[PATCH v6 0/5] fdinfo memory stats

[PATCH v6 0/5] fdinfo memory stats

[Tvrtko Ursulin]

From: Tvrtko Ursulin <tvrtko.ursulin@intel.com>

A short series to enable fdinfo memory stats for i915.

I added tracking of most classes of objects (user objects, page tables, context
state, ring buffers) which contribute to client's memory footprint and am
accouting their memory use along the similar lines as in Rob's msm code, just
that with i915 specific code we can show a memory region breakdown and so
support discrete and multi-tile GPUs properly. And also reflect that our objects
can have multiple allowed backing stores.

The existing helper Rob added is then used to dump the per memory region stats
to fdinfo.

The basic objects-per-client infrastructure can later be extended to cover all
objects and so avoid needing to walk the IDR under the client's file table lock,
which would further avoid distburbing the running clients by parallel fdinfo
readers.

Example fdinfo format:

# cat /proc/1383/fdinfo/8
pos:    0
flags:  02100002
mnt_id: 21
ino:    397
drm-driver:     i915
drm-client-id:  18
drm-pdev:       0000:00:02.0
drm-total-system:       125 MiB
drm-shared-system:      16 MiB
drm-active-system:      110 MiB
drm-resident-system:    125 MiB
drm-purgeable-system:   2 MiB
drm-total-stolen-system:        0
drm-shared-stolen-system:       0
drm-active-stolen-system:       0
drm-resident-stolen-system:     0
drm-purgeable-stolen-system:    0
drm-engine-render:      25662044495 ns
drm-engine-copy:        0 ns
drm-engine-video:       0 ns
drm-engine-video-enhance:       0 ns

Example gputop output:

DRM minor 0
 PID     SMEM  SMEMRSS   render     copy     video    NAME
1233     124M     124M |████████||        ||        ||        | neverball
1130      59M      59M |█▌      ||        ||        ||        | Xorg
1207      12M      12M |        ||        ||        ||        | xfwm4

Or with Wayland:

DRM minor 0
 PID      MEM      RSS    render        copy         video    video-enhance NAME
2093     191M     191M |▊          ||           ||           ||           | gnome-shell
DRM minor 128
 PID      MEM      RSS    render        copy         video    video-enhance NAME
2551      71M      71M |██▉        ||           ||           ||           | neverball
2553      50M      50M |           ||           ||           ||           | Xwayland

v2:
 * Now actually per client.

v3:
 * Track imported dma-buf objects.

v4:
 * Rely on DRM GEM handles for tracking user objects.
 * Fix internal object accounting (no placements).

v5:
 * Fixed brain fart of overwriting the loop cursor.
 * Fixed object destruction racing with fdinfo reads.
 * Take reference to GEM context while using it.

v6:
 * Rebase, cover letter update.

Tvrtko Ursulin (5):
  drm/i915: Add ability for tracking buffer objects per client
  drm/i915: Record which client owns a VM
  drm/i915: Track page table backing store usage
  drm/i915: Account ring buffer and context state storage
  drm/i915: Implement fdinfo memory stats printing

 drivers/gpu/drm/i915/gem/i915_gem_context.c   |  11 +-
 .../gpu/drm/i915/gem/i915_gem_context_types.h |   3 +
 drivers/gpu/drm/i915/gem/i915_gem_object.c    |  13 +-
 .../gpu/drm/i915/gem/i915_gem_object_types.h  |  12 ++
 .../gpu/drm/i915/gem/selftests/mock_context.c |   4 +-
 drivers/gpu/drm/i915/gt/intel_context.c       |  14 ++
 drivers/gpu/drm/i915/gt/intel_gtt.c           |   6 +
 drivers/gpu/drm/i915/gt/intel_gtt.h           |   1 +
 drivers/gpu/drm/i915/i915_drm_client.c        | 131 ++++++++++++++++++
 drivers/gpu/drm/i915/i915_drm_client.h        |  41 ++++++
 10 files changed, 228 insertions(+), 8 deletions(-)

-- 
2.39.2

[PATCH 1/5] drm/i915: Add ability for tracking buffer objects per client

[Tvrtko Ursulin]

From: Tvrtko Ursulin <tvrtko.ursulin@intel.com>

In order to show per client memory usage lets add some infrastructure
which enables tracking buffer objects owned by clients.

We add a per client list protected by a new per client lock and to support
delayed destruction (post client exit) we make tracked objects hold
references to the owning client.

Also, object memory region teardown is moved to the existing RCU free
callback to allow safe dereference from the fdinfo RCU read section.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
---
 drivers/gpu/drm/i915/gem/i915_gem_object.c    | 13 +++++--
 .../gpu/drm/i915/gem/i915_gem_object_types.h  | 12 +++++++
 drivers/gpu/drm/i915/i915_drm_client.c        | 36 +++++++++++++++++++
 drivers/gpu/drm/i915/i915_drm_client.h        | 32 +++++++++++++++++
 4 files changed, 90 insertions(+), 3 deletions(-)

diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object.c b/drivers/gpu/drm/i915/gem/i915_gem_object.c
index 97ac6fb37958..3dc4fbb67d2b 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_object.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_object.c
@@ -105,6 +105,10 @@ void i915_gem_object_init(struct drm_i915_gem_object *obj,
 
 	INIT_LIST_HEAD(&obj->mm.link);
 
+#ifdef CONFIG_PROC_FS
+	INIT_LIST_HEAD(&obj->client_link);
+#endif
+
 	INIT_LIST_HEAD(&obj->lut_list);
 	spin_lock_init(&obj->lut_lock);
 
@@ -292,6 +296,10 @@ void __i915_gem_free_object_rcu(struct rcu_head *head)
 		container_of(head, typeof(*obj), rcu);
 	struct drm_i915_private *i915 = to_i915(obj->base.dev);
 
+	/* We need to keep this alive for RCU read access from fdinfo. */
+	if (obj->mm.n_placements > 1)
+		kfree(obj->mm.placements);
+
 	i915_gem_object_free(obj);
 
 	GEM_BUG_ON(!atomic_read(&i915->mm.free_count));
@@ -388,9 +396,6 @@ void __i915_gem_free_object(struct drm_i915_gem_object *obj)
 	if (obj->ops->release)
 		obj->ops->release(obj);
 
-	if (obj->mm.n_placements > 1)
-		kfree(obj->mm.placements);
-
 	if (obj->shares_resv_from)
 		i915_vm_resv_put(obj->shares_resv_from);
 
@@ -441,6 +446,8 @@ static void i915_gem_free_object(struct drm_gem_object *gem_obj)
 
 	GEM_BUG_ON(i915_gem_object_is_framebuffer(obj));
 
+	i915_drm_client_remove_object(obj);
+
 	/*
 	 * Before we free the object, make sure any pure RCU-only
 	 * read-side critical sections are complete, e.g.
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object_types.h b/drivers/gpu/drm/i915/gem/i915_gem_object_types.h
index e72c57716bee..8de2b91b3edf 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_object_types.h
+++ b/drivers/gpu/drm/i915/gem/i915_gem_object_types.h
@@ -300,6 +300,18 @@ struct drm_i915_gem_object {
 	 */
 	struct i915_address_space *shares_resv_from;
 
+#ifdef CONFIG_PROC_FS
+	/**
+	 * @client: @i915_drm_client which created the object
+	 */
+	struct i915_drm_client *client;
+
+	/**
+	 * @client_link: Link into @i915_drm_client.objects_list
+	 */
+	struct list_head client_link;
+#endif
+
 	union {
 		struct rcu_head rcu;
 		struct llist_node freed;
diff --git a/drivers/gpu/drm/i915/i915_drm_client.c b/drivers/gpu/drm/i915/i915_drm_client.c
index 2a44b3876cb5..2e5e69edc0f9 100644
--- a/drivers/gpu/drm/i915/i915_drm_client.c
+++ b/drivers/gpu/drm/i915/i915_drm_client.c
@@ -28,6 +28,10 @@ struct i915_drm_client *i915_drm_client_alloc(void)
 	kref_init(&client->kref);
 	spin_lock_init(&client->ctx_lock);
 	INIT_LIST_HEAD(&client->ctx_list);
+#ifdef CONFIG_PROC_FS
+	spin_lock_init(&client->objects_lock);
+	INIT_LIST_HEAD(&client->objects_list);
+#endif
 
 	return client;
 }
@@ -108,4 +112,36 @@ void i915_drm_client_fdinfo(struct drm_printer *p, struct drm_file *file)
 	for (i = 0; i < ARRAY_SIZE(uabi_class_names); i++)
 		show_client_class(p, i915, file_priv->client, i);
 }
+
+void i915_drm_client_add_object(struct i915_drm_client *client,
+				struct drm_i915_gem_object *obj)
+{
+	unsigned long flags;
+
+	GEM_WARN_ON(obj->client);
+	GEM_WARN_ON(!list_empty(&obj->client_link));
+
+	spin_lock_irqsave(&client->objects_lock, flags);
+	obj->client = i915_drm_client_get(client);
+	list_add_tail_rcu(&obj->client_link, &client->objects_list);
+	spin_unlock_irqrestore(&client->objects_lock, flags);
+}
+
+bool i915_drm_client_remove_object(struct drm_i915_gem_object *obj)
+{
+	struct i915_drm_client *client = fetch_and_zero(&obj->client);
+	unsigned long flags;
+
+	/* Object may not be associated with a client. */
+	if (!client)
+		return false;
+
+	spin_lock_irqsave(&client->objects_lock, flags);
+	list_del_rcu(&obj->client_link);
+	spin_unlock_irqrestore(&client->objects_lock, flags);
+
+	i915_drm_client_put(client);
+
+	return true;
+}
 #endif
diff --git a/drivers/gpu/drm/i915/i915_drm_client.h b/drivers/gpu/drm/i915/i915_drm_client.h
index 67816c912bca..5f58fdf7dcb8 100644
--- a/drivers/gpu/drm/i915/i915_drm_client.h
+++ b/drivers/gpu/drm/i915/i915_drm_client.h
@@ -12,6 +12,9 @@
 
 #include <uapi/drm/i915_drm.h>
 
+#include "i915_file_private.h"
+#include "gem/i915_gem_object_types.h"
+
 #define I915_LAST_UABI_ENGINE_CLASS I915_ENGINE_CLASS_COMPUTE
 
 struct drm_file;
@@ -25,6 +28,20 @@ struct i915_drm_client {
 	spinlock_t ctx_lock; /* For add/remove from ctx_list. */
 	struct list_head ctx_list; /* List of contexts belonging to client. */
 
+#ifdef CONFIG_PROC_FS
+	/**
+	 * @objects_lock: lock protecting @objects_list
+	 */
+	spinlock_t objects_lock;
+
+	/**
+	 * @objects_list: list of objects created by this client
+	 *
+	 * Protected by @objects_lock.
+	 */
+	struct list_head objects_list;
+#endif
+
 	/**
 	 * @past_runtime: Accumulation of pphwsp runtimes from closed contexts.
 	 */
@@ -49,4 +66,19 @@ struct i915_drm_client *i915_drm_client_alloc(void);
 
 void i915_drm_client_fdinfo(struct drm_printer *p, struct drm_file *file);
 
+#ifdef CONFIG_PROC_FS
+void i915_drm_client_add_object(struct i915_drm_client *client,
+				struct drm_i915_gem_object *obj);
+bool i915_drm_client_remove_object(struct drm_i915_gem_object *obj);
+#else
+static inline void i915_drm_client_add_object(struct i915_drm_client *client,
+					      struct drm_i915_gem_object *obj)
+{
+}
+
+static inline bool i915_drm_client_remove_object(struct drm_i915_gem_object *obj)
+{
+}
+#endif
+
 #endif /* !__I915_DRM_CLIENT_H__ */
-- 
2.39.2

[PATCH 2/5] drm/i915: Record which client owns a VM

[Tvrtko Ursulin]

From: Tvrtko Ursulin <tvrtko.ursulin@intel.com>

To enable accounting of indirect client memory usage (such as page tables)
in the following patch, lets start recording the creator of each PPGTT.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Aravind Iddamsetty <aravind.iddamsetty@intel.com>
---
 drivers/gpu/drm/i915/gem/i915_gem_context.c       | 11 ++++++++---
 drivers/gpu/drm/i915/gem/i915_gem_context_types.h |  3 +++
 drivers/gpu/drm/i915/gem/selftests/mock_context.c |  4 ++--
 drivers/gpu/drm/i915/gt/intel_gtt.h               |  1 +
 4 files changed, 14 insertions(+), 5 deletions(-)

diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context.c b/drivers/gpu/drm/i915/gem/i915_gem_context.c
index 9a9ff84c90d7..35cf6608180e 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_context.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_context.c
@@ -279,7 +279,8 @@ static int proto_context_set_protected(struct drm_i915_private *i915,
 }
 
 static struct i915_gem_proto_context *
-proto_context_create(struct drm_i915_private *i915, unsigned int flags)
+proto_context_create(struct drm_i915_file_private *fpriv,
+		     struct drm_i915_private *i915, unsigned int flags)
 {
 	struct i915_gem_proto_context *pc, *err;
 
@@ -287,6 +288,7 @@ proto_context_create(struct drm_i915_private *i915, unsigned int flags)
 	if (!pc)
 		return ERR_PTR(-ENOMEM);
 
+	pc->fpriv = fpriv;
 	pc->num_user_engines = -1;
 	pc->user_engines = NULL;
 	pc->user_flags = BIT(UCONTEXT_BANNABLE) |
@@ -1621,6 +1623,7 @@ i915_gem_create_context(struct drm_i915_private *i915,
 			err = PTR_ERR(ppgtt);
 			goto err_ctx;
 		}
+		ppgtt->vm.fpriv = pc->fpriv;
 		vm = &ppgtt->vm;
 	}
 	if (vm)
@@ -1740,7 +1743,7 @@ int i915_gem_context_open(struct drm_i915_private *i915,
 	/* 0 reserved for invalid/unassigned ppgtt */
 	xa_init_flags(&file_priv->vm_xa, XA_FLAGS_ALLOC1);
 
-	pc = proto_context_create(i915, 0);
+	pc = proto_context_create(file_priv, i915, 0);
 	if (IS_ERR(pc)) {
 		err = PTR_ERR(pc);
 		goto err;
@@ -1822,6 +1825,7 @@ int i915_gem_vm_create_ioctl(struct drm_device *dev, void *data,
 
 	GEM_BUG_ON(id == 0); /* reserved for invalid/unassigned ppgtt */
 	args->vm_id = id;
+	ppgtt->vm.fpriv = file_priv;
 	return 0;
 
 err_put:
@@ -2284,7 +2288,8 @@ int i915_gem_context_create_ioctl(struct drm_device *dev, void *data,
 		return -EIO;
 	}
 
-	ext_data.pc = proto_context_create(i915, args->flags);
+	ext_data.pc = proto_context_create(file->driver_priv, i915,
+					   args->flags);
 	if (IS_ERR(ext_data.pc))
 		return PTR_ERR(ext_data.pc);
 
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context_types.h b/drivers/gpu/drm/i915/gem/i915_gem_context_types.h
index cb78214a7dcd..c573c067779f 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_context_types.h
+++ b/drivers/gpu/drm/i915/gem/i915_gem_context_types.h
@@ -188,6 +188,9 @@ struct i915_gem_proto_engine {
  * CONTEXT_CREATE_SET_PARAM during GEM_CONTEXT_CREATE.
  */
 struct i915_gem_proto_context {
+	/** @fpriv: Client which creates the context */
+	struct drm_i915_file_private *fpriv;
+
 	/** @vm: See &i915_gem_context.vm */
 	struct i915_address_space *vm;
 
diff --git a/drivers/gpu/drm/i915/gem/selftests/mock_context.c b/drivers/gpu/drm/i915/gem/selftests/mock_context.c
index 8ac6726ec16b..125584ada282 100644
--- a/drivers/gpu/drm/i915/gem/selftests/mock_context.c
+++ b/drivers/gpu/drm/i915/gem/selftests/mock_context.c
@@ -83,7 +83,7 @@ live_context(struct drm_i915_private *i915, struct file *file)
 	int err;
 	u32 id;
 
-	pc = proto_context_create(i915, 0);
+	pc = proto_context_create(fpriv, i915, 0);
 	if (IS_ERR(pc))
 		return ERR_CAST(pc);
 
@@ -152,7 +152,7 @@ kernel_context(struct drm_i915_private *i915,
 	struct i915_gem_context *ctx;
 	struct i915_gem_proto_context *pc;
 
-	pc = proto_context_create(i915, 0);
+	pc = proto_context_create(NULL, i915, 0);
 	if (IS_ERR(pc))
 		return ERR_CAST(pc);
 
diff --git a/drivers/gpu/drm/i915/gt/intel_gtt.h b/drivers/gpu/drm/i915/gt/intel_gtt.h
index 4d6296cdbcfd..7192a534a654 100644
--- a/drivers/gpu/drm/i915/gt/intel_gtt.h
+++ b/drivers/gpu/drm/i915/gt/intel_gtt.h
@@ -248,6 +248,7 @@ struct i915_address_space {
 	struct drm_mm mm;
 	struct intel_gt *gt;
 	struct drm_i915_private *i915;
+	struct drm_i915_file_private *fpriv;
 	struct device *dma;
 	u64 total;		/* size addr space maps (ex. 2GB for ggtt) */
 	u64 reserved;		/* size addr space reserved */
-- 
2.39.2

[PATCH 3/5] drm/i915: Track page table backing store usage

[Tvrtko Ursulin]

From: Tvrtko Ursulin <tvrtko.ursulin@intel.com>

Account page table backing store against the owning client memory usage
stats.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Aravind Iddamsetty <aravind.iddamsetty@intel.com>
---
 drivers/gpu/drm/i915/gt/intel_gtt.c | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/drivers/gpu/drm/i915/gt/intel_gtt.c b/drivers/gpu/drm/i915/gt/intel_gtt.c
index 731d9f2bbc56..065099362a98 100644
--- a/drivers/gpu/drm/i915/gt/intel_gtt.c
+++ b/drivers/gpu/drm/i915/gt/intel_gtt.c
@@ -58,6 +58,9 @@ struct drm_i915_gem_object *alloc_pt_lmem(struct i915_address_space *vm, int sz)
 	if (!IS_ERR(obj)) {
 		obj->base.resv = i915_vm_resv_get(vm);
 		obj->shares_resv_from = vm;
+
+		if (vm->fpriv)
+			i915_drm_client_add_object(vm->fpriv->client, obj);
 	}
 
 	return obj;
@@ -79,6 +82,9 @@ struct drm_i915_gem_object *alloc_pt_dma(struct i915_address_space *vm, int sz)
 	if (!IS_ERR(obj)) {
 		obj->base.resv = i915_vm_resv_get(vm);
 		obj->shares_resv_from = vm;
+
+		if (vm->fpriv)
+			i915_drm_client_add_object(vm->fpriv->client, obj);
 	}
 
 	return obj;
-- 
2.39.2

[PATCH 4/5] drm/i915: Account ring buffer and context state storage

[Tvrtko Ursulin]

From: Tvrtko Ursulin <tvrtko.ursulin@intel.com>

Account ring buffers and logical context space against the owning client
memory usage stats.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Aravind Iddamsetty <aravind.iddamsetty@intel.com>
---
 drivers/gpu/drm/i915/gt/intel_context.c | 14 ++++++++++++++
 drivers/gpu/drm/i915/i915_drm_client.c  | 10 ++++++++++
 drivers/gpu/drm/i915/i915_drm_client.h  |  9 +++++++++
 3 files changed, 33 insertions(+)

diff --git a/drivers/gpu/drm/i915/gt/intel_context.c b/drivers/gpu/drm/i915/gt/intel_context.c
index a53b26178f0a..a2f1245741bb 100644
--- a/drivers/gpu/drm/i915/gt/intel_context.c
+++ b/drivers/gpu/drm/i915/gt/intel_context.c
@@ -6,6 +6,7 @@
 #include "gem/i915_gem_context.h"
 #include "gem/i915_gem_pm.h"
 
+#include "i915_drm_client.h"
 #include "i915_drv.h"
 #include "i915_trace.h"
 
@@ -50,6 +51,7 @@ intel_context_create(struct intel_engine_cs *engine)
 
 int intel_context_alloc_state(struct intel_context *ce)
 {
+	struct i915_gem_context *ctx;
 	int err = 0;
 
 	if (mutex_lock_interruptible(&ce->pin_mutex))
@@ -66,6 +68,18 @@ int intel_context_alloc_state(struct intel_context *ce)
 			goto unlock;
 
 		set_bit(CONTEXT_ALLOC_BIT, &ce->flags);
+
+		rcu_read_lock();
+		ctx = rcu_dereference(ce->gem_context);
+		if (ctx && !kref_get_unless_zero(&ctx->ref))
+			ctx = NULL;
+		rcu_read_unlock();
+		if (ctx) {
+			if (ctx->client)
+				i915_drm_client_add_context_objects(ctx->client,
+								    ce);
+			i915_gem_context_put(ctx);
+		}
 	}
 
 unlock:
diff --git a/drivers/gpu/drm/i915/i915_drm_client.c b/drivers/gpu/drm/i915/i915_drm_client.c
index 2e5e69edc0f9..a61356012df8 100644
--- a/drivers/gpu/drm/i915/i915_drm_client.c
+++ b/drivers/gpu/drm/i915/i915_drm_client.c
@@ -144,4 +144,14 @@ bool i915_drm_client_remove_object(struct drm_i915_gem_object *obj)
 
 	return true;
 }
+
+void i915_drm_client_add_context_objects(struct i915_drm_client *client,
+					 struct intel_context *ce)
+{
+	if (ce->state)
+		i915_drm_client_add_object(client, ce->state->obj);
+
+	if (ce->ring != ce->engine->legacy.ring && ce->ring->vma)
+		i915_drm_client_add_object(client, ce->ring->vma->obj);
+}
 #endif
diff --git a/drivers/gpu/drm/i915/i915_drm_client.h b/drivers/gpu/drm/i915/i915_drm_client.h
index 5f58fdf7dcb8..69cedfcd3d69 100644
--- a/drivers/gpu/drm/i915/i915_drm_client.h
+++ b/drivers/gpu/drm/i915/i915_drm_client.h
@@ -14,6 +14,7 @@
 
 #include "i915_file_private.h"
 #include "gem/i915_gem_object_types.h"
+#include "gt/intel_context_types.h"
 
 #define I915_LAST_UABI_ENGINE_CLASS I915_ENGINE_CLASS_COMPUTE
 
@@ -70,6 +71,8 @@ void i915_drm_client_fdinfo(struct drm_printer *p, struct drm_file *file);
 void i915_drm_client_add_object(struct i915_drm_client *client,
 				struct drm_i915_gem_object *obj);
 bool i915_drm_client_remove_object(struct drm_i915_gem_object *obj);
+void i915_drm_client_add_context_objects(struct i915_drm_client *client,
+					 struct intel_context *ce);
 #else
 static inline void i915_drm_client_add_object(struct i915_drm_client *client,
 					      struct drm_i915_gem_object *obj)
@@ -79,6 +82,12 @@ static inline void i915_drm_client_add_object(struct i915_drm_client *client,
 static inline bool i915_drm_client_remove_object(struct drm_i915_gem_object *obj)
 {
 }
+
+static inline void
+i915_drm_client_add_context_objects(struct i915_drm_client *client,
+				    struct intel_context *ce)
+{
+}
 #endif
 
 #endif /* !__I915_DRM_CLIENT_H__ */
-- 
2.39.2

[PATCH 5/5] drm/i915: Implement fdinfo memory stats printing

[Tvrtko Ursulin]

From: Tvrtko Ursulin <tvrtko.ursulin@intel.com>

Use the newly added drm_print_memory_stats helper to show memory
utilisation of our objects in drm/driver specific fdinfo output.

To collect the stats we walk the per memory regions object lists
and accumulate object size into the respective drm_memory_stats
categories.

Objects with multiple possible placements are reported in multiple
regions for total and shared sizes, while other categories are
counted only for the currently active region.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Aravind Iddamsetty <aravind.iddamsetty@intel.com>
Cc: Rob Clark <robdclark@gmail.com>
---
 drivers/gpu/drm/i915/i915_drm_client.c | 85 ++++++++++++++++++++++++++
 1 file changed, 85 insertions(+)

diff --git a/drivers/gpu/drm/i915/i915_drm_client.c b/drivers/gpu/drm/i915/i915_drm_client.c
index a61356012df8..9e7a6075ee25 100644
--- a/drivers/gpu/drm/i915/i915_drm_client.c
+++ b/drivers/gpu/drm/i915/i915_drm_client.c
@@ -45,6 +45,89 @@ void __i915_drm_client_free(struct kref *kref)
 }
 
 #ifdef CONFIG_PROC_FS
+static void
+obj_meminfo(struct drm_i915_gem_object *obj,
+	    struct drm_memory_stats stats[INTEL_REGION_UNKNOWN])
+{
+	struct intel_memory_region *mr;
+	u64 sz = obj->base.size;
+	enum intel_region_id id;
+	unsigned int i;
+
+	/* Attribute size and shared to all possible memory regions. */
+	for (i = 0; i < obj->mm.n_placements; i++) {
+		mr = obj->mm.placements[i];
+		id = mr->id;
+
+		if (obj->base.handle_count > 1)
+			stats[id].shared += sz;
+		else
+			stats[id].private += sz;
+	}
+
+	/* Attribute other categories to only the current region. */
+	mr = obj->mm.region;
+	if (mr)
+		id = mr->id;
+	else
+		id = INTEL_REGION_SMEM;
+
+	if (!obj->mm.n_placements) {
+		if (obj->base.handle_count > 1)
+			stats[id].shared += sz;
+		else
+			stats[id].private += sz;
+	}
+
+	if (i915_gem_object_has_pages(obj)) {
+		stats[id].resident += sz;
+
+		if (!dma_resv_test_signaled(obj->base.resv,
+					    dma_resv_usage_rw(true)))
+			stats[id].active += sz;
+		else if (i915_gem_object_is_shrinkable(obj) &&
+			 obj->mm.madv == I915_MADV_DONTNEED)
+			stats[id].purgeable += sz;
+	}
+}
+
+static void show_meminfo(struct drm_printer *p, struct drm_file *file)
+{
+	struct drm_memory_stats stats[INTEL_REGION_UNKNOWN] = {};
+	struct drm_i915_file_private *fpriv = file->driver_priv;
+	struct i915_drm_client *client = fpriv->client;
+	struct drm_i915_private *i915 = fpriv->i915;
+	struct drm_i915_gem_object *obj;
+	struct intel_memory_region *mr;
+	struct list_head *pos;
+	unsigned int id;
+
+	/* Public objects. */
+	spin_lock(&file->table_lock);
+	idr_for_each_entry(&file->object_idr, obj, id)
+		obj_meminfo(obj, stats);
+	spin_unlock(&file->table_lock);
+
+	/* Internal objects. */
+	rcu_read_lock();
+	list_for_each_rcu(pos, &client->objects_list) {
+		obj = i915_gem_object_get_rcu(list_entry(pos, typeof(*obj),
+							 client_link));
+		if (!obj)
+			continue;
+		obj_meminfo(obj, stats);
+		i915_gem_object_put(obj);
+	}
+	rcu_read_unlock();
+
+	for_each_memory_region(mr, i915, id)
+		drm_print_memory_stats(p,
+				       &stats[id],
+				       DRM_GEM_OBJECT_RESIDENT |
+				       DRM_GEM_OBJECT_PURGEABLE,
+				       mr->name);
+}
+
 static const char * const uabi_class_names[] = {
 	[I915_ENGINE_CLASS_RENDER] = "render",
 	[I915_ENGINE_CLASS_COPY] = "copy",
@@ -106,6 +189,8 @@ void i915_drm_client_fdinfo(struct drm_printer *p, struct drm_file *file)
 	 * ******************************************************************
 	 */
 
+	show_meminfo(p, file);
+
 	if (GRAPHICS_VER(i915) < 8)
 		return;
 
-- 
2.39.2

Re: [PATCH 5/5] drm/i915: Implement fdinfo memory stats printing

[Iddamsetty, Aravind]

On 27-07-2023 15:43, Tvrtko Ursulin wrote:
> From: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
> 
> Use the newly added drm_print_memory_stats helper to show memory
> utilisation of our objects in drm/driver specific fdinfo output.
> 
> To collect the stats we walk the per memory regions object lists
> and accumulate object size into the respective drm_memory_stats
> categories.
> 
> Objects with multiple possible placements are reported in multiple
> regions for total and shared sizes, while other categories are
> counted only for the currently active region.
> 
> Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
> Cc: Aravind Iddamsetty <aravind.iddamsetty@intel.com>
> Cc: Rob Clark <robdclark@gmail.com>> ---
>  drivers/gpu/drm/i915/i915_drm_client.c | 85 ++++++++++++++++++++++++++
>  1 file changed, 85 insertions(+)
> 
> diff --git a/drivers/gpu/drm/i915/i915_drm_client.c b/drivers/gpu/drm/i915/i915_drm_client.c
> index a61356012df8..9e7a6075ee25 100644
> --- a/drivers/gpu/drm/i915/i915_drm_client.c
> +++ b/drivers/gpu/drm/i915/i915_drm_client.c
> @@ -45,6 +45,89 @@ void __i915_drm_client_free(struct kref *kref)
>  }
>  
>  #ifdef CONFIG_PROC_FS
> +static void
> +obj_meminfo(struct drm_i915_gem_object *obj,
> +	    struct drm_memory_stats stats[INTEL_REGION_UNKNOWN])
> +{
> +	struct intel_memory_region *mr;
> +	u64 sz = obj->base.size;
> +	enum intel_region_id id;
> +	unsigned int i;
> +
> +	/* Attribute size and shared to all possible memory regions. */
> +	for (i = 0; i < obj->mm.n_placements; i++) {
> +		mr = obj->mm.placements[i];
> +		id = mr->id;
> +
> +		if (obj->base.handle_count > 1)
> +			stats[id].shared += sz;
> +		else
> +			stats[id].private += sz;
> +	}
> +
> +	/* Attribute other categories to only the current region. */
> +	mr = obj->mm.region;
> +	if (mr)
> +		id = mr->id;
> +	else
> +		id = INTEL_REGION_SMEM;
> +
> +	if (!obj->mm.n_placements) {

I guess we do not expect to have n_placements set to public objects, is
that right?

Thanks,
Aravind.
> +		if (obj->base.handle_count > 1)
> +			stats[id].shared += sz;
> +		else
> +			stats[id].private += sz;
> +	}
> +
> +	if (i915_gem_object_has_pages(obj)) {
> +		stats[id].resident += sz;
> +
> +		if (!dma_resv_test_signaled(obj->base.resv,
> +					    dma_resv_usage_rw(true)))
> +			stats[id].active += sz;
> +		else if (i915_gem_object_is_shrinkable(obj) &&
> +			 obj->mm.madv == I915_MADV_DONTNEED)
> +			stats[id].purgeable += sz;
> +	}
> +}
> +
> +static void show_meminfo(struct drm_printer *p, struct drm_file *file)
> +{
> +	struct drm_memory_stats stats[INTEL_REGION_UNKNOWN] = {};
> +	struct drm_i915_file_private *fpriv = file->driver_priv;
> +	struct i915_drm_client *client = fpriv->client;
> +	struct drm_i915_private *i915 = fpriv->i915;
> +	struct drm_i915_gem_object *obj;
> +	struct intel_memory_region *mr;
> +	struct list_head *pos;
> +	unsigned int id;
> +
> +	/* Public objects. */
> +	spin_lock(&file->table_lock);
> +	idr_for_each_entry(&file->object_idr, obj, id)
> +		obj_meminfo(obj, stats);
> +	spin_unlock(&file->table_lock);
> +
> +	/* Internal objects. */
> +	rcu_read_lock();
> +	list_for_each_rcu(pos, &client->objects_list) {
> +		obj = i915_gem_object_get_rcu(list_entry(pos, typeof(*obj),
> +							 client_link));
> +		if (!obj)
> +			continue;
> +		obj_meminfo(obj, stats);
> +		i915_gem_object_put(obj);
> +	}
> +	rcu_read_unlock();
> +
> +	for_each_memory_region(mr, i915, id)
> +		drm_print_memory_stats(p,
> +				       &stats[id],
> +				       DRM_GEM_OBJECT_RESIDENT |
> +				       DRM_GEM_OBJECT_PURGEABLE,
> +				       mr->name);
> +}
> +
>  static const char * const uabi_class_names[] = {
>  	[I915_ENGINE_CLASS_RENDER] = "render",
>  	[I915_ENGINE_CLASS_COPY] = "copy",
> @@ -106,6 +189,8 @@ void i915_drm_client_fdinfo(struct drm_printer *p, struct drm_file *file)
>  	 * ******************************************************************
>  	 */
>  
> +	show_meminfo(p, file);
> +
>  	if (GRAPHICS_VER(i915) < 8)
>  		return;
>

Re: [Intel-gfx] [PATCH 1/5] drm/i915: Add ability for tracking buffer objects per client

[Iddamsetty, Aravind]

On 27-07-2023 15:43, Tvrtko Ursulin wrote:
> From: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
> 
> In order to show per client memory usage lets add some infrastructure
> which enables tracking buffer objects owned by clients.
> 
> We add a per client list protected by a new per client lock and to support
> delayed destruction (post client exit) we make tracked objects hold
> references to the owning client.
> 
> Also, object memory region teardown is moved to the existing RCU free
> callback to allow safe dereference from the fdinfo RCU read section.

This is same as the earlier series, which I had reviewed but forgot to
give r-b. sorry for the delay.

Reviewed-by: Aravind Iddamsetty <aravind.iddamsetty@intel.com>

Thanks,
Aravind.
> 
> Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
> ---
>  drivers/gpu/drm/i915/gem/i915_gem_object.c    | 13 +++++--
>  .../gpu/drm/i915/gem/i915_gem_object_types.h  | 12 +++++++
>  drivers/gpu/drm/i915/i915_drm_client.c        | 36 +++++++++++++++++++
>  drivers/gpu/drm/i915/i915_drm_client.h        | 32 +++++++++++++++++
>  4 files changed, 90 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object.c b/drivers/gpu/drm/i915/gem/i915_gem_object.c
> index 97ac6fb37958..3dc4fbb67d2b 100644
> --- a/drivers/gpu/drm/i915/gem/i915_gem_object.c
> +++ b/drivers/gpu/drm/i915/gem/i915_gem_object.c
> @@ -105,6 +105,10 @@ void i915_gem_object_init(struct drm_i915_gem_object *obj,
>  
>  	INIT_LIST_HEAD(&obj->mm.link);
>  
> +#ifdef CONFIG_PROC_FS
> +	INIT_LIST_HEAD(&obj->client_link);
> +#endif
> +
>  	INIT_LIST_HEAD(&obj->lut_list);
>  	spin_lock_init(&obj->lut_lock);
>  
> @@ -292,6 +296,10 @@ void __i915_gem_free_object_rcu(struct rcu_head *head)
>  		container_of(head, typeof(*obj), rcu);
>  	struct drm_i915_private *i915 = to_i915(obj->base.dev);
>  
> +	/* We need to keep this alive for RCU read access from fdinfo. */
> +	if (obj->mm.n_placements > 1)
> +		kfree(obj->mm.placements);
> +
>  	i915_gem_object_free(obj);
>  
>  	GEM_BUG_ON(!atomic_read(&i915->mm.free_count));
> @@ -388,9 +396,6 @@ void __i915_gem_free_object(struct drm_i915_gem_object *obj)
>  	if (obj->ops->release)
>  		obj->ops->release(obj);
>  
> -	if (obj->mm.n_placements > 1)
> -		kfree(obj->mm.placements);
> -
>  	if (obj->shares_resv_from)
>  		i915_vm_resv_put(obj->shares_resv_from);
>  
> @@ -441,6 +446,8 @@ static void i915_gem_free_object(struct drm_gem_object *gem_obj)
>  
>  	GEM_BUG_ON(i915_gem_object_is_framebuffer(obj));
>  
> +	i915_drm_client_remove_object(obj);
> +
>  	/*
>  	 * Before we free the object, make sure any pure RCU-only
>  	 * read-side critical sections are complete, e.g.
> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object_types.h b/drivers/gpu/drm/i915/gem/i915_gem_object_types.h
> index e72c57716bee..8de2b91b3edf 100644
> --- a/drivers/gpu/drm/i915/gem/i915_gem_object_types.h
> +++ b/drivers/gpu/drm/i915/gem/i915_gem_object_types.h
> @@ -300,6 +300,18 @@ struct drm_i915_gem_object {
>  	 */
>  	struct i915_address_space *shares_resv_from;
>  
> +#ifdef CONFIG_PROC_FS
> +	/**
> +	 * @client: @i915_drm_client which created the object
> +	 */
> +	struct i915_drm_client *client;
> +
> +	/**
> +	 * @client_link: Link into @i915_drm_client.objects_list
> +	 */
> +	struct list_head client_link;
> +#endif
> +
>  	union {
>  		struct rcu_head rcu;
>  		struct llist_node freed;
> diff --git a/drivers/gpu/drm/i915/i915_drm_client.c b/drivers/gpu/drm/i915/i915_drm_client.c
> index 2a44b3876cb5..2e5e69edc0f9 100644
> --- a/drivers/gpu/drm/i915/i915_drm_client.c
> +++ b/drivers/gpu/drm/i915/i915_drm_client.c
> @@ -28,6 +28,10 @@ struct i915_drm_client *i915_drm_client_alloc(void)
>  	kref_init(&client->kref);
>  	spin_lock_init(&client->ctx_lock);
>  	INIT_LIST_HEAD(&client->ctx_list);
> +#ifdef CONFIG_PROC_FS
> +	spin_lock_init(&client->objects_lock);
> +	INIT_LIST_HEAD(&client->objects_list);
> +#endif
>  
>  	return client;
>  }
> @@ -108,4 +112,36 @@ void i915_drm_client_fdinfo(struct drm_printer *p, struct drm_file *file)
>  	for (i = 0; i < ARRAY_SIZE(uabi_class_names); i++)
>  		show_client_class(p, i915, file_priv->client, i);
>  }
> +
> +void i915_drm_client_add_object(struct i915_drm_client *client,
> +				struct drm_i915_gem_object *obj)
> +{
> +	unsigned long flags;
> +
> +	GEM_WARN_ON(obj->client);
> +	GEM_WARN_ON(!list_empty(&obj->client_link));
> +
> +	spin_lock_irqsave(&client->objects_lock, flags);
> +	obj->client = i915_drm_client_get(client);
> +	list_add_tail_rcu(&obj->client_link, &client->objects_list);
> +	spin_unlock_irqrestore(&client->objects_lock, flags);
> +}
> +
> +bool i915_drm_client_remove_object(struct drm_i915_gem_object *obj)
> +{
> +	struct i915_drm_client *client = fetch_and_zero(&obj->client);
> +	unsigned long flags;
> +
> +	/* Object may not be associated with a client. */
> +	if (!client)
> +		return false;
> +
> +	spin_lock_irqsave(&client->objects_lock, flags);
> +	list_del_rcu(&obj->client_link);
> +	spin_unlock_irqrestore(&client->objects_lock, flags);
> +
> +	i915_drm_client_put(client);
> +
> +	return true;
> +}
>  #endif
> diff --git a/drivers/gpu/drm/i915/i915_drm_client.h b/drivers/gpu/drm/i915/i915_drm_client.h
> index 67816c912bca..5f58fdf7dcb8 100644
> --- a/drivers/gpu/drm/i915/i915_drm_client.h
> +++ b/drivers/gpu/drm/i915/i915_drm_client.h
> @@ -12,6 +12,9 @@
>  
>  #include <uapi/drm/i915_drm.h>
>  
> +#include "i915_file_private.h"
> +#include "gem/i915_gem_object_types.h"
> +
>  #define I915_LAST_UABI_ENGINE_CLASS I915_ENGINE_CLASS_COMPUTE
>  
>  struct drm_file;
> @@ -25,6 +28,20 @@ struct i915_drm_client {
>  	spinlock_t ctx_lock; /* For add/remove from ctx_list. */
>  	struct list_head ctx_list; /* List of contexts belonging to client. */
>  
> +#ifdef CONFIG_PROC_FS
> +	/**
> +	 * @objects_lock: lock protecting @objects_list
> +	 */
> +	spinlock_t objects_lock;
> +
> +	/**
> +	 * @objects_list: list of objects created by this client
> +	 *
> +	 * Protected by @objects_lock.
> +	 */
> +	struct list_head objects_list;
> +#endif
> +
>  	/**
>  	 * @past_runtime: Accumulation of pphwsp runtimes from closed contexts.
>  	 */
> @@ -49,4 +66,19 @@ struct i915_drm_client *i915_drm_client_alloc(void);
>  
>  void i915_drm_client_fdinfo(struct drm_printer *p, struct drm_file *file);
>  
> +#ifdef CONFIG_PROC_FS
> +void i915_drm_client_add_object(struct i915_drm_client *client,
> +				struct drm_i915_gem_object *obj);
> +bool i915_drm_client_remove_object(struct drm_i915_gem_object *obj);
> +#else
> +static inline void i915_drm_client_add_object(struct i915_drm_client *client,
> +					      struct drm_i915_gem_object *obj)
> +{
> +}
> +
> +static inline bool i915_drm_client_remove_object(struct drm_i915_gem_object *obj)
> +{
> +}
> +#endif
> +
>  #endif /* !__I915_DRM_CLIENT_H__ */

Re: [PATCH 5/5] drm/i915: Implement fdinfo memory stats printing

[Tvrtko Ursulin]

On 03/08/2023 06:15, Iddamsetty, Aravind wrote:
> On 27-07-2023 15:43, Tvrtko Ursulin wrote:
>> From: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
>>
>> Use the newly added drm_print_memory_stats helper to show memory
>> utilisation of our objects in drm/driver specific fdinfo output.
>>
>> To collect the stats we walk the per memory regions object lists
>> and accumulate object size into the respective drm_memory_stats
>> categories.
>>
>> Objects with multiple possible placements are reported in multiple
>> regions for total and shared sizes, while other categories are
>> counted only for the currently active region.
>>
>> Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
>> Cc: Aravind Iddamsetty <aravind.iddamsetty@intel.com>
>> Cc: Rob Clark <robdclark@gmail.com>> ---
>>   drivers/gpu/drm/i915/i915_drm_client.c | 85 ++++++++++++++++++++++++++
>>   1 file changed, 85 insertions(+)
>>
>> diff --git a/drivers/gpu/drm/i915/i915_drm_client.c b/drivers/gpu/drm/i915/i915_drm_client.c
>> index a61356012df8..9e7a6075ee25 100644
>> --- a/drivers/gpu/drm/i915/i915_drm_client.c
>> +++ b/drivers/gpu/drm/i915/i915_drm_client.c
>> @@ -45,6 +45,89 @@ void __i915_drm_client_free(struct kref *kref)
>>   }
>>   
>>   #ifdef CONFIG_PROC_FS
>> +static void
>> +obj_meminfo(struct drm_i915_gem_object *obj,
>> +	    struct drm_memory_stats stats[INTEL_REGION_UNKNOWN])
>> +{
>> +	struct intel_memory_region *mr;
>> +	u64 sz = obj->base.size;
>> +	enum intel_region_id id;
>> +	unsigned int i;
>> +
>> +	/* Attribute size and shared to all possible memory regions. */
>> +	for (i = 0; i < obj->mm.n_placements; i++) {
>> +		mr = obj->mm.placements[i];
>> +		id = mr->id;
>> +
>> +		if (obj->base.handle_count > 1)
>> +			stats[id].shared += sz;
>> +		else
>> +			stats[id].private += sz;
>> +	}
>> +
>> +	/* Attribute other categories to only the current region. */
>> +	mr = obj->mm.region;
>> +	if (mr)
>> +		id = mr->id;
>> +	else
>> +		id = INTEL_REGION_SMEM;
>> +
>> +	if (!obj->mm.n_placements) {
> 
> I guess we do not expect to have n_placements set to public objects, is
> that right?

I think they are the only ones which can have placements. It is via 
I915_GEM_CREATE_EXT_MEMORY_REGIONS userspace is able to create them.

My main conundrum in this patch is a few lines above, the loop which 
adds shared and private.

Question is, if an object can be either smem or lmem, how do we want to 
report it? This patch adds the size for all possible regions and 
resident and active only to the currently active. But perhaps that is 
wrong. Maybe I should change it is only against the active region and 
multiple regions are just ignored. Then if object is migrated do access 
patterns or memory pressure, the total size would migrate too.

I think I was trying to achieve something here (have more visibility on 
what kind of backing store clients are allocating) which maybe does not 
work to well with the current categories.

Namely if userspace allocates say one 1MiB object with placement in 
either smem or lmem, and it is currently resident in lmem, I wanted it 
to show as:

  total-smem: 1 MiB
  resident-smem: 0
  total-lmem: 1 MiB
  resident-lmem: 1 MiB

To constantly show how in theory client could be using memory from 
either region. Maybe that is misleading and should instead be:

  total-smem: 0
  resident-smem: 0
  total-lmem: 1 MiB
  resident-lmem: 1 MiB

?

And then if/when the same object gets migrated to smem it changes to 
(lets assume it is also not resident any more but got swapped out):

  total-smem: 1 MiB
  resident-smem: 0
  total-lmem: 0
  resident-lmem: 0

Regards,

Tvrtko

>> +		if (obj->base.handle_count > 1)
>> +			stats[id].shared += sz;
>> +		else
>> +			stats[id].private += sz;
>> +	}
>> +
>> +	if (i915_gem_object_has_pages(obj)) {
>> +		stats[id].resident += sz;
>> +
>> +		if (!dma_resv_test_signaled(obj->base.resv,
>> +					    dma_resv_usage_rw(true)))
>> +			stats[id].active += sz;
>> +		else if (i915_gem_object_is_shrinkable(obj) &&
>> +			 obj->mm.madv == I915_MADV_DONTNEED)
>> +			stats[id].purgeable += sz;
>> +	}
>> +}
>> +
>> +static void show_meminfo(struct drm_printer *p, struct drm_file *file)
>> +{
>> +	struct drm_memory_stats stats[INTEL_REGION_UNKNOWN] = {};
>> +	struct drm_i915_file_private *fpriv = file->driver_priv;
>> +	struct i915_drm_client *client = fpriv->client;
>> +	struct drm_i915_private *i915 = fpriv->i915;
>> +	struct drm_i915_gem_object *obj;
>> +	struct intel_memory_region *mr;
>> +	struct list_head *pos;
>> +	unsigned int id;
>> +
>> +	/* Public objects. */
>> +	spin_lock(&file->table_lock);
>> +	idr_for_each_entry(&file->object_idr, obj, id)
>> +		obj_meminfo(obj, stats);
>> +	spin_unlock(&file->table_lock);
>> +
>> +	/* Internal objects. */
>> +	rcu_read_lock();
>> +	list_for_each_rcu(pos, &client->objects_list) {
>> +		obj = i915_gem_object_get_rcu(list_entry(pos, typeof(*obj),
>> +							 client_link));
>> +		if (!obj)
>> +			continue;
>> +		obj_meminfo(obj, stats);
>> +		i915_gem_object_put(obj);
>> +	}
>> +	rcu_read_unlock();
>> +
>> +	for_each_memory_region(mr, i915, id)
>> +		drm_print_memory_stats(p,
>> +				       &stats[id],
>> +				       DRM_GEM_OBJECT_RESIDENT |
>> +				       DRM_GEM_OBJECT_PURGEABLE,
>> +				       mr->name);
>> +}
>> +
>>   static const char * const uabi_class_names[] = {
>>   	[I915_ENGINE_CLASS_RENDER] = "render",
>>   	[I915_ENGINE_CLASS_COPY] = "copy",
>> @@ -106,6 +189,8 @@ void i915_drm_client_fdinfo(struct drm_printer *p, struct drm_file *file)
>>   	 * ******************************************************************
>>   	 */
>>   
>> +	show_meminfo(p, file);
>> +
>>   	if (GRAPHICS_VER(i915) < 8)
>>   		return;
>>
> 
> 
> 

Re: [PATCH 5/5] drm/i915: Implement fdinfo memory stats printing

[Iddamsetty, Aravind]

On 03-08-2023 14:19, Tvrtko Ursulin wrote:
> 
> On 03/08/2023 06:15, Iddamsetty, Aravind wrote:
>> On 27-07-2023 15:43, Tvrtko Ursulin wrote:
>>> From: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
>>>
>>> Use the newly added drm_print_memory_stats helper to show memory
>>> utilisation of our objects in drm/driver specific fdinfo output.
>>>
>>> To collect the stats we walk the per memory regions object lists
>>> and accumulate object size into the respective drm_memory_stats
>>> categories.
>>>
>>> Objects with multiple possible placements are reported in multiple
>>> regions for total and shared sizes, while other categories are
>>> counted only for the currently active region.
>>>
>>> Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
>>> Cc: Aravind Iddamsetty <aravind.iddamsetty@intel.com>
>>> Cc: Rob Clark <robdclark@gmail.com>> ---
>>>   drivers/gpu/drm/i915/i915_drm_client.c | 85 ++++++++++++++++++++++++++
>>>   1 file changed, 85 insertions(+)
>>>
>>> diff --git a/drivers/gpu/drm/i915/i915_drm_client.c
>>> b/drivers/gpu/drm/i915/i915_drm_client.c
>>> index a61356012df8..9e7a6075ee25 100644
>>> --- a/drivers/gpu/drm/i915/i915_drm_client.c
>>> +++ b/drivers/gpu/drm/i915/i915_drm_client.c
>>> @@ -45,6 +45,89 @@ void __i915_drm_client_free(struct kref *kref)
>>>   }
>>>     #ifdef CONFIG_PROC_FS
>>> +static void
>>> +obj_meminfo(struct drm_i915_gem_object *obj,
>>> +        struct drm_memory_stats stats[INTEL_REGION_UNKNOWN])
>>> +{
>>> +    struct intel_memory_region *mr;
>>> +    u64 sz = obj->base.size;
>>> +    enum intel_region_id id;
>>> +    unsigned int i;
>>> +
>>> +    /* Attribute size and shared to all possible memory regions. */
>>> +    for (i = 0; i < obj->mm.n_placements; i++) {
>>> +        mr = obj->mm.placements[i];
>>> +        id = mr->id;
>>> +
>>> +        if (obj->base.handle_count > 1)
>>> +            stats[id].shared += sz;
>>> +        else
>>> +            stats[id].private += sz;
>>> +    }
>>> +
>>> +    /* Attribute other categories to only the current region. */
>>> +    mr = obj->mm.region;
>>> +    if (mr)
>>> +        id = mr->id;
>>> +    else
>>> +        id = INTEL_REGION_SMEM;
>>> +
>>> +    if (!obj->mm.n_placements) {
>>
>> I guess we do not expect to have n_placements set to public objects, is
>> that right?
> 
> I think they are the only ones which can have placements. It is via
> I915_GEM_CREATE_EXT_MEMORY_REGIONS userspace is able to create them.
> 
> My main conundrum in this patch is a few lines above, the loop which
> adds shared and private.
> 
> Question is, if an object can be either smem or lmem, how do we want to
> report it? This patch adds the size for all possible regions and
> resident and active only to the currently active. But perhaps that is
> wrong. Maybe I should change it is only against the active region and
> multiple regions are just ignored. Then if object is migrated do access
> patterns or memory pressure, the total size would migrate too.
> 
> I think I was trying to achieve something here (have more visibility on
> what kind of backing store clients are allocating) which maybe does not
> work to well with the current categories.
> 
> Namely if userspace allocates say one 1MiB object with placement in
> either smem or lmem, and it is currently resident in lmem, I wanted it
> to show as:
> 
>  total-smem: 1 MiB
>  resident-smem: 0
>  total-lmem: 1 MiB
>  resident-lmem: 1 MiB
> 
> To constantly show how in theory client could be using memory from
> either region. Maybe that is misleading and should instead be:
> 
>  total-smem: 0
>  resident-smem: 0
>  total-lmem: 1 MiB
>  resident-lmem: 1 MiB
> 
> ?

I think the current implementation will not match with the memregion
info in query ioctl as well. While what you say is true I'm not sure if
there can be a client who is tracking the allocation say for an obj who
has 2 placements LMEM and SMEM, and might assume since I had made a
reservation in SMEM it shall not fail when i try to migrate there later.

Thanks,
Aravind.

> 
> And then if/when the same object gets migrated to smem it changes to
> (lets assume it is also not resident any more but got swapped out):
> 
>  total-smem: 1 MiB
>  resident-smem: 0
>  total-lmem: 0
>  resident-lmem: 0
> 
> Regards,
> 
> Tvrtko
> 
>>> +        if (obj->base.handle_count > 1)
>>> +            stats[id].shared += sz;
>>> +        else
>>> +            stats[id].private += sz;
>>> +    }
>>> +
>>> +    if (i915_gem_object_has_pages(obj)) {
>>> +        stats[id].resident += sz;
>>> +
>>> +        if (!dma_resv_test_signaled(obj->base.resv,
>>> +                        dma_resv_usage_rw(true)))
>>> +            stats[id].active += sz;
>>> +        else if (i915_gem_object_is_shrinkable(obj) &&
>>> +             obj->mm.madv == I915_MADV_DONTNEED)
>>> +            stats[id].purgeable += sz;
>>> +    }
>>> +}
>>> +
>>> +static void show_meminfo(struct drm_printer *p, struct drm_file *file)
>>> +{
>>> +    struct drm_memory_stats stats[INTEL_REGION_UNKNOWN] = {};
>>> +    struct drm_i915_file_private *fpriv = file->driver_priv;
>>> +    struct i915_drm_client *client = fpriv->client;
>>> +    struct drm_i915_private *i915 = fpriv->i915;
>>> +    struct drm_i915_gem_object *obj;
>>> +    struct intel_memory_region *mr;
>>> +    struct list_head *pos;
>>> +    unsigned int id;
>>> +
>>> +    /* Public objects. */
>>> +    spin_lock(&file->table_lock);
>>> +    idr_for_each_entry(&file->object_idr, obj, id)
>>> +        obj_meminfo(obj, stats);
>>> +    spin_unlock(&file->table_lock);
>>> +
>>> +    /* Internal objects. */
>>> +    rcu_read_lock();
>>> +    list_for_each_rcu(pos, &client->objects_list) {
>>> +        obj = i915_gem_object_get_rcu(list_entry(pos, typeof(*obj),
>>> +                             client_link));
>>> +        if (!obj)
>>> +            continue;
>>> +        obj_meminfo(obj, stats);
>>> +        i915_gem_object_put(obj);
>>> +    }
>>> +    rcu_read_unlock();
>>> +
>>> +    for_each_memory_region(mr, i915, id)
>>> +        drm_print_memory_stats(p,
>>> +                       &stats[id],
>>> +                       DRM_GEM_OBJECT_RESIDENT |
>>> +                       DRM_GEM_OBJECT_PURGEABLE,
>>> +                       mr->name);
>>> +}
>>> +
>>>   static const char * const uabi_class_names[] = {
>>>       [I915_ENGINE_CLASS_RENDER] = "render",
>>>       [I915_ENGINE_CLASS_COPY] = "copy",
>>> @@ -106,6 +189,8 @@ void i915_drm_client_fdinfo(struct drm_printer
>>> *p, struct drm_file *file)
>>>        *
>>> ******************************************************************
>>>        */
>>>   +    show_meminfo(p, file);
>>> +
>>>       if (GRAPHICS_VER(i915) < 8)
>>>           return;
>>>
>>
>>
>>

[PATCH 3/5] drm/i915: Track page table backing store usage

[Tvrtko Ursulin]

From: Tvrtko Ursulin <tvrtko.ursulin@intel.com>

Account page table backing store against the owning client memory usage
stats.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Aravind Iddamsetty <aravind.iddamsetty@intel.com>
---
 drivers/gpu/drm/i915/gt/intel_gtt.c | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/drivers/gpu/drm/i915/gt/intel_gtt.c b/drivers/gpu/drm/i915/gt/intel_gtt.c
index 13944a14ea2d..c3f29999b379 100644
--- a/drivers/gpu/drm/i915/gt/intel_gtt.c
+++ b/drivers/gpu/drm/i915/gt/intel_gtt.c
@@ -58,6 +58,9 @@ struct drm_i915_gem_object *alloc_pt_lmem(struct i915_address_space *vm, int sz)
 	if (!IS_ERR(obj)) {
 		obj->base.resv = i915_vm_resv_get(vm);
 		obj->shares_resv_from = vm;
+
+		if (vm->fpriv)
+			i915_drm_client_add_object(vm->fpriv->client, obj);
 	}
 
 	return obj;
@@ -79,6 +82,9 @@ struct drm_i915_gem_object *alloc_pt_dma(struct i915_address_space *vm, int sz)
 	if (!IS_ERR(obj)) {
 		obj->base.resv = i915_vm_resv_get(vm);
 		obj->shares_resv_from = vm;
+
+		if (vm->fpriv)
+			i915_drm_client_add_object(vm->fpriv->client, obj);
 	}
 
 	return obj;
-- 
2.39.2

[PATCH 3/5] drm/i915: Track page table backing store usage

[Tvrtko Ursulin]

From: Tvrtko Ursulin <tvrtko.ursulin@intel.com>

Account page table backing store against the owning client memory usage
stats.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
---
 drivers/gpu/drm/i915/gt/intel_gtt.c | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/drivers/gpu/drm/i915/gt/intel_gtt.c b/drivers/gpu/drm/i915/gt/intel_gtt.c
index 2f6a9be0ffe6..126269a0d728 100644
--- a/drivers/gpu/drm/i915/gt/intel_gtt.c
+++ b/drivers/gpu/drm/i915/gt/intel_gtt.c
@@ -58,6 +58,9 @@ struct drm_i915_gem_object *alloc_pt_lmem(struct i915_address_space *vm, int sz)
 	if (!IS_ERR(obj)) {
 		obj->base.resv = i915_vm_resv_get(vm);
 		obj->shares_resv_from = vm;
+
+		if (vm->fpriv)
+			i915_drm_client_add_object(vm->fpriv->client, obj);
 	}
 
 	return obj;
@@ -79,6 +82,9 @@ struct drm_i915_gem_object *alloc_pt_dma(struct i915_address_space *vm, int sz)
 	if (!IS_ERR(obj)) {
 		obj->base.resv = i915_vm_resv_get(vm);
 		obj->shares_resv_from = vm;
+
+		if (vm->fpriv)
+			i915_drm_client_add_object(vm->fpriv->client, obj);
 	}
 
 	return obj;
-- 
2.39.2

[PATCH 3/5] drm/i915: Track page table backing store usage

[Tvrtko Ursulin]

From: Tvrtko Ursulin <tvrtko.ursulin@intel.com>

Account page table backing store against the owning client memory usage
stats.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
---
 drivers/gpu/drm/i915/gt/intel_gtt.c | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/drivers/gpu/drm/i915/gt/intel_gtt.c b/drivers/gpu/drm/i915/gt/intel_gtt.c
index 2f6a9be0ffe6..126269a0d728 100644
--- a/drivers/gpu/drm/i915/gt/intel_gtt.c
+++ b/drivers/gpu/drm/i915/gt/intel_gtt.c
@@ -58,6 +58,9 @@ struct drm_i915_gem_object *alloc_pt_lmem(struct i915_address_space *vm, int sz)
 	if (!IS_ERR(obj)) {
 		obj->base.resv = i915_vm_resv_get(vm);
 		obj->shares_resv_from = vm;
+
+		if (vm->fpriv)
+			i915_drm_client_add_object(vm->fpriv->client, obj);
 	}
 
 	return obj;
@@ -79,6 +82,9 @@ struct drm_i915_gem_object *alloc_pt_dma(struct i915_address_space *vm, int sz)
 	if (!IS_ERR(obj)) {
 		obj->base.resv = i915_vm_resv_get(vm);
 		obj->shares_resv_from = vm;
+
+		if (vm->fpriv)
+			i915_drm_client_add_object(vm->fpriv->client, obj);
 	}
 
 	return obj;
-- 
2.39.2

[PATCH 3/5] drm/i915: Track page table backing store usage

[Tvrtko Ursulin]

From: Tvrtko Ursulin <tvrtko.ursulin@intel.com>

Account page table backing store against the owning client memory usage
stats.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
---
 drivers/gpu/drm/i915/gt/intel_gtt.c | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/drivers/gpu/drm/i915/gt/intel_gtt.c b/drivers/gpu/drm/i915/gt/intel_gtt.c
index 2f6a9be0ffe6..126269a0d728 100644
--- a/drivers/gpu/drm/i915/gt/intel_gtt.c
+++ b/drivers/gpu/drm/i915/gt/intel_gtt.c
@@ -58,6 +58,9 @@ struct drm_i915_gem_object *alloc_pt_lmem(struct i915_address_space *vm, int sz)
 	if (!IS_ERR(obj)) {
 		obj->base.resv = i915_vm_resv_get(vm);
 		obj->shares_resv_from = vm;
+
+		if (vm->fpriv)
+			i915_drm_client_add_object(vm->fpriv->client, obj);
 	}
 
 	return obj;
@@ -79,6 +82,9 @@ struct drm_i915_gem_object *alloc_pt_dma(struct i915_address_space *vm, int sz)
 	if (!IS_ERR(obj)) {
 		obj->base.resv = i915_vm_resv_get(vm);
 		obj->shares_resv_from = vm;
+
+		if (vm->fpriv)
+			i915_drm_client_add_object(vm->fpriv->client, obj);
 	}
 
 	return obj;
-- 
2.39.2