575beb0397a75d545d71b Mon Sep 17 00:00:00 2001
From: "Jesse.zhang@amd.com" <Jesse.zhang@amd.com>
Date: Tue, 5 Nov 2024 15:22:56 +0800
Subject: drm/amdgpu: Add sysfs interface for gc reset mask

Add two sysfs interfaces for gfx and compute:
gfx_reset_mask
compute_reset_mask

These interfaces are read-only and show the resets supported by the IP.
For example, full adapter reset (mode1/mode2/BACO/etc),
soft reset, queue reset, and pipe reset.

V2: the sysfs node returns a text string instead of some flags (Christian)
v3: add a generic helper which takes the ring as parameter
    and print the strings in the order they are applied (Christian)

    check amdgpu_gpu_recovery  before creating sysfs file itself,
    and initialize supported_reset_types in IP version files (Lijo)
v4: Fixing uninitialized variables (Tim)

Signed-off-by: Jesse Zhang <Jesse.Zhang@amd.com>
Suggested-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Tim Huang <tim.huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu.h        |  8 ++++
 drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 44 +++++++++++++++++++
 drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c    | 70 ++++++++++++++++++++++++++++++
 drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h    |  2 +
 drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c     |  5 +++
 drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c     | 18 ++++++++
 drivers/gpu/drm/amd/amdgpu/gfx_v12_0.c     |  6 +++
 drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c      |  6 +++
 drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c    | 13 ++++++
 9 files changed, 172 insertions(+)

(limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h')

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
index 7645e498faa4..d8bc6da50016 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
@@ -299,6 +299,12 @@ extern int amdgpu_wbrf;
 #define AMDGPU_RESET_VCE			(1 << 13)
 #define AMDGPU_RESET_VCE1			(1 << 14)
 
+/* reset mask */
+#define AMDGPU_RESET_TYPE_FULL (1 << 0) /* full adapter reset, mode1/mode2/BACO/etc. */
+#define AMDGPU_RESET_TYPE_SOFT_RESET (1 << 1) /* IP level soft reset */
+#define AMDGPU_RESET_TYPE_PER_QUEUE (1 << 2) /* per queue */
+#define AMDGPU_RESET_TYPE_PER_PIPE (1 << 3) /* per pipe */
+
 /* max cursor sizes (in pixels) */
 #define CIK_CURSOR_WIDTH 128
 #define CIK_CURSOR_HEIGHT 128
@@ -1464,6 +1470,8 @@ struct dma_fence *amdgpu_device_get_gang(struct amdgpu_device *adev);
 struct dma_fence *amdgpu_device_switch_gang(struct amdgpu_device *adev,
 					    struct dma_fence *gang);
 bool amdgpu_device_has_display_hardware(struct amdgpu_device *adev);
+ssize_t amdgpu_get_soft_full_reset_mask(struct amdgpu_ring *ring);
+ssize_t amdgpu_show_reset_mask(char *buf, uint32_t supported_reset);
 
 /* atpx handler */
 #if defined(CONFIG_VGA_SWITCHEROO)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
index 0450eab6ade7..38a17571da75 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
@@ -6715,3 +6715,47 @@ uint32_t amdgpu_device_wait_on_rreg(struct amdgpu_device *adev,
 	}
 	return ret;
 }
+
+ssize_t amdgpu_get_soft_full_reset_mask(struct amdgpu_ring *ring)
+{
+	ssize_t size = 0;
+
+	if (!ring || !ring->adev)
+		return size;
+
+	if (amdgpu_device_should_recover_gpu(ring->adev))
+		size |= AMDGPU_RESET_TYPE_FULL;
+
+	if (unlikely(!ring->adev->debug_disable_soft_recovery) &&
+	    !amdgpu_sriov_vf(ring->adev) && ring->funcs->soft_recovery)
+		size |= AMDGPU_RESET_TYPE_SOFT_RESET;
+
+	return size;
+}
+
+ssize_t amdgpu_show_reset_mask(char *buf, uint32_t supported_reset)
+{
+	ssize_t size = 0;
+
+	if (supported_reset == 0) {
+		size += sysfs_emit_at(buf, size, "unsupported");
+		size += sysfs_emit_at(buf, size, "\n");
+		return size;
+
+	}
+
+	if (supported_reset & AMDGPU_RESET_TYPE_SOFT_RESET)
+		size += sysfs_emit_at(buf, size, "soft ");
+
+	if (supported_reset & AMDGPU_RESET_TYPE_PER_QUEUE)
+		size += sysfs_emit_at(buf, size, "queue ");
+
+	if (supported_reset & AMDGPU_RESET_TYPE_PER_PIPE)
+		size += sysfs_emit_at(buf, size, "pipe ");
+
+	if (supported_reset & AMDGPU_RESET_TYPE_FULL)
+		size += sysfs_emit_at(buf, size, "full ");
+
+	size += sysfs_emit_at(buf, size, "\n");
+	return size;
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
index 2f3f09dfb1fd..6cc6484bde06 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
@@ -1588,6 +1588,32 @@ static ssize_t amdgpu_gfx_set_enforce_isolation(struct device *dev,
 	return count;
 }
 
+static ssize_t amdgpu_gfx_get_gfx_reset_mask(struct device *dev,
+						struct device_attribute *attr,
+						char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+
+	if (!adev)
+		return -ENODEV;
+
+	return amdgpu_show_reset_mask(buf, adev->gfx.gfx_supported_reset);
+}
+
+static ssize_t amdgpu_gfx_get_compute_reset_mask(struct device *dev,
+						struct device_attribute *attr,
+						char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+
+	if (!adev)
+		return -ENODEV;
+
+	return amdgpu_show_reset_mask(buf, adev->gfx.compute_supported_reset);
+}
+
 static DEVICE_ATTR(run_cleaner_shader, 0200,
 		   NULL, amdgpu_gfx_set_run_cleaner_shader);
 
@@ -1601,6 +1627,11 @@ static DEVICE_ATTR(current_compute_partition, 0644,
 
 static DEVICE_ATTR(available_compute_partition, 0444,
 		   amdgpu_gfx_get_available_compute_partition, NULL);
+static DEVICE_ATTR(gfx_reset_mask, 0444,
+		   amdgpu_gfx_get_gfx_reset_mask, NULL);
+
+static DEVICE_ATTR(compute_reset_mask, 0444,
+		   amdgpu_gfx_get_compute_reset_mask, NULL);
 
 static int amdgpu_gfx_sysfs_xcp_init(struct amdgpu_device *adev)
 {
@@ -1666,6 +1697,40 @@ static void amdgpu_gfx_sysfs_isolation_shader_fini(struct amdgpu_device *adev)
 		device_remove_file(adev->dev, &dev_attr_run_cleaner_shader);
 }
 
+static int amdgpu_gfx_sysfs_reset_mask_init(struct amdgpu_device *adev)
+{
+	int r = 0;
+
+	if (!amdgpu_gpu_recovery)
+		return r;
+
+	if (adev->gfx.num_gfx_rings) {
+		r = device_create_file(adev->dev, &dev_attr_gfx_reset_mask);
+		if (r)
+			return r;
+	}
+
+	if (adev->gfx.num_compute_rings) {
+		r = device_create_file(adev->dev, &dev_attr_compute_reset_mask);
+		if (r)
+			return r;
+	}
+
+	return r;
+}
+
+static void amdgpu_gfx_sysfs_reset_mask_fini(struct amdgpu_device *adev)
+{
+	if (!amdgpu_gpu_recovery)
+		return;
+
+	if (adev->gfx.num_gfx_rings)
+		device_remove_file(adev->dev, &dev_attr_gfx_reset_mask);
+
+	if (adev->gfx.num_compute_rings)
+		device_remove_file(adev->dev, &dev_attr_compute_reset_mask);
+}
+
 int amdgpu_gfx_sysfs_init(struct amdgpu_device *adev)
 {
 	int r;
@@ -1680,6 +1745,10 @@ int amdgpu_gfx_sysfs_init(struct amdgpu_device *adev)
 	if (r)
 		dev_err(adev->dev, "failed to create isolation sysfs files");
 
+	r = amdgpu_gfx_sysfs_reset_mask_init(adev);
+	if (r)
+		dev_err(adev->dev, "failed to create reset mask sysfs files");
+
 	return r;
 }
 
@@ -1687,6 +1756,7 @@ void amdgpu_gfx_sysfs_fini(struct amdgpu_device *adev)
 {
 	amdgpu_gfx_sysfs_xcp_fini(adev);
 	amdgpu_gfx_sysfs_isolation_shader_fini(adev);
+	amdgpu_gfx_sysfs_reset_mask_fini(adev);
 }
 
 int amdgpu_gfx_cleaner_shader_sw_init(struct amdgpu_device *adev,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h
index fd73e527f446..8b5bd63b5773 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h
@@ -424,6 +424,8 @@ struct amdgpu_gfx {
 	/* reset mask */
 	uint32_t                        grbm_soft_reset;
 	uint32_t                        srbm_soft_reset;
+	uint32_t 			gfx_supported_reset;
+	uint32_t 			compute_supported_reset;
 
 	/* gfx off */
 	bool                            gfx_off_state;      /* true: enabled, false: disabled */
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c
index d1a18ca584dd..24dce803a829 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c
@@ -4825,6 +4825,11 @@ static int gfx_v10_0_sw_init(struct amdgpu_ip_block *ip_block)
 			}
 		}
 	}
+	/* TODO: Add queue reset mask when FW fully supports it */
+	adev->gfx.gfx_supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->gfx.gfx_ring[0]);
+	adev->gfx.compute_supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->gfx.compute_ring[0]);
 
 	r = amdgpu_gfx_kiq_init(adev, GFX10_MEC_HPD_SIZE, 0);
 	if (r) {
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
index 62e4c446793d..28fe8d23c91f 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
@@ -1691,6 +1691,24 @@ static int gfx_v11_0_sw_init(struct amdgpu_ip_block *ip_block)
 		}
 	}
 
+	adev->gfx.gfx_supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->gfx.gfx_ring[0]);
+	adev->gfx.compute_supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->gfx.compute_ring[0]);
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(11, 0, 0):
+	case IP_VERSION(11, 0, 2):
+	case IP_VERSION(11, 0, 3):
+		if ((adev->gfx.me_fw_version >= 2280) &&
+			    (adev->gfx.mec_fw_version >= 2410)) {
+				adev->gfx.compute_supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE;
+				adev->gfx.gfx_supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE;
+		}
+		break;
+	default:
+		break;
+	}
+
 	if (!adev->enable_mes_kiq) {
 		r = amdgpu_gfx_kiq_init(adev, GFX11_MEC_HPD_SIZE, 0);
 		if (r) {
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v12_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v12_0.c
index 1b99f90cd193..fe7c48f2fb2a 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v12_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v12_0.c
@@ -1437,6 +1437,12 @@ static int gfx_v12_0_sw_init(struct amdgpu_ip_block *ip_block)
 		}
 	}
 
+	/* TODO: Add queue reset mask when FW fully supports it */
+	adev->gfx.gfx_supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->gfx.gfx_ring[0]);
+	adev->gfx.compute_supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->gfx.compute_ring[0]);
+
 	if (!adev->enable_mes_kiq) {
 		r = amdgpu_gfx_kiq_init(adev, GFX12_MEC_HPD_SIZE, 0);
 		if (r) {
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
index a880dce16ae2..0b6f09f2cc9b 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
@@ -2374,6 +2374,12 @@ static int gfx_v9_0_sw_init(struct amdgpu_ip_block *ip_block)
 		}
 	}
 
+	/* TODO: Add queue reset mask when FW fully supports it */
+	adev->gfx.gfx_supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->gfx.gfx_ring[0]);
+	adev->gfx.compute_supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->gfx.compute_ring[0]);
+
 	r = amdgpu_gfx_kiq_init(adev, GFX9_MEC_HPD_SIZE, 0);
 	if (r) {
 		DRM_ERROR("Failed to init KIQ BOs!\n");
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 983088805c3a..e2b3dda57030 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
@@ -1157,6 +1157,19 @@ static int gfx_v9_4_3_sw_init(struct amdgpu_ip_block *ip_block)
 			return r;
 	}
 
+	adev->gfx.compute_supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->gfx.compute_ring[0]);
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(9, 4, 3):
+	case IP_VERSION(9, 4, 4):
+		if (adev->gfx.mec_fw_version >= 155) {
+			adev->gfx.compute_supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE;
+			adev->gfx.compute_supported_reset |= AMDGPU_RESET_TYPE_PER_PIPE;
+		}
+		break;
+	default:
+		break;
+	}
 	r = gfx_v9_4_3_gpu_early_init(adev);
 	if (r)
 		return r;
-- 
cgit