[PATCH 1/4] drm/ttm: add debugfs directory v2

[PATCH 1/4] drm/ttm: add debugfs directory v2

[Christian König]

As far as I can tell the buffer_count was never used by an
userspace application.

The number of BOs in the system is far better suited in
debugfs than sysfs and we now should be able to add other
information here as well.

v2: add that additionally to sysfs

Signed-off-by: Christian König <christian.koenig@amd.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
---
 drivers/gpu/drm/ttm/ttm_bo.c     | 48 ++------------------------------
 drivers/gpu/drm/ttm/ttm_module.c |  4 +++
 drivers/gpu/drm/ttm/ttm_module.h |  6 ++--
 3 files changed, 11 insertions(+), 47 deletions(-)

diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c
index b65f4b12f986..c289a6a37ff9 100644
--- a/drivers/gpu/drm/ttm/ttm_bo.c
+++ b/drivers/gpu/drm/ttm/ttm_bo.c
@@ -44,8 +44,6 @@
 
 #include "ttm_module.h"
 
-static void ttm_bo_global_kobj_release(struct kobject *kobj);
-
 /*
  * ttm_global_mutex - protecting the global BO state
  */
@@ -54,11 +52,6 @@ unsigned ttm_bo_glob_use_count;
 struct ttm_bo_global ttm_bo_glob;
 EXPORT_SYMBOL(ttm_bo_glob);
 
-static struct attribute ttm_bo_count = {
-	.name = "bo_count",
-	.mode = S_IRUGO
-};
-
 /* default destructor */
 static void ttm_bo_default_destroy(struct ttm_buffer_object *bo)
 {
@@ -84,32 +77,6 @@ static void ttm_bo_mem_space_debug(struct ttm_buffer_object *bo,
 	}
 }
 
-static ssize_t ttm_bo_global_show(struct kobject *kobj,
-				  struct attribute *attr,
-				  char *buffer)
-{
-	struct ttm_bo_global *glob =
-		container_of(kobj, struct ttm_bo_global, kobj);
-
-	return snprintf(buffer, PAGE_SIZE, "%d\n",
-				atomic_read(&glob->bo_count));
-}
-
-static struct attribute *ttm_bo_global_attrs[] = {
-	&ttm_bo_count,
-	NULL
-};
-
-static const struct sysfs_ops ttm_bo_global_ops = {
-	.show = &ttm_bo_global_show
-};
-
-static struct kobj_type ttm_bo_glob_kobj_type  = {
-	.release = &ttm_bo_global_kobj_release,
-	.sysfs_ops = &ttm_bo_global_ops,
-	.default_attrs = ttm_bo_global_attrs
-};
-
 static void ttm_bo_del_from_lru(struct ttm_buffer_object *bo)
 {
 	struct ttm_bo_device *bdev = bo->bdev;
@@ -1228,14 +1195,6 @@ size_t ttm_bo_dma_acc_size(struct ttm_bo_device *bdev,
 }
 EXPORT_SYMBOL(ttm_bo_dma_acc_size);
 
-static void ttm_bo_global_kobj_release(struct kobject *kobj)
-{
-	struct ttm_bo_global *glob =
-		container_of(kobj, struct ttm_bo_global, kobj);
-
-	__free_page(glob->dummy_read_page);
-}
-
 static void ttm_bo_global_release(void)
 {
 	struct ttm_bo_global *glob = &ttm_bo_glob;
@@ -1247,6 +1206,7 @@ static void ttm_bo_global_release(void)
 	kobject_del(&glob->kobj);
 	kobject_put(&glob->kobj);
 	ttm_mem_global_release(&ttm_mem_glob);
+	__free_page(glob->dummy_read_page);
 	memset(glob, 0, sizeof(*glob));
 out:
 	mutex_unlock(&ttm_global_mutex);
@@ -1279,10 +1239,8 @@ static int ttm_bo_global_init(void)
 	INIT_LIST_HEAD(&glob->device_list);
 	atomic_set(&glob->bo_count, 0);
 
-	ret = kobject_init_and_add(
-		&glob->kobj, &ttm_bo_glob_kobj_type, ttm_get_kobj(), "buffer_objects");
-	if (unlikely(ret != 0))
-		kobject_put(&glob->kobj);
+	debugfs_create_atomic_t("buffer_objects", 0444, ttm_debugfs_root,
+				&glob->bo_count);
 out:
 	mutex_unlock(&ttm_global_mutex);
 	return ret;
diff --git a/drivers/gpu/drm/ttm/ttm_module.c b/drivers/gpu/drm/ttm/ttm_module.c
index c0906437cb1c..f6566603a60f 100644
--- a/drivers/gpu/drm/ttm/ttm_module.c
+++ b/drivers/gpu/drm/ttm/ttm_module.c
@@ -32,12 +32,14 @@
 #include <linux/module.h>
 #include <linux/device.h>
 #include <linux/sched.h>
+#include <linux/debugfs.h>
 #include <drm/drm_sysfs.h>
 
 #include "ttm_module.h"
 
 static DECLARE_WAIT_QUEUE_HEAD(exit_q);
 static atomic_t device_released;
+struct dentry *ttm_debugfs_root;
 
 static struct device_type ttm_drm_class_type = {
 	.name = "ttm",
@@ -77,6 +79,7 @@ static int __init ttm_init(void)
 	if (unlikely(ret != 0))
 		goto out_no_dev_reg;
 
+	ttm_debugfs_root = debugfs_create_dir("ttm", NULL);
 	return 0;
 out_no_dev_reg:
 	atomic_set(&device_released, 1);
@@ -94,6 +97,7 @@ static void __exit ttm_exit(void)
 	 */
 
 	wait_event(exit_q, atomic_read(&device_released) == 1);
+	debugfs_remove(ttm_debugfs_root);
 }
 
 module_init(ttm_init);
diff --git a/drivers/gpu/drm/ttm/ttm_module.h b/drivers/gpu/drm/ttm/ttm_module.h
index 45fa318c1585..2f03c2fcf570 100644
--- a/drivers/gpu/drm/ttm/ttm_module.h
+++ b/drivers/gpu/drm/ttm/ttm_module.h
@@ -31,10 +31,12 @@
 #ifndef _TTM_MODULE_H_
 #define _TTM_MODULE_H_
 
-#include <linux/kernel.h>
+#define TTM_PFX "[TTM] "
+
 struct kobject;
+struct dentry;
 
-#define TTM_PFX "[TTM] "
 extern struct kobject *ttm_get_kobj(void);
+extern struct dentry *ttm_debugfs_root;
 
 #endif /* _TTM_MODULE_H_ */
-- 
2.25.1

_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel

[PATCH 2/4] drm/ttm: add a debugfs file for the global page pools

[Christian König]

Instead of printing this on the per device pool.

Signed-off-by: Christian König <christian.koenig@amd.com>
---
 drivers/gpu/drm/ttm/ttm_pool.c | 70 ++++++++++++++++++++++++----------
 1 file changed, 50 insertions(+), 20 deletions(-)

diff --git a/drivers/gpu/drm/ttm/ttm_pool.c b/drivers/gpu/drm/ttm/ttm_pool.c
index 7b2f60616750..1d61e8fc0e81 100644
--- a/drivers/gpu/drm/ttm/ttm_pool.c
+++ b/drivers/gpu/drm/ttm/ttm_pool.c
@@ -42,6 +42,8 @@
 #include <drm/ttm/ttm_bo_driver.h>
 #include <drm/ttm/ttm_tt.h>
 
+#include "ttm_module.h"
+
 /**
  * struct ttm_pool_dma - Helper object for coherent DMA mappings
  *
@@ -543,6 +545,17 @@ static unsigned int ttm_pool_type_count(struct ttm_pool_type *pt)
 	return count;
 }
 
+/* Print a nice header for the order */
+static void ttm_pool_debugfs_header(struct seq_file *m)
+{
+	unsigned int i;
+
+	seq_puts(m, "\t ");
+	for (i = 0; i < MAX_ORDER; ++i)
+		seq_printf(m, " ---%2u---", i);
+	seq_puts(m, "\n");
+}
+
 /* Dump information about the different pool types */
 static void ttm_pool_debugfs_orders(struct ttm_pool_type *pt,
 				    struct seq_file *m)
@@ -554,6 +567,35 @@ static void ttm_pool_debugfs_orders(struct ttm_pool_type *pt,
 	seq_puts(m, "\n");
 }
 
+/* Dump the total amount of allocated pages */
+static void ttm_pool_debugfs_footer(struct seq_file *m)
+{
+	seq_printf(m, "\ntotal\t: %8lu of %8lu\n",
+		   atomic_long_read(&allocated_pages), page_pool_size);
+}
+
+/* Dump the information for the global pools */
+static int ttm_pool_debugfs_globals_show(struct seq_file *m, void *data)
+{
+	ttm_pool_debugfs_header(m);
+
+	spin_lock(&shrinker_lock);
+	seq_puts(m, "wc\t:");
+	ttm_pool_debugfs_orders(global_write_combined, m);
+	seq_puts(m, "uc\t:");
+	ttm_pool_debugfs_orders(global_uncached, m);
+	seq_puts(m, "wc 32\t:");
+	ttm_pool_debugfs_orders(global_dma32_write_combined, m);
+	seq_puts(m, "uc 32\t:");
+	ttm_pool_debugfs_orders(global_dma32_uncached, m);
+	spin_unlock(&shrinker_lock);
+
+	ttm_pool_debugfs_footer(m);
+
+	return 0;
+}
+DEFINE_SHOW_ATTRIBUTE(ttm_pool_debugfs_globals);
+
 /**
  * ttm_pool_debugfs - Debugfs dump function for a pool
  *
@@ -566,23 +608,9 @@ int ttm_pool_debugfs(struct ttm_pool *pool, struct seq_file *m)
 {
 	unsigned int i;
 
-	spin_lock(&shrinker_lock);
-
-	seq_puts(m, "\t ");
-	for (i = 0; i < MAX_ORDER; ++i)
-		seq_printf(m, " ---%2u---", i);
-	seq_puts(m, "\n");
-
-	seq_puts(m, "wc\t:");
-	ttm_pool_debugfs_orders(global_write_combined, m);
-	seq_puts(m, "uc\t:");
-	ttm_pool_debugfs_orders(global_uncached, m);
-
-	seq_puts(m, "wc 32\t:");
-	ttm_pool_debugfs_orders(global_dma32_write_combined, m);
-	seq_puts(m, "uc 32\t:");
-	ttm_pool_debugfs_orders(global_dma32_uncached, m);
+	ttm_pool_debugfs_header(m);
 
+	spin_lock(&shrinker_lock);
 	for (i = 0; i < TTM_NUM_CACHING_TYPES; ++i) {
 		seq_puts(m, "DMA ");
 		switch (i) {
@@ -598,12 +626,9 @@ int ttm_pool_debugfs(struct ttm_pool *pool, struct seq_file *m)
 		}
 		ttm_pool_debugfs_orders(pool->caching[i].orders, m);
 	}
-
-	seq_printf(m, "\ntotal\t: %8lu of %8lu\n",
-		   atomic_long_read(&allocated_pages), page_pool_size);
-
 	spin_unlock(&shrinker_lock);
 
+	ttm_pool_debugfs_footer(m);
 	return 0;
 }
 EXPORT_SYMBOL(ttm_pool_debugfs);
@@ -660,6 +685,11 @@ int ttm_pool_mgr_init(unsigned long num_pages)
 				   ttm_uncached, i);
 	}
 
+#ifdef CONFIG_DEBUG_FS
+	debugfs_create_file("page_pool", 0444, ttm_debugfs_root, NULL,
+			    &ttm_pool_debugfs_globals_fops);
+#endif
+
 	mm_shrinker.count_objects = ttm_pool_shrinker_count;
 	mm_shrinker.scan_objects = ttm_pool_shrinker_scan;
 	mm_shrinker.seeks = 1;
-- 
2.25.1

_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel

[PATCH 3/4] drm/ttm: add debugfs entry to test pool shrinker v2

[Christian König]

Useful for testing.

v2: add fs_reclaim_acquire()/_release()

Signed-off-by: Christian König <christian.koenig@amd.com>
---
 drivers/gpu/drm/ttm/ttm_pool.c | 53 ++++++++++++++++++++++------------
 1 file changed, 35 insertions(+), 18 deletions(-)

diff --git a/drivers/gpu/drm/ttm/ttm_pool.c b/drivers/gpu/drm/ttm/ttm_pool.c
index 1d61e8fc0e81..98ecb9c9842c 100644
--- a/drivers/gpu/drm/ttm/ttm_pool.c
+++ b/drivers/gpu/drm/ttm/ttm_pool.c
@@ -33,6 +33,7 @@
 
 #include <linux/module.h>
 #include <linux/dma-mapping.h>
+#include <linux/sched/mm.h>
 
 #ifdef CONFIG_X86
 #include <asm/set_memory.h>
@@ -529,6 +530,28 @@ void ttm_pool_fini(struct ttm_pool *pool)
 }
 EXPORT_SYMBOL(ttm_pool_fini);
 
+/* As long as pages are available make sure to release at least one */
+static unsigned long ttm_pool_shrinker_scan(struct shrinker *shrink,
+					    struct shrink_control *sc)
+{
+	unsigned long num_freed = 0;
+
+	do
+		num_freed += ttm_pool_shrink();
+	while (!num_freed && atomic_long_read(&allocated_pages));
+
+	return num_freed;
+}
+
+/* Return the number of pages available or SHRINK_EMPTY if we have none */
+static unsigned long ttm_pool_shrinker_count(struct shrinker *shrink,
+					     struct shrink_control *sc)
+{
+	unsigned long num_pages = atomic_long_read(&allocated_pages);
+
+	return num_pages ? num_pages : SHRINK_EMPTY;
+}
+
 #ifdef CONFIG_DEBUG_FS
 /* Count the number of pages available in a pool_type */
 static unsigned int ttm_pool_type_count(struct ttm_pool_type *pt)
@@ -633,29 +656,21 @@ int ttm_pool_debugfs(struct ttm_pool *pool, struct seq_file *m)
 }
 EXPORT_SYMBOL(ttm_pool_debugfs);
 
-#endif
-
-/* As long as pages are available make sure to release at least one */
-static unsigned long ttm_pool_shrinker_scan(struct shrinker *shrink,
-					    struct shrink_control *sc)
+/* Test the shrinker functions and dump the result */
+static int ttm_pool_debugfs_shrink_show(struct seq_file *m, void *data)
 {
-	unsigned long num_freed = 0;
+	struct shrink_control sc = { .gfp_mask = GFP_NOFS };
 
-	do
-		num_freed += ttm_pool_shrink();
-	while (!num_freed && atomic_long_read(&allocated_pages));
+	fs_reclaim_acquire(GFP_KERNEL);
+	seq_printf(m, "%lu/%lu\n", ttm_pool_shrinker_count(&mm_shrinker, &sc),
+		   ttm_pool_shrinker_scan(&mm_shrinker, &sc));
+	fs_reclaim_release(GFP_KERNEL);
 
-	return num_freed;
+	return 0;
 }
+DEFINE_SHOW_ATTRIBUTE(ttm_pool_debugfs_shrink);
 
-/* Return the number of pages available or SHRINK_EMPTY if we have none */
-static unsigned long ttm_pool_shrinker_count(struct shrinker *shrink,
-					     struct shrink_control *sc)
-{
-	unsigned long num_pages = atomic_long_read(&allocated_pages);
-
-	return num_pages ? num_pages : SHRINK_EMPTY;
-}
+#endif
 
 /**
  * ttm_pool_mgr_init - Initialize globals
@@ -688,6 +703,8 @@ int ttm_pool_mgr_init(unsigned long num_pages)
 #ifdef CONFIG_DEBUG_FS
 	debugfs_create_file("page_pool", 0444, ttm_debugfs_root, NULL,
 			    &ttm_pool_debugfs_globals_fops);
+	debugfs_create_file("page_pool_shrink", 0400, ttm_debugfs_root, NULL,
+			    &ttm_pool_debugfs_shrink_fops);
 #endif
 
 	mm_shrinker.count_objects = ttm_pool_shrinker_count;
-- 
2.25.1

_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel

[PATCH 4/4] drm/ttm: optimize ttm pool shrinker a bit

[Christian König]

Only initialize the DMA coherent pools if they are used.

Signed-off-by: Christian König <christian.koenig@amd.com>
---
 drivers/gpu/drm/ttm/ttm_pool.c | 23 ++++++++++++++++-------
 1 file changed, 16 insertions(+), 7 deletions(-)

diff --git a/drivers/gpu/drm/ttm/ttm_pool.c b/drivers/gpu/drm/ttm/ttm_pool.c
index 98ecb9c9842c..e0617717113f 100644
--- a/drivers/gpu/drm/ttm/ttm_pool.c
+++ b/drivers/gpu/drm/ttm/ttm_pool.c
@@ -505,10 +505,12 @@ void ttm_pool_init(struct ttm_pool *pool, struct device *dev,
 	pool->use_dma_alloc = use_dma_alloc;
 	pool->use_dma32 = use_dma32;
 
-	for (i = 0; i < TTM_NUM_CACHING_TYPES; ++i)
-		for (j = 0; j < MAX_ORDER; ++j)
-			ttm_pool_type_init(&pool->caching[i].orders[j],
-					   pool, i, j);
+	if (use_dma_alloc) {
+		for (i = 0; i < TTM_NUM_CACHING_TYPES; ++i)
+			for (j = 0; j < MAX_ORDER; ++j)
+				ttm_pool_type_init(&pool->caching[i].orders[j],
+						   pool, i, j);
+	}
 }
 EXPORT_SYMBOL(ttm_pool_init);
 
@@ -524,9 +526,11 @@ void ttm_pool_fini(struct ttm_pool *pool)
 {
 	unsigned int i, j;
 
-	for (i = 0; i < TTM_NUM_CACHING_TYPES; ++i)
-		for (j = 0; j < MAX_ORDER; ++j)
-			ttm_pool_type_fini(&pool->caching[i].orders[j]);
+	if (pool->use_dma_alloc) {
+		for (i = 0; i < TTM_NUM_CACHING_TYPES; ++i)
+			for (j = 0; j < MAX_ORDER; ++j)
+				ttm_pool_type_fini(&pool->caching[i].orders[j]);
+	}
 }
 EXPORT_SYMBOL(ttm_pool_fini);
 
@@ -631,6 +635,11 @@ int ttm_pool_debugfs(struct ttm_pool *pool, struct seq_file *m)
 {
 	unsigned int i;
 
+	if (!pool->use_dma_alloc) {
+		seq_puts(m, "unused\n");
+		return 0;
+	}
+
 	ttm_pool_debugfs_header(m);
 
 	spin_lock(&shrinker_lock);
-- 
2.25.1

_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel

Re: [PATCH 2/4] drm/ttm: add a debugfs file for the global page pools

[Daniel Vetter]

On Tue, Jan 19, 2021 at 01:18:19PM +0100, Christian König wrote:
> Instead of printing this on the per device pool.
> 
> Signed-off-by: Christian König <christian.koenig@amd.com>

Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>

btw I think ttm should also set up the per-bdev debugfs stuff, feels silly
having that boilerplate. Same for the per-ttm_manager thing we have I
think.
-Daniel

> ---
>  drivers/gpu/drm/ttm/ttm_pool.c | 70 ++++++++++++++++++++++++----------
>  1 file changed, 50 insertions(+), 20 deletions(-)
> 
> diff --git a/drivers/gpu/drm/ttm/ttm_pool.c b/drivers/gpu/drm/ttm/ttm_pool.c
> index 7b2f60616750..1d61e8fc0e81 100644
> --- a/drivers/gpu/drm/ttm/ttm_pool.c
> +++ b/drivers/gpu/drm/ttm/ttm_pool.c
> @@ -42,6 +42,8 @@
>  #include <drm/ttm/ttm_bo_driver.h>
>  #include <drm/ttm/ttm_tt.h>
>  
> +#include "ttm_module.h"
> +
>  /**
>   * struct ttm_pool_dma - Helper object for coherent DMA mappings
>   *
> @@ -543,6 +545,17 @@ static unsigned int ttm_pool_type_count(struct ttm_pool_type *pt)
>  	return count;
>  }
>  
> +/* Print a nice header for the order */
> +static void ttm_pool_debugfs_header(struct seq_file *m)
> +{
> +	unsigned int i;
> +
> +	seq_puts(m, "\t ");
> +	for (i = 0; i < MAX_ORDER; ++i)
> +		seq_printf(m, " ---%2u---", i);
> +	seq_puts(m, "\n");
> +}
> +
>  /* Dump information about the different pool types */
>  static void ttm_pool_debugfs_orders(struct ttm_pool_type *pt,
>  				    struct seq_file *m)
> @@ -554,6 +567,35 @@ static void ttm_pool_debugfs_orders(struct ttm_pool_type *pt,
>  	seq_puts(m, "\n");
>  }
>  
> +/* Dump the total amount of allocated pages */
> +static void ttm_pool_debugfs_footer(struct seq_file *m)
> +{
> +	seq_printf(m, "\ntotal\t: %8lu of %8lu\n",
> +		   atomic_long_read(&allocated_pages), page_pool_size);
> +}
> +
> +/* Dump the information for the global pools */
> +static int ttm_pool_debugfs_globals_show(struct seq_file *m, void *data)
> +{
> +	ttm_pool_debugfs_header(m);
> +
> +	spin_lock(&shrinker_lock);
> +	seq_puts(m, "wc\t:");
> +	ttm_pool_debugfs_orders(global_write_combined, m);
> +	seq_puts(m, "uc\t:");
> +	ttm_pool_debugfs_orders(global_uncached, m);
> +	seq_puts(m, "wc 32\t:");
> +	ttm_pool_debugfs_orders(global_dma32_write_combined, m);
> +	seq_puts(m, "uc 32\t:");
> +	ttm_pool_debugfs_orders(global_dma32_uncached, m);
> +	spin_unlock(&shrinker_lock);
> +
> +	ttm_pool_debugfs_footer(m);
> +
> +	return 0;
> +}
> +DEFINE_SHOW_ATTRIBUTE(ttm_pool_debugfs_globals);
> +
>  /**
>   * ttm_pool_debugfs - Debugfs dump function for a pool
>   *
> @@ -566,23 +608,9 @@ int ttm_pool_debugfs(struct ttm_pool *pool, struct seq_file *m)
>  {
>  	unsigned int i;
>  
> -	spin_lock(&shrinker_lock);
> -
> -	seq_puts(m, "\t ");
> -	for (i = 0; i < MAX_ORDER; ++i)
> -		seq_printf(m, " ---%2u---", i);
> -	seq_puts(m, "\n");
> -
> -	seq_puts(m, "wc\t:");
> -	ttm_pool_debugfs_orders(global_write_combined, m);
> -	seq_puts(m, "uc\t:");
> -	ttm_pool_debugfs_orders(global_uncached, m);
> -
> -	seq_puts(m, "wc 32\t:");
> -	ttm_pool_debugfs_orders(global_dma32_write_combined, m);
> -	seq_puts(m, "uc 32\t:");
> -	ttm_pool_debugfs_orders(global_dma32_uncached, m);
> +	ttm_pool_debugfs_header(m);
>  
> +	spin_lock(&shrinker_lock);
>  	for (i = 0; i < TTM_NUM_CACHING_TYPES; ++i) {
>  		seq_puts(m, "DMA ");
>  		switch (i) {
> @@ -598,12 +626,9 @@ int ttm_pool_debugfs(struct ttm_pool *pool, struct seq_file *m)
>  		}
>  		ttm_pool_debugfs_orders(pool->caching[i].orders, m);
>  	}
> -
> -	seq_printf(m, "\ntotal\t: %8lu of %8lu\n",
> -		   atomic_long_read(&allocated_pages), page_pool_size);
> -
>  	spin_unlock(&shrinker_lock);
>  
> +	ttm_pool_debugfs_footer(m);
>  	return 0;
>  }
>  EXPORT_SYMBOL(ttm_pool_debugfs);
> @@ -660,6 +685,11 @@ int ttm_pool_mgr_init(unsigned long num_pages)
>  				   ttm_uncached, i);
>  	}
>  
> +#ifdef CONFIG_DEBUG_FS
> +	debugfs_create_file("page_pool", 0444, ttm_debugfs_root, NULL,
> +			    &ttm_pool_debugfs_globals_fops);
> +#endif
> +
>  	mm_shrinker.count_objects = ttm_pool_shrinker_count;
>  	mm_shrinker.scan_objects = ttm_pool_shrinker_scan;
>  	mm_shrinker.seeks = 1;
> -- 
> 2.25.1
> 

-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel

Re: [PATCH 3/4] drm/ttm: add debugfs entry to test pool shrinker v2

[Daniel Vetter]

On Tue, Jan 19, 2021 at 01:18:20PM +0100, Christian König wrote:
> Useful for testing.
> 
> v2: add fs_reclaim_acquire()/_release()
> 
> Signed-off-by: Christian König <christian.koenig@amd.com>
> ---
>  drivers/gpu/drm/ttm/ttm_pool.c | 53 ++++++++++++++++++++++------------
>  1 file changed, 35 insertions(+), 18 deletions(-)
> 
> diff --git a/drivers/gpu/drm/ttm/ttm_pool.c b/drivers/gpu/drm/ttm/ttm_pool.c
> index 1d61e8fc0e81..98ecb9c9842c 100644
> --- a/drivers/gpu/drm/ttm/ttm_pool.c
> +++ b/drivers/gpu/drm/ttm/ttm_pool.c
> @@ -33,6 +33,7 @@
>  
>  #include <linux/module.h>
>  #include <linux/dma-mapping.h>
> +#include <linux/sched/mm.h>
>  
>  #ifdef CONFIG_X86
>  #include <asm/set_memory.h>
> @@ -529,6 +530,28 @@ void ttm_pool_fini(struct ttm_pool *pool)
>  }
>  EXPORT_SYMBOL(ttm_pool_fini);
>  
> +/* As long as pages are available make sure to release at least one */
> +static unsigned long ttm_pool_shrinker_scan(struct shrinker *shrink,
> +					    struct shrink_control *sc)
> +{
> +	unsigned long num_freed = 0;
> +
> +	do
> +		num_freed += ttm_pool_shrink();
> +	while (!num_freed && atomic_long_read(&allocated_pages));
> +
> +	return num_freed;
> +}
> +
> +/* Return the number of pages available or SHRINK_EMPTY if we have none */
> +static unsigned long ttm_pool_shrinker_count(struct shrinker *shrink,
> +					     struct shrink_control *sc)
> +{
> +	unsigned long num_pages = atomic_long_read(&allocated_pages);
> +
> +	return num_pages ? num_pages : SHRINK_EMPTY;
> +}
> +
>  #ifdef CONFIG_DEBUG_FS
>  /* Count the number of pages available in a pool_type */
>  static unsigned int ttm_pool_type_count(struct ttm_pool_type *pt)
> @@ -633,29 +656,21 @@ int ttm_pool_debugfs(struct ttm_pool *pool, struct seq_file *m)
>  }
>  EXPORT_SYMBOL(ttm_pool_debugfs);
>  
> -#endif
> -
> -/* As long as pages are available make sure to release at least one */
> -static unsigned long ttm_pool_shrinker_scan(struct shrinker *shrink,
> -					    struct shrink_control *sc)
> +/* Test the shrinker functions and dump the result */
> +static int ttm_pool_debugfs_shrink_show(struct seq_file *m, void *data)
>  {
> -	unsigned long num_freed = 0;
> +	struct shrink_control sc = { .gfp_mask = GFP_NOFS };
>  
> -	do
> -		num_freed += ttm_pool_shrink();
> -	while (!num_freed && atomic_long_read(&allocated_pages));
> +	fs_reclaim_acquire(GFP_KERNEL);
> +	seq_printf(m, "%lu/%lu\n", ttm_pool_shrinker_count(&mm_shrinker, &sc),
> +		   ttm_pool_shrinker_scan(&mm_shrinker, &sc));
> +	fs_reclaim_release(GFP_KERNEL);
>  
> -	return num_freed;
> +	return 0;
>  }
> +DEFINE_SHOW_ATTRIBUTE(ttm_pool_debugfs_shrink);

Shrinking everything is a bit a hammer, but probably the right size we
need for testing :-)

Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>

>  
> -/* Return the number of pages available or SHRINK_EMPTY if we have none */
> -static unsigned long ttm_pool_shrinker_count(struct shrinker *shrink,
> -					     struct shrink_control *sc)
> -{
> -	unsigned long num_pages = atomic_long_read(&allocated_pages);
> -
> -	return num_pages ? num_pages : SHRINK_EMPTY;
> -}
> +#endif
>  
>  /**
>   * ttm_pool_mgr_init - Initialize globals
> @@ -688,6 +703,8 @@ int ttm_pool_mgr_init(unsigned long num_pages)
>  #ifdef CONFIG_DEBUG_FS
>  	debugfs_create_file("page_pool", 0444, ttm_debugfs_root, NULL,
>  			    &ttm_pool_debugfs_globals_fops);
> +	debugfs_create_file("page_pool_shrink", 0400, ttm_debugfs_root, NULL,
> +			    &ttm_pool_debugfs_shrink_fops);
>  #endif
>  
>  	mm_shrinker.count_objects = ttm_pool_shrinker_count;
> -- 
> 2.25.1
> 

-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel

Re: [PATCH 4/4] drm/ttm: optimize ttm pool shrinker a bit

[Daniel Vetter]

On Tue, Jan 19, 2021 at 01:18:21PM +0100, Christian König wrote:
> Only initialize the DMA coherent pools if they are used.
> 
> Signed-off-by: Christian König <christian.koenig@amd.com>

Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>

> ---
>  drivers/gpu/drm/ttm/ttm_pool.c | 23 ++++++++++++++++-------
>  1 file changed, 16 insertions(+), 7 deletions(-)
> 
> diff --git a/drivers/gpu/drm/ttm/ttm_pool.c b/drivers/gpu/drm/ttm/ttm_pool.c
> index 98ecb9c9842c..e0617717113f 100644
> --- a/drivers/gpu/drm/ttm/ttm_pool.c
> +++ b/drivers/gpu/drm/ttm/ttm_pool.c
> @@ -505,10 +505,12 @@ void ttm_pool_init(struct ttm_pool *pool, struct device *dev,
>  	pool->use_dma_alloc = use_dma_alloc;
>  	pool->use_dma32 = use_dma32;
>  
> -	for (i = 0; i < TTM_NUM_CACHING_TYPES; ++i)
> -		for (j = 0; j < MAX_ORDER; ++j)
> -			ttm_pool_type_init(&pool->caching[i].orders[j],
> -					   pool, i, j);
> +	if (use_dma_alloc) {
> +		for (i = 0; i < TTM_NUM_CACHING_TYPES; ++i)
> +			for (j = 0; j < MAX_ORDER; ++j)
> +				ttm_pool_type_init(&pool->caching[i].orders[j],
> +						   pool, i, j);
> +	}
>  }
>  EXPORT_SYMBOL(ttm_pool_init);
>  
> @@ -524,9 +526,11 @@ void ttm_pool_fini(struct ttm_pool *pool)
>  {
>  	unsigned int i, j;
>  
> -	for (i = 0; i < TTM_NUM_CACHING_TYPES; ++i)
> -		for (j = 0; j < MAX_ORDER; ++j)
> -			ttm_pool_type_fini(&pool->caching[i].orders[j]);
> +	if (pool->use_dma_alloc) {
> +		for (i = 0; i < TTM_NUM_CACHING_TYPES; ++i)
> +			for (j = 0; j < MAX_ORDER; ++j)
> +				ttm_pool_type_fini(&pool->caching[i].orders[j]);
> +	}
>  }
>  EXPORT_SYMBOL(ttm_pool_fini);
>  
> @@ -631,6 +635,11 @@ int ttm_pool_debugfs(struct ttm_pool *pool, struct seq_file *m)
>  {
>  	unsigned int i;
>  
> +	if (!pool->use_dma_alloc) {
> +		seq_puts(m, "unused\n");
> +		return 0;
> +	}
> +
>  	ttm_pool_debugfs_header(m);
>  
>  	spin_lock(&shrinker_lock);
> -- 
> 2.25.1
> 

-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel

Re: [PATCH 2/4] drm/ttm: add a debugfs file for the global page pools

[Daniel Vetter]

On Fri, Dec 18, 2020 at 06:55:36PM +0100, Christian König wrote:
> Instead of printing this on the per device pool.
> 
> Signed-off-by: Christian König <christian.koenig@amd.com>

Hm shouldn't we remove this from radeon/amdgpu too and drop the
EXPORT_SYMBOL for ttm_pool_debugfs? I'm kinda confused why we still have
both ...
-Daniel


> ---
>  drivers/gpu/drm/ttm/ttm_pool.c | 70 ++++++++++++++++++++++++----------
>  1 file changed, 50 insertions(+), 20 deletions(-)
> 
> diff --git a/drivers/gpu/drm/ttm/ttm_pool.c b/drivers/gpu/drm/ttm/ttm_pool.c
> index 7b2f60616750..1d61e8fc0e81 100644
> --- a/drivers/gpu/drm/ttm/ttm_pool.c
> +++ b/drivers/gpu/drm/ttm/ttm_pool.c
> @@ -42,6 +42,8 @@
>  #include <drm/ttm/ttm_bo_driver.h>
>  #include <drm/ttm/ttm_tt.h>
>  
> +#include "ttm_module.h"
> +
>  /**
>   * struct ttm_pool_dma - Helper object for coherent DMA mappings
>   *
> @@ -543,6 +545,17 @@ static unsigned int ttm_pool_type_count(struct ttm_pool_type *pt)
>  	return count;
>  }
>  
> +/* Print a nice header for the order */
> +static void ttm_pool_debugfs_header(struct seq_file *m)
> +{
> +	unsigned int i;
> +
> +	seq_puts(m, "\t ");
> +	for (i = 0; i < MAX_ORDER; ++i)
> +		seq_printf(m, " ---%2u---", i);
> +	seq_puts(m, "\n");
> +}
> +
>  /* Dump information about the different pool types */
>  static void ttm_pool_debugfs_orders(struct ttm_pool_type *pt,
>  				    struct seq_file *m)
> @@ -554,6 +567,35 @@ static void ttm_pool_debugfs_orders(struct ttm_pool_type *pt,
>  	seq_puts(m, "\n");
>  }
>  
> +/* Dump the total amount of allocated pages */
> +static void ttm_pool_debugfs_footer(struct seq_file *m)
> +{
> +	seq_printf(m, "\ntotal\t: %8lu of %8lu\n",
> +		   atomic_long_read(&allocated_pages), page_pool_size);
> +}
> +
> +/* Dump the information for the global pools */
> +static int ttm_pool_debugfs_globals_show(struct seq_file *m, void *data)
> +{
> +	ttm_pool_debugfs_header(m);
> +
> +	spin_lock(&shrinker_lock);
> +	seq_puts(m, "wc\t:");
> +	ttm_pool_debugfs_orders(global_write_combined, m);
> +	seq_puts(m, "uc\t:");
> +	ttm_pool_debugfs_orders(global_uncached, m);
> +	seq_puts(m, "wc 32\t:");
> +	ttm_pool_debugfs_orders(global_dma32_write_combined, m);
> +	seq_puts(m, "uc 32\t:");
> +	ttm_pool_debugfs_orders(global_dma32_uncached, m);
> +	spin_unlock(&shrinker_lock);
> +
> +	ttm_pool_debugfs_footer(m);
> +
> +	return 0;
> +}
> +DEFINE_SHOW_ATTRIBUTE(ttm_pool_debugfs_globals);
> +
>  /**
>   * ttm_pool_debugfs - Debugfs dump function for a pool
>   *
> @@ -566,23 +608,9 @@ int ttm_pool_debugfs(struct ttm_pool *pool, struct seq_file *m)
>  {
>  	unsigned int i;
>  
> -	spin_lock(&shrinker_lock);
> -
> -	seq_puts(m, "\t ");
> -	for (i = 0; i < MAX_ORDER; ++i)
> -		seq_printf(m, " ---%2u---", i);
> -	seq_puts(m, "\n");
> -
> -	seq_puts(m, "wc\t:");
> -	ttm_pool_debugfs_orders(global_write_combined, m);
> -	seq_puts(m, "uc\t:");
> -	ttm_pool_debugfs_orders(global_uncached, m);
> -
> -	seq_puts(m, "wc 32\t:");
> -	ttm_pool_debugfs_orders(global_dma32_write_combined, m);
> -	seq_puts(m, "uc 32\t:");
> -	ttm_pool_debugfs_orders(global_dma32_uncached, m);
> +	ttm_pool_debugfs_header(m);
>  
> +	spin_lock(&shrinker_lock);
>  	for (i = 0; i < TTM_NUM_CACHING_TYPES; ++i) {
>  		seq_puts(m, "DMA ");
>  		switch (i) {
> @@ -598,12 +626,9 @@ int ttm_pool_debugfs(struct ttm_pool *pool, struct seq_file *m)
>  		}
>  		ttm_pool_debugfs_orders(pool->caching[i].orders, m);
>  	}
> -
> -	seq_printf(m, "\ntotal\t: %8lu of %8lu\n",
> -		   atomic_long_read(&allocated_pages), page_pool_size);
> -
>  	spin_unlock(&shrinker_lock);
>  
> +	ttm_pool_debugfs_footer(m);
>  	return 0;
>  }
>  EXPORT_SYMBOL(ttm_pool_debugfs);
> @@ -660,6 +685,11 @@ int ttm_pool_mgr_init(unsigned long num_pages)
>  				   ttm_uncached, i);
>  	}
>  
> +#ifdef CONFIG_DEBUG_FS
> +	debugfs_create_file("page_pool", 0444, ttm_debugfs_root, NULL,
> +			    &ttm_pool_debugfs_globals_fops);
> +#endif
> +
>  	mm_shrinker.count_objects = ttm_pool_shrinker_count;
>  	mm_shrinker.scan_objects = ttm_pool_shrinker_scan;
>  	mm_shrinker.seeks = 1;
> -- 
> 2.25.1
> 

-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel

[PATCH 2/4] drm/ttm: add a debugfs file for the global page pools

[Christian König]

Instead of printing this on the per device pool.

Signed-off-by: Christian König <christian.koenig@amd.com>
---
 drivers/gpu/drm/ttm/ttm_pool.c | 70 ++++++++++++++++++++++++----------
 1 file changed, 50 insertions(+), 20 deletions(-)

diff --git a/drivers/gpu/drm/ttm/ttm_pool.c b/drivers/gpu/drm/ttm/ttm_pool.c
index 7b2f60616750..1d61e8fc0e81 100644
--- a/drivers/gpu/drm/ttm/ttm_pool.c
+++ b/drivers/gpu/drm/ttm/ttm_pool.c
@@ -42,6 +42,8 @@
 #include <drm/ttm/ttm_bo_driver.h>
 #include <drm/ttm/ttm_tt.h>
 
+#include "ttm_module.h"
+
 /**
  * struct ttm_pool_dma - Helper object for coherent DMA mappings
  *
@@ -543,6 +545,17 @@ static unsigned int ttm_pool_type_count(struct ttm_pool_type *pt)
 	return count;
 }
 
+/* Print a nice header for the order */
+static void ttm_pool_debugfs_header(struct seq_file *m)
+{
+	unsigned int i;
+
+	seq_puts(m, "\t ");
+	for (i = 0; i < MAX_ORDER; ++i)
+		seq_printf(m, " ---%2u---", i);
+	seq_puts(m, "\n");
+}
+
 /* Dump information about the different pool types */
 static void ttm_pool_debugfs_orders(struct ttm_pool_type *pt,
 				    struct seq_file *m)
@@ -554,6 +567,35 @@ static void ttm_pool_debugfs_orders(struct ttm_pool_type *pt,
 	seq_puts(m, "\n");
 }
 
+/* Dump the total amount of allocated pages */
+static void ttm_pool_debugfs_footer(struct seq_file *m)
+{
+	seq_printf(m, "\ntotal\t: %8lu of %8lu\n",
+		   atomic_long_read(&allocated_pages), page_pool_size);
+}
+
+/* Dump the information for the global pools */
+static int ttm_pool_debugfs_globals_show(struct seq_file *m, void *data)
+{
+	ttm_pool_debugfs_header(m);
+
+	spin_lock(&shrinker_lock);
+	seq_puts(m, "wc\t:");
+	ttm_pool_debugfs_orders(global_write_combined, m);
+	seq_puts(m, "uc\t:");
+	ttm_pool_debugfs_orders(global_uncached, m);
+	seq_puts(m, "wc 32\t:");
+	ttm_pool_debugfs_orders(global_dma32_write_combined, m);
+	seq_puts(m, "uc 32\t:");
+	ttm_pool_debugfs_orders(global_dma32_uncached, m);
+	spin_unlock(&shrinker_lock);
+
+	ttm_pool_debugfs_footer(m);
+
+	return 0;
+}
+DEFINE_SHOW_ATTRIBUTE(ttm_pool_debugfs_globals);
+
 /**
  * ttm_pool_debugfs - Debugfs dump function for a pool
  *
@@ -566,23 +608,9 @@ int ttm_pool_debugfs(struct ttm_pool *pool, struct seq_file *m)
 {
 	unsigned int i;
 
-	spin_lock(&shrinker_lock);
-
-	seq_puts(m, "\t ");
-	for (i = 0; i < MAX_ORDER; ++i)
-		seq_printf(m, " ---%2u---", i);
-	seq_puts(m, "\n");
-
-	seq_puts(m, "wc\t:");
-	ttm_pool_debugfs_orders(global_write_combined, m);
-	seq_puts(m, "uc\t:");
-	ttm_pool_debugfs_orders(global_uncached, m);
-
-	seq_puts(m, "wc 32\t:");
-	ttm_pool_debugfs_orders(global_dma32_write_combined, m);
-	seq_puts(m, "uc 32\t:");
-	ttm_pool_debugfs_orders(global_dma32_uncached, m);
+	ttm_pool_debugfs_header(m);
 
+	spin_lock(&shrinker_lock);
 	for (i = 0; i < TTM_NUM_CACHING_TYPES; ++i) {
 		seq_puts(m, "DMA ");
 		switch (i) {
@@ -598,12 +626,9 @@ int ttm_pool_debugfs(struct ttm_pool *pool, struct seq_file *m)
 		}
 		ttm_pool_debugfs_orders(pool->caching[i].orders, m);
 	}
-
-	seq_printf(m, "\ntotal\t: %8lu of %8lu\n",
-		   atomic_long_read(&allocated_pages), page_pool_size);
-
 	spin_unlock(&shrinker_lock);
 
+	ttm_pool_debugfs_footer(m);
 	return 0;
 }
 EXPORT_SYMBOL(ttm_pool_debugfs);
@@ -660,6 +685,11 @@ int ttm_pool_mgr_init(unsigned long num_pages)
 				   ttm_uncached, i);
 	}
 
+#ifdef CONFIG_DEBUG_FS
+	debugfs_create_file("page_pool", 0444, ttm_debugfs_root, NULL,
+			    &ttm_pool_debugfs_globals_fops);
+#endif
+
 	mm_shrinker.count_objects = ttm_pool_shrinker_count;
 	mm_shrinker.scan_objects = ttm_pool_shrinker_scan;
 	mm_shrinker.seeks = 1;
-- 
2.25.1

_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel