[PATCH 1/4] drm/i915: export error state to string conversion

[PATCH 1/4] drm/i915: export error state to string conversion

[Mika Kuoppala]

In preparation for accessing error state from sysfs, export
error state to string conversion function. Also tuck buffer
error handling inside the function.

Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com>
---
 drivers/gpu/drm/i915/i915_debugfs.c |   24 ++++++++----------------
 drivers/gpu/drm/i915/i915_drv.h     |    7 +++++++
 2 files changed, 15 insertions(+), 16 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
index 0e7e3c0..77bfd51 100644
--- a/drivers/gpu/drm/i915/i915_debugfs.c
+++ b/drivers/gpu/drm/i915/i915_debugfs.c
@@ -739,15 +739,8 @@ static void i915_ring_error_state(struct drm_i915_error_state_buf *m,
 	err_printf(m, "  ring->tail: 0x%08x\n", error->cpu_ring_tail[ring]);
 }
 
-struct i915_error_state_file_priv {
-	struct drm_device *dev;
-	struct drm_i915_error_state *error;
-};
-
-
-static int i915_error_state(struct i915_error_state_file_priv *error_priv,
-			    struct drm_i915_error_state_buf *m)
-
+int i915_error_state_to_str(struct drm_i915_error_state_buf *m,
+			    const struct i915_error_state_file_priv *error_priv)
 {
 	struct drm_device *dev = error_priv->dev;
 	drm_i915_private_t *dev_priv = dev->dev_private;
@@ -757,7 +750,7 @@ static int i915_error_state(struct i915_error_state_file_priv *error_priv,
 
 	if (!error) {
 		err_printf(m, "no error state collected\n");
-		return 0;
+		goto out;
 	}
 
 	err_printf(m, "Time: %ld s %ld us\n", error->time.tv_sec,
@@ -867,6 +860,10 @@ static int i915_error_state(struct i915_error_state_file_priv *error_priv,
 	if (error->display)
 		intel_display_print_error_state(m, dev, error->display);
 
+out:
+	if (m->bytes == 0 && m->err)
+		return m->err;
+
 	return 0;
 }
 
@@ -960,15 +957,10 @@ static ssize_t i915_error_state_read(struct file *file, char __user *userbuf,
 
 	error_str.start = *pos;
 
-	ret = i915_error_state(error_priv, &error_str);
+	ret = i915_error_state_to_str(&error_str, error_priv);
 	if (ret)
 		goto out;
 
-	if (error_str.bytes == 0 && error_str.err) {
-		ret = error_str.err;
-		goto out;
-	}
-
 	ret_count = simple_read_from_buffer(userbuf, count, &tmp_pos,
 					    error_str.buf,
 					    error_str.bytes);
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index 215aa63..538b36b 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -857,6 +857,11 @@ struct drm_i915_error_state_buf {
 	loff_t pos;
 };
 
+struct i915_error_state_file_priv {
+	struct drm_device *dev;
+	struct drm_i915_error_state *error;
+};
+
 struct i915_gpu_error {
 	/* For hangcheck timer */
 #define DRM_I915_HANGCHECK_PERIOD 1500 /* in ms */
@@ -1850,6 +1855,8 @@ int i915_debugfs_init(struct drm_minor *minor);
 void i915_debugfs_cleanup(struct drm_minor *minor);
 __printf(2, 3)
 void i915_error_printf(struct drm_i915_error_state_buf *e, const char *f, ...);
+int i915_error_state_to_str(struct drm_i915_error_state_buf *estr,
+			    const struct i915_error_state_file_priv *error);
 
 /* i915_suspend.c */
 extern int i915_save_state(struct drm_device *dev);
-- 
1.7.9.5

[PATCH 2/4] drm/i915: export error state ref handling

[Mika Kuoppala]

In preparation for sysfs error state access,
export ref error state ref counting interface.

Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com>
---
 drivers/gpu/drm/i915/i915_debugfs.c |   31 ++++++++++++++++++++++---------
 drivers/gpu/drm/i915/i915_drv.h     |    3 +++
 2 files changed, 25 insertions(+), 9 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
index 77bfd51..1e22820 100644
--- a/drivers/gpu/drm/i915/i915_debugfs.c
+++ b/drivers/gpu/drm/i915/i915_debugfs.c
@@ -889,12 +889,30 @@ i915_error_state_write(struct file *filp,
 	return cnt;
 }
 
+void i915_error_state_get(struct drm_device *dev,
+			  struct i915_error_state_file_priv *error_priv)
+{
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	unsigned long flags;
+
+	spin_lock_irqsave(&dev_priv->gpu_error.lock, flags);
+	error_priv->error = dev_priv->gpu_error.first_error;
+	if (error_priv->error)
+		kref_get(&error_priv->error->ref);
+	spin_unlock_irqrestore(&dev_priv->gpu_error.lock, flags);
+
+}
+
+void i915_error_state_put(struct i915_error_state_file_priv *error_priv)
+{
+	if (error_priv->error)
+		kref_put(&error_priv->error->ref, i915_error_state_free);
+}
+
 static int i915_error_state_open(struct inode *inode, struct file *file)
 {
 	struct drm_device *dev = inode->i_private;
-	drm_i915_private_t *dev_priv = dev->dev_private;
 	struct i915_error_state_file_priv *error_priv;
-	unsigned long flags;
 
 	error_priv = kzalloc(sizeof(*error_priv), GFP_KERNEL);
 	if (!error_priv)
@@ -902,11 +920,7 @@ static int i915_error_state_open(struct inode *inode, struct file *file)
 
 	error_priv->dev = dev;
 
-	spin_lock_irqsave(&dev_priv->gpu_error.lock, flags);
-	error_priv->error = dev_priv->gpu_error.first_error;
-	if (error_priv->error)
-		kref_get(&error_priv->error->ref);
-	spin_unlock_irqrestore(&dev_priv->gpu_error.lock, flags);
+	i915_error_state_get(dev, error_priv);
 
 	file->private_data = error_priv;
 
@@ -917,8 +931,7 @@ static int i915_error_state_release(struct inode *inode, struct file *file)
 {
 	struct i915_error_state_file_priv *error_priv = file->private_data;
 
-	if (error_priv->error)
-		kref_put(&error_priv->error->ref, i915_error_state_free);
+	i915_error_state_put(error_priv);
 	kfree(error_priv);
 
 	return 0;
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index 538b36b..adc2841 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -1857,6 +1857,9 @@ __printf(2, 3)
 void i915_error_printf(struct drm_i915_error_state_buf *e, const char *f, ...);
 int i915_error_state_to_str(struct drm_i915_error_state_buf *estr,
 			    const struct i915_error_state_file_priv *error);
+void i915_error_state_get(struct drm_device *dev,
+			  struct i915_error_state_file_priv *error_priv);
+void i915_error_state_put(struct i915_error_state_file_priv *error_priv);
 
 /* i915_suspend.c */
 extern int i915_save_state(struct drm_device *dev);
-- 
1.7.9.5

[PATCH 3/4] drm/i915: introduce i915_error_state_buf_init

[Mika Kuoppala]

Make function for struct i915_error_state_buf initialization
and export it, for sysfs and debugfs.

Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com>
---
 drivers/gpu/drm/i915/i915_debugfs.c |   50 +++++++++++++++++++++--------------
 drivers/gpu/drm/i915/i915_drv.h     |    7 +++++
 2 files changed, 37 insertions(+), 20 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
index 1e22820..6ea6747 100644
--- a/drivers/gpu/drm/i915/i915_debugfs.c
+++ b/drivers/gpu/drm/i915/i915_debugfs.c
@@ -937,38 +937,48 @@ static int i915_error_state_release(struct inode *inode, struct file *file)
 	return 0;
 }
 
-static ssize_t i915_error_state_read(struct file *file, char __user *userbuf,
-				     size_t count, loff_t *pos)
+int i915_error_state_buf_init(struct drm_i915_error_state_buf *ebuf,
+			      size_t count, loff_t pos)
 {
-	struct i915_error_state_file_priv *error_priv = file->private_data;
-	struct drm_i915_error_state_buf error_str;
-	loff_t tmp_pos = 0;
-	ssize_t ret_count = 0;
-	int ret = 0;
-
-	memset(&error_str, 0, sizeof(error_str));
+	memset(ebuf, 0, sizeof(*ebuf));
 
 	/* We need to have enough room to store any i915_error_state printf
 	 * so that we can move it to start position.
 	 */
-	error_str.size = count + 1 > PAGE_SIZE ? count + 1 : PAGE_SIZE;
-	error_str.buf = kmalloc(error_str.size,
+	ebuf->size = count + 1 > PAGE_SIZE ? count + 1 : PAGE_SIZE;
+	ebuf->buf = kmalloc(ebuf->size,
 				GFP_TEMPORARY | __GFP_NORETRY | __GFP_NOWARN);
 
-	if (error_str.buf == NULL) {
-		error_str.size = PAGE_SIZE;
-		error_str.buf = kmalloc(error_str.size, GFP_TEMPORARY);
+	if (ebuf->buf == NULL) {
+		ebuf->size = PAGE_SIZE;
+		ebuf->buf = kmalloc(ebuf->size, GFP_TEMPORARY);
 	}
 
-	if (error_str.buf == NULL) {
-		error_str.size = 128;
-		error_str.buf = kmalloc(error_str.size, GFP_TEMPORARY);
+	if (ebuf->buf == NULL) {
+		ebuf->size = 128;
+		ebuf->buf = kmalloc(ebuf->size, GFP_TEMPORARY);
 	}
 
-	if (error_str.buf == NULL)
+	if (ebuf->buf == NULL)
 		return -ENOMEM;
 
-	error_str.start = *pos;
+	ebuf->start = pos;
+
+	return 0;
+}
+
+static ssize_t i915_error_state_read(struct file *file, char __user *userbuf,
+				     size_t count, loff_t *pos)
+{
+	struct i915_error_state_file_priv *error_priv = file->private_data;
+	struct drm_i915_error_state_buf error_str;
+	loff_t tmp_pos = 0;
+	ssize_t ret_count = 0;
+	int ret;
+
+	ret = i915_error_state_buf_init(&error_str, count, *pos);
+	if (ret)
+		return ret;
 
 	ret = i915_error_state_to_str(&error_str, error_priv);
 	if (ret)
@@ -983,7 +993,7 @@ static ssize_t i915_error_state_read(struct file *file, char __user *userbuf,
 	else
 		*pos = error_str.start + ret_count;
 out:
-	kfree(error_str.buf);
+	i915_error_state_buf_release(&error_str);
 	return ret ?: ret_count;
 }
 
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index adc2841..b8e16d1 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -1860,6 +1860,13 @@ int i915_error_state_to_str(struct drm_i915_error_state_buf *estr,
 void i915_error_state_get(struct drm_device *dev,
 			  struct i915_error_state_file_priv *error_priv);
 void i915_error_state_put(struct i915_error_state_file_priv *error_priv);
+int i915_error_state_buf_init(struct drm_i915_error_state_buf *eb,
+			      size_t count, loff_t pos);
+static inline void i915_error_state_buf_release(
+	struct drm_i915_error_state_buf *eb)
+{
+	kfree(eb->buf);
+}
 
 /* i915_suspend.c */
 extern int i915_save_state(struct drm_device *dev);
-- 
1.7.9.5

[PATCH 4/4] drm/i915: add i915_error_state sysfs entry

[Mika Kuoppala]

As getting error state doesn't anymore require big kmallocs,
make error state accessible also from sysfs.

Suggested-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com>
---
 drivers/gpu/drm/i915/i915_sysfs.c |   49 +++++++++++++++++++++++++++++++++++++
 1 file changed, 49 insertions(+)

diff --git a/drivers/gpu/drm/i915/i915_sysfs.c b/drivers/gpu/drm/i915/i915_sysfs.c
index 6875b56..d4c754b 100644
--- a/drivers/gpu/drm/i915/i915_sysfs.c
+++ b/drivers/gpu/drm/i915/i915_sysfs.c
@@ -409,6 +409,49 @@ static const struct attribute *gen6_attrs[] = {
 	NULL,
 };
 
+static ssize_t error_state_read(struct file *filp, struct kobject *kobj,
+				struct bin_attribute *attr, char *buf,
+				loff_t off, size_t count)
+{
+
+	struct device *kdev = container_of(kobj, struct device, kobj);
+	struct drm_minor *minor = container_of(kdev, struct drm_minor, kdev);
+	struct drm_device *dev = minor->dev;
+	struct i915_error_state_file_priv error_priv;
+	struct drm_i915_error_state_buf error_str;
+	ssize_t ret_count = 0;
+	int ret;
+
+	memset(&error_priv, 0, sizeof(error_priv));
+
+	ret = i915_error_state_buf_init(&error_str, count, off);
+	if (ret)
+		return ret;
+
+	error_priv.dev = dev;
+	i915_error_state_get(dev, &error_priv);
+
+	ret = i915_error_state_to_str(&error_str, &error_priv);
+	if (ret)
+		goto out;
+
+	ret_count = count < error_str.bytes ? count : error_str.bytes;
+
+	memcpy(buf, error_str.buf, ret_count);
+out:
+	i915_error_state_buf_release(&error_str);
+	i915_error_state_put(&error_priv);
+
+	return ret ?: ret_count;
+}
+
+static struct bin_attribute error_state_attr = {
+	.attr.name = "i915_error_state",
+	.attr.mode = 0444,
+	.size = 0,
+	.read = error_state_read,
+};
+
 void i915_setup_sysfs(struct drm_device *dev)
 {
 	int ret;
@@ -432,10 +475,16 @@ void i915_setup_sysfs(struct drm_device *dev)
 		if (ret)
 			DRM_ERROR("gen6 sysfs setup failed\n");
 	}
+
+	ret = sysfs_create_bin_file(&dev->primary->kdev.kobj,
+				    &error_state_attr);
+	if (ret)
+		DRM_ERROR("error_state sysfs setup failed\n");
 }
 
 void i915_teardown_sysfs(struct drm_device *dev)
 {
+	sysfs_remove_bin_file(&dev->primary->kdev.kobj, &error_state_attr);
 	sysfs_remove_files(&dev->primary->kdev.kobj, gen6_attrs);
 	device_remove_bin_file(&dev->primary->kdev,  &dpf_attrs);
 #ifdef CONFIG_PM
-- 
1.7.9.5

Re: [PATCH 4/4] drm/i915: add i915_error_state sysfs entry

[Chris Wilson]

On Thu, Jun 06, 2013 at 03:18:42PM +0300, Mika Kuoppala wrote:
> As getting error state doesn't anymore require big kmallocs,
> make error state accessible also from sysfs.
> 
> Suggested-by: Daniel Vetter <daniel.vetter@ffwll.ch>
> Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com>

Don't forget the ability to clear an error-state in order to capture a
fresh one.
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre

Re: [PATCH 4/4] drm/i915: add i915_error_state sysfs entry

[Daniel Vetter]

On Thu, Jun 06, 2013 at 03:18:42PM +0300, Mika Kuoppala wrote:
> As getting error state doesn't anymore require big kmallocs,
> make error state accessible also from sysfs.
> 
> Suggested-by: Daniel Vetter <daniel.vetter@ffwll.ch>
> Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com>
> ---
>  drivers/gpu/drm/i915/i915_sysfs.c |   49 +++++++++++++++++++++++++++++++++++++
>  1 file changed, 49 insertions(+)
> 
> diff --git a/drivers/gpu/drm/i915/i915_sysfs.c b/drivers/gpu/drm/i915/i915_sysfs.c
> index 6875b56..d4c754b 100644
> --- a/drivers/gpu/drm/i915/i915_sysfs.c
> +++ b/drivers/gpu/drm/i915/i915_sysfs.c
> @@ -409,6 +409,49 @@ static const struct attribute *gen6_attrs[] = {
>  	NULL,
>  };
>  
> +static ssize_t error_state_read(struct file *filp, struct kobject *kobj,
> +				struct bin_attribute *attr, char *buf,
> +				loff_t off, size_t count)
> +{
> +
> +	struct device *kdev = container_of(kobj, struct device, kobj);
> +	struct drm_minor *minor = container_of(kdev, struct drm_minor, kdev);
> +	struct drm_device *dev = minor->dev;
> +	struct i915_error_state_file_priv error_priv;
> +	struct drm_i915_error_state_buf error_str;
> +	ssize_t ret_count = 0;
> +	int ret;
> +
> +	memset(&error_priv, 0, sizeof(error_priv));
> +
> +	ret = i915_error_state_buf_init(&error_str, count, off);
> +	if (ret)
> +		return ret;
> +
> +	error_priv.dev = dev;
> +	i915_error_state_get(dev, &error_priv);
> +
> +	ret = i915_error_state_to_str(&error_str, &error_priv);
> +	if (ret)
> +		goto out;
> +
> +	ret_count = count < error_str.bytes ? count : error_str.bytes;
> +
> +	memcpy(buf, error_str.buf, ret_count);
> +out:
> +	i915_error_state_buf_release(&error_str);
> +	i915_error_state_put(&error_priv);
> +
> +	return ret ?: ret_count;
> +}
> +
> +static struct bin_attribute error_state_attr = {
> +	.attr.name = "i915_error_state",
> +	.attr.mode = 0444,

Error state could easily leak sensitive information, so I think we should
only allow root to read it. Also maybe a new name for it, i915_ is kinda
meh ... I vote for gpu_error_state.
-Daniel

> +	.size = 0,
> +	.read = error_state_read,
> +};
> +
>  void i915_setup_sysfs(struct drm_device *dev)
>  {
>  	int ret;
> @@ -432,10 +475,16 @@ void i915_setup_sysfs(struct drm_device *dev)
>  		if (ret)
>  			DRM_ERROR("gen6 sysfs setup failed\n");
>  	}
> +
> +	ret = sysfs_create_bin_file(&dev->primary->kdev.kobj,
> +				    &error_state_attr);
> +	if (ret)
> +		DRM_ERROR("error_state sysfs setup failed\n");
>  }
>  
>  void i915_teardown_sysfs(struct drm_device *dev)
>  {
> +	sysfs_remove_bin_file(&dev->primary->kdev.kobj, &error_state_attr);
>  	sysfs_remove_files(&dev->primary->kdev.kobj, gen6_attrs);
>  	device_remove_bin_file(&dev->primary->kdev,  &dpf_attrs);
>  #ifdef CONFIG_PM
> -- 
> 1.7.9.5
> 
> _______________________________________________
> Intel-gfx mailing list
> Intel-gfx@lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/intel-gfx

-- 
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch

[PATCH 4/4] drm/i915: add error_state sysfs entry

[Mika Kuoppala]

As getting error state doesn't anymore require big kmallocs,
make error state accessible also from sysfs.

v2: - error state clearing (Chris Wilson)
    - user hint, proper access mode bits and name (Daniel Vetter)

Suggested-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com>
---
 drivers/gpu/drm/i915/i915_irq.c   |    3 +-
 drivers/gpu/drm/i915/i915_sysfs.c |   71 +++++++++++++++++++++++++++++++++++++
 2 files changed, 72 insertions(+), 2 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c
index 63996aa..22ed519 100644
--- a/drivers/gpu/drm/i915/i915_irq.c
+++ b/drivers/gpu/drm/i915/i915_irq.c
@@ -1881,8 +1881,7 @@ static void i915_capture_error_state(struct drm_device *dev)
 	}
 
 	DRM_INFO("capturing error event; look for more information in "
-		 "/sys/kernel/debug/dri/%d/i915_error_state\n",
-		 dev->primary->index);
+		 "/sys/class/drm/card%d/error_state\n", dev->primary->index);
 
 	kref_init(&error->ref);
 	error->eir = I915_READ(EIR);
diff --git a/drivers/gpu/drm/i915/i915_sysfs.c b/drivers/gpu/drm/i915/i915_sysfs.c
index 6875b56..50416d1 100644
--- a/drivers/gpu/drm/i915/i915_sysfs.c
+++ b/drivers/gpu/drm/i915/i915_sysfs.c
@@ -409,6 +409,71 @@ static const struct attribute *gen6_attrs[] = {
 	NULL,
 };
 
+static ssize_t error_state_read(struct file *filp, struct kobject *kobj,
+				struct bin_attribute *attr, char *buf,
+				loff_t off, size_t count)
+{
+
+	struct device *kdev = container_of(kobj, struct device, kobj);
+	struct drm_minor *minor = container_of(kdev, struct drm_minor, kdev);
+	struct drm_device *dev = minor->dev;
+	struct i915_error_state_file_priv error_priv;
+	struct drm_i915_error_state_buf error_str;
+	ssize_t ret_count = 0;
+	int ret;
+
+	memset(&error_priv, 0, sizeof(error_priv));
+
+	ret = i915_error_state_buf_init(&error_str, count, off);
+	if (ret)
+		return ret;
+
+	error_priv.dev = dev;
+	i915_error_state_get(dev, &error_priv);
+
+	ret = i915_error_state_to_str(&error_str, &error_priv);
+	if (ret)
+		goto out;
+
+	ret_count = count < error_str.bytes ? count : error_str.bytes;
+
+	memcpy(buf, error_str.buf, ret_count);
+out:
+	i915_error_state_buf_release(&error_str);
+	i915_error_state_put(&error_priv);
+
+	return ret ?: ret_count;
+}
+
+static ssize_t error_state_write(struct file *file, struct kobject *kobj,
+				 struct bin_attribute *attr, char *buf,
+				 loff_t off, size_t count)
+{
+	struct device *kdev = container_of(kobj, struct device, kobj);
+	struct drm_minor *minor = container_of(kdev, struct drm_minor, kdev);
+	struct drm_device *dev = minor->dev;
+	int ret;
+
+	DRM_DEBUG_DRIVER("Resetting error state\n");
+
+	ret = mutex_lock_interruptible(&dev->struct_mutex);
+	if (ret)
+		return ret;
+
+	i915_destroy_error_state(dev);
+	mutex_unlock(&dev->struct_mutex);
+
+	return count;
+}
+
+static struct bin_attribute error_state_attr = {
+	.attr.name = "error_state",
+	.attr.mode = S_IRUSR | S_IWUSR,
+	.size = 0,
+	.read = error_state_read,
+	.write = error_state_write,
+};
+
 void i915_setup_sysfs(struct drm_device *dev)
 {
 	int ret;
@@ -432,10 +497,16 @@ void i915_setup_sysfs(struct drm_device *dev)
 		if (ret)
 			DRM_ERROR("gen6 sysfs setup failed\n");
 	}
+
+	ret = sysfs_create_bin_file(&dev->primary->kdev.kobj,
+				    &error_state_attr);
+	if (ret)
+		DRM_ERROR("error_state sysfs setup failed\n");
 }
 
 void i915_teardown_sysfs(struct drm_device *dev)
 {
+	sysfs_remove_bin_file(&dev->primary->kdev.kobj, &error_state_attr);
 	sysfs_remove_files(&dev->primary->kdev.kobj, gen6_attrs);
 	device_remove_bin_file(&dev->primary->kdev,  &dpf_attrs);
 #ifdef CONFIG_PM
-- 
1.7.9.5

Re: [PATCH 4/4] drm/i915: add error_state sysfs entry

[Chris Wilson]

On Thu, Jun 06, 2013 at 04:33:06PM +0300, Mika Kuoppala wrote:
> As getting error state doesn't anymore require big kmallocs,
> make error state accessible also from sysfs.
> 
> v2: - error state clearing (Chris Wilson)
>     - user hint, proper access mode bits and name (Daniel Vetter)
> 
> Suggested-by: Daniel Vetter <daniel.vetter@ffwll.ch>
> Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com>
> ---
> +static ssize_t error_state_read(struct file *filp, struct kobject *kobj,
> +				struct bin_attribute *attr, char *buf,
> +				loff_t off, size_t count)
> +{
> +
> +	struct device *kdev = container_of(kobj, struct device, kobj);
> +	struct drm_minor *minor = container_of(kdev, struct drm_minor, kdev);
> +	struct drm_device *dev = minor->dev;
> +	struct i915_error_state_file_priv error_priv;
> +	struct drm_i915_error_state_buf error_str;
> +	ssize_t ret_count = 0;
> +	int ret;
> +
> +	memset(&error_priv, 0, sizeof(error_priv));
> +
> +	ret = i915_error_state_buf_init(&error_str, count, off);
> +	if (ret)
> +		return ret;
> +
> +	error_priv.dev = dev;
> +	i915_error_state_get(dev, &error_priv);
> +
> +	ret = i915_error_state_to_str(&error_str, &error_priv);
> +	if (ret)
> +		goto out;
> +
> +	ret_count = count < error_str.bytes ? count : error_str.bytes;
> +
> +	memcpy(buf, error_str.buf, ret_count);
> +out:
> +	i915_error_state_buf_release(&error_str);
> +	i915_error_state_put(&error_priv);

Not quite oniony enough.
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre

[PATCH 4/4] drm/i915: add error_state sysfs entry

[Mika Kuoppala]

As getting error state doesn't anymore require big kmallocs,
make error state accessible also from sysfs.

v2: - error state clearing (Chris Wilson)
    - user hint, proper access mode bits and name (Daniel Vetter)

v3: release resources in proper order (Chris Wilson)

Suggested-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com>
---
 drivers/gpu/drm/i915/i915_irq.c   |    3 +-
 drivers/gpu/drm/i915/i915_sysfs.c |   71 +++++++++++++++++++++++++++++++++++++
 2 files changed, 72 insertions(+), 2 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c
index 63996aa..22ed519 100644
--- a/drivers/gpu/drm/i915/i915_irq.c
+++ b/drivers/gpu/drm/i915/i915_irq.c
@@ -1881,8 +1881,7 @@ static void i915_capture_error_state(struct drm_device *dev)
 	}
 
 	DRM_INFO("capturing error event; look for more information in "
-		 "/sys/kernel/debug/dri/%d/i915_error_state\n",
-		 dev->primary->index);
+		 "/sys/class/drm/card%d/error_state\n", dev->primary->index);
 
 	kref_init(&error->ref);
 	error->eir = I915_READ(EIR);
diff --git a/drivers/gpu/drm/i915/i915_sysfs.c b/drivers/gpu/drm/i915/i915_sysfs.c
index 6875b56..ffaf5da 100644
--- a/drivers/gpu/drm/i915/i915_sysfs.c
+++ b/drivers/gpu/drm/i915/i915_sysfs.c
@@ -409,6 +409,71 @@ static const struct attribute *gen6_attrs[] = {
 	NULL,
 };
 
+static ssize_t error_state_read(struct file *filp, struct kobject *kobj,
+				struct bin_attribute *attr, char *buf,
+				loff_t off, size_t count)
+{
+
+	struct device *kdev = container_of(kobj, struct device, kobj);
+	struct drm_minor *minor = container_of(kdev, struct drm_minor, kdev);
+	struct drm_device *dev = minor->dev;
+	struct i915_error_state_file_priv error_priv;
+	struct drm_i915_error_state_buf error_str;
+	ssize_t ret_count = 0;
+	int ret;
+
+	memset(&error_priv, 0, sizeof(error_priv));
+
+	ret = i915_error_state_buf_init(&error_str, count, off);
+	if (ret)
+		return ret;
+
+	error_priv.dev = dev;
+	i915_error_state_get(dev, &error_priv);
+
+	ret = i915_error_state_to_str(&error_str, &error_priv);
+	if (ret)
+		goto out;
+
+	ret_count = count < error_str.bytes ? count : error_str.bytes;
+
+	memcpy(buf, error_str.buf, ret_count);
+out:
+	i915_error_state_put(&error_priv);
+	i915_error_state_buf_release(&error_str);
+
+	return ret ?: ret_count;
+}
+
+static ssize_t error_state_write(struct file *file, struct kobject *kobj,
+				 struct bin_attribute *attr, char *buf,
+				 loff_t off, size_t count)
+{
+	struct device *kdev = container_of(kobj, struct device, kobj);
+	struct drm_minor *minor = container_of(kdev, struct drm_minor, kdev);
+	struct drm_device *dev = minor->dev;
+	int ret;
+
+	DRM_DEBUG_DRIVER("Resetting error state\n");
+
+	ret = mutex_lock_interruptible(&dev->struct_mutex);
+	if (ret)
+		return ret;
+
+	i915_destroy_error_state(dev);
+	mutex_unlock(&dev->struct_mutex);
+
+	return count;
+}
+
+static struct bin_attribute error_state_attr = {
+	.attr.name = "error_state",
+	.attr.mode = S_IRUSR | S_IWUSR,
+	.size = 0,
+	.read = error_state_read,
+	.write = error_state_write,
+};
+
 void i915_setup_sysfs(struct drm_device *dev)
 {
 	int ret;
@@ -432,10 +497,16 @@ void i915_setup_sysfs(struct drm_device *dev)
 		if (ret)
 			DRM_ERROR("gen6 sysfs setup failed\n");
 	}
+
+	ret = sysfs_create_bin_file(&dev->primary->kdev.kobj,
+				    &error_state_attr);
+	if (ret)
+		DRM_ERROR("error_state sysfs setup failed\n");
 }
 
 void i915_teardown_sysfs(struct drm_device *dev)
 {
+	sysfs_remove_bin_file(&dev->primary->kdev.kobj, &error_state_attr);
 	sysfs_remove_files(&dev->primary->kdev.kobj, gen6_attrs);
 	device_remove_bin_file(&dev->primary->kdev,  &dpf_attrs);
 #ifdef CONFIG_PM
-- 
1.7.9.5

Re: [PATCH 4/4] drm/i915: add error_state sysfs entry

[Chris Wilson]

On Thu, Jun 06, 2013 at 05:38:54PM +0300, Mika Kuoppala wrote:
> As getting error state doesn't anymore require big kmallocs,
> make error state accessible also from sysfs.
> 
> v2: - error state clearing (Chris Wilson)
>     - user hint, proper access mode bits and name (Daniel Vetter)
> 
> v3: release resources in proper order (Chris Wilson)
> 
> Suggested-by: Daniel Vetter <daniel.vetter@ffwll.ch>
> Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com>

I am happy overall here. We could take the plunge and use this as an
excuse to move all the error-state string handling into a new file,
which would be useful for when we pull the hangcheck mechanism out of
i915_irq.c.

Perhaps the more pressing bikeshed is that after years of
i915_error_state, I'm tired of the name and think that the 'state' was
just redundant, so perhaps '/sys/class/drm/card0/error' is a nice
minimalist and modern looking name.
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre

Re: [PATCH 4/4] drm/i915: add error_state sysfs entry

[Daniel Vetter]

On Mon, Jul 01, 2013 at 05:19:59PM +0100, Chris Wilson wrote:
> On Thu, Jun 06, 2013 at 05:38:54PM +0300, Mika Kuoppala wrote:
> > As getting error state doesn't anymore require big kmallocs,
> > make error state accessible also from sysfs.
> > 
> > v2: - error state clearing (Chris Wilson)
> >     - user hint, proper access mode bits and name (Daniel Vetter)
> > 
> > v3: release resources in proper order (Chris Wilson)
> > 
> > Suggested-by: Daniel Vetter <daniel.vetter@ffwll.ch>
> > Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com>
> 
> I am happy overall here. We could take the plunge and use this as an
> excuse to move all the error-state string handling into a new file,
> which would be useful for when we pull the hangcheck mechanism out of
> i915_irq.c.
> 
> Perhaps the more pressing bikeshed is that after years of
> i915_error_state, I'm tired of the name and think that the 'state' was
> just redundant, so perhaps '/sys/class/drm/card0/error' is a nice
> minimalist and modern looking name.

Series merged with this bikeshed applied and your r-b from irc added.

Thanks, Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch