* [PATCH 1/2] drm: Warn when leaking flip events on close
@ 2014-08-06 11:02 ville.syrjala
2014-08-06 11:02 ` [PATCH 2/2] drm/i915: Free pending page flip events at .preclose() ville.syrjala
0 siblings, 1 reply; 5+ messages in thread
From: ville.syrjala @ 2014-08-06 11:02 UTC (permalink / raw)
To: intel-gfx; +Cc: dri-devel
From: Ville Syrjälä <ville.syrjala@linux.intel.com>
Warn when there are events on the file_priv->event_list just before
file_priv gets freed. This can occur if the driver doesn't clean up
pending page flip events in ->preclose().
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
---
drivers/gpu/drm/drm_fops.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/drivers/gpu/drm/drm_fops.c b/drivers/gpu/drm/drm_fops.c
index 8f91062..0fa4dad 100644
--- a/drivers/gpu/drm/drm_fops.c
+++ b/drivers/gpu/drm/drm_fops.c
@@ -529,6 +529,8 @@ int drm_release(struct inode *inode, struct file *filp)
if (drm_core_check_feature(dev, DRIVER_PRIME))
drm_prime_destroy_file_private(&file_priv->prime);
+ WARN_ON(!list_empty(&file_priv->event_list));
+
put_pid(file_priv->pid);
kfree(file_priv);
--
1.8.5.5
_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/dri-devel
^ permalink raw reply related [flat|nested] 5+ messages in thread
* [PATCH 2/2] drm/i915: Free pending page flip events at .preclose()
2014-08-06 11:02 [PATCH 1/2] drm: Warn when leaking flip events on close ville.syrjala
@ 2014-08-06 11:02 ` ville.syrjala
2014-08-06 12:52 ` Daniel Vetter
2014-08-06 14:10 ` [PATCH 2/2] drm/i915: Free pending page flip events at .preclose() Daniel Vetter
0 siblings, 2 replies; 5+ messages in thread
From: ville.syrjala @ 2014-08-06 11:02 UTC (permalink / raw)
To: intel-gfx; +Cc: dri-devel
From: Ville Syrjälä <ville.syrjala@linux.intel.com>
If there are pending page flips when the fd gets closed those page
flips may have events associated to them. When the page flip eventually
completes it will queue the event to file_priv->event_list, but that
may be too late and file_priv->event_list has already been cleaned up.
Thus we leak a bit of kernel memory in the form of the event structure.
To avoid such problems clear out such pending events from
intel_crtc->unpin_work at ->preclose(). Any event that already made it
to file_priv->event_list will get cleaned up by the drm_release_events()
a bit later.
We can ignore the file_priv->event_space accounting since file_priv is
going away. This is already how drm core deals with pending vblank
events, which are maintained by the drm core.
What saves us from a total disaster (ie. dereferencing and alrady
freed file_priv) is the fact that the fb descruction triggers a modeset
and there we wait for pending flips.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
---
drivers/gpu/drm/i915/i915_dma.c | 3 +++
drivers/gpu/drm/i915/intel_display.c | 22 ++++++++++++++++++++++
drivers/gpu/drm/i915/intel_drv.h | 1 +
3 files changed, 26 insertions(+)
diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c
index 2e7f03a..c965698 100644
--- a/drivers/gpu/drm/i915/i915_dma.c
+++ b/drivers/gpu/drm/i915/i915_dma.c
@@ -1981,6 +1981,9 @@ void i915_driver_preclose(struct drm_device *dev, struct drm_file *file)
i915_gem_context_close(dev, file);
i915_gem_release(dev, file);
mutex_unlock(&dev->struct_mutex);
+
+ if (drm_core_check_feature(dev, DRIVER_MODESET))
+ intel_modeset_preclose(dev, file);
}
void i915_driver_postclose(struct drm_device *dev, struct drm_file *file)
diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c
index 883af0b..4230e4a 100644
--- a/drivers/gpu/drm/i915/intel_display.c
+++ b/drivers/gpu/drm/i915/intel_display.c
@@ -13416,3 +13416,25 @@ intel_display_print_error_state(struct drm_i915_error_state_buf *m,
err_printf(m, " VSYNC: %08x\n", error->transcoder[i].vsync);
}
}
+
+void intel_modeset_preclose(struct drm_device *dev, struct drm_file *file)
+{
+ struct intel_crtc *crtc;
+
+ for_each_intel_crtc(dev, crtc) {
+ struct intel_unpin_work *work;
+ unsigned long irqflags;
+
+ spin_lock_irqsave(&dev->event_lock, irqflags);
+
+ work = crtc->unpin_work;
+
+ if (work && work->event &&
+ work->event->base.file_priv == file) {
+ kfree(work->event);
+ work->event = NULL;
+ }
+
+ spin_unlock_irqrestore(&dev->event_lock, irqflags);
+ }
+}
diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h
index 28d185d..8f04ba8 100644
--- a/drivers/gpu/drm/i915/intel_drv.h
+++ b/drivers/gpu/drm/i915/intel_drv.h
@@ -888,6 +888,7 @@ void intel_mode_from_pipe_config(struct drm_display_mode *mode,
struct intel_crtc_config *pipe_config);
int intel_format_to_fourcc(int format);
void intel_crtc_wait_for_pending_flips(struct drm_crtc *crtc);
+void intel_modeset_preclose(struct drm_device *dev, struct drm_file *file);
/* intel_dp.c */
--
1.8.5.5
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 5+ messages in thread
* Re: [PATCH 2/2] drm/i915: Free pending page flip events at .preclose()
2014-08-06 11:02 ` [PATCH 2/2] drm/i915: Free pending page flip events at .preclose() ville.syrjala
@ 2014-08-06 12:52 ` Daniel Vetter
2014-08-06 13:12 ` [PATCH igt] tests: Add kms_flip_event_leak test ville.syrjala
2014-08-06 14:10 ` [PATCH 2/2] drm/i915: Free pending page flip events at .preclose() Daniel Vetter
1 sibling, 1 reply; 5+ messages in thread
From: Daniel Vetter @ 2014-08-06 12:52 UTC (permalink / raw)
To: ville.syrjala; +Cc: intel-gfx, dri-devel
On Wed, Aug 06, 2014 at 02:02:51PM +0300, ville.syrjala@linux.intel.com wrote:
> From: Ville Syrjälä <ville.syrjala@linux.intel.com>
>
> If there are pending page flips when the fd gets closed those page
> flips may have events associated to them. When the page flip eventually
> completes it will queue the event to file_priv->event_list, but that
> may be too late and file_priv->event_list has already been cleaned up.
> Thus we leak a bit of kernel memory in the form of the event structure.
>
> To avoid such problems clear out such pending events from
> intel_crtc->unpin_work at ->preclose(). Any event that already made it
> to file_priv->event_list will get cleaned up by the drm_release_events()
> a bit later.
>
> We can ignore the file_priv->event_space accounting since file_priv is
> going away. This is already how drm core deals with pending vblank
> events, which are maintained by the drm core.
>
> What saves us from a total disaster (ie. dereferencing and alrady
> freed file_priv) is the fact that the fb descruction triggers a modeset
> and there we wait for pending flips.
>
> Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Do we have an igt for this?
-Daniel
> ---
> drivers/gpu/drm/i915/i915_dma.c | 3 +++
> drivers/gpu/drm/i915/intel_display.c | 22 ++++++++++++++++++++++
> drivers/gpu/drm/i915/intel_drv.h | 1 +
> 3 files changed, 26 insertions(+)
>
> diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c
> index 2e7f03a..c965698 100644
> --- a/drivers/gpu/drm/i915/i915_dma.c
> +++ b/drivers/gpu/drm/i915/i915_dma.c
> @@ -1981,6 +1981,9 @@ void i915_driver_preclose(struct drm_device *dev, struct drm_file *file)
> i915_gem_context_close(dev, file);
> i915_gem_release(dev, file);
> mutex_unlock(&dev->struct_mutex);
> +
> + if (drm_core_check_feature(dev, DRIVER_MODESET))
> + intel_modeset_preclose(dev, file);
> }
>
> void i915_driver_postclose(struct drm_device *dev, struct drm_file *file)
> diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c
> index 883af0b..4230e4a 100644
> --- a/drivers/gpu/drm/i915/intel_display.c
> +++ b/drivers/gpu/drm/i915/intel_display.c
> @@ -13416,3 +13416,25 @@ intel_display_print_error_state(struct drm_i915_error_state_buf *m,
> err_printf(m, " VSYNC: %08x\n", error->transcoder[i].vsync);
> }
> }
> +
> +void intel_modeset_preclose(struct drm_device *dev, struct drm_file *file)
> +{
> + struct intel_crtc *crtc;
> +
> + for_each_intel_crtc(dev, crtc) {
> + struct intel_unpin_work *work;
> + unsigned long irqflags;
> +
> + spin_lock_irqsave(&dev->event_lock, irqflags);
> +
> + work = crtc->unpin_work;
> +
> + if (work && work->event &&
> + work->event->base.file_priv == file) {
> + kfree(work->event);
> + work->event = NULL;
> + }
> +
> + spin_unlock_irqrestore(&dev->event_lock, irqflags);
> + }
> +}
> diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h
> index 28d185d..8f04ba8 100644
> --- a/drivers/gpu/drm/i915/intel_drv.h
> +++ b/drivers/gpu/drm/i915/intel_drv.h
> @@ -888,6 +888,7 @@ void intel_mode_from_pipe_config(struct drm_display_mode *mode,
> struct intel_crtc_config *pipe_config);
> int intel_format_to_fourcc(int format);
> void intel_crtc_wait_for_pending_flips(struct drm_crtc *crtc);
> +void intel_modeset_preclose(struct drm_device *dev, struct drm_file *file);
>
>
> /* intel_dp.c */
> --
> 1.8.5.5
>
> _______________________________________________
> 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
^ permalink raw reply [flat|nested] 5+ messages in thread
* [PATCH igt] tests: Add kms_flip_event_leak test
2014-08-06 12:52 ` Daniel Vetter
@ 2014-08-06 13:12 ` ville.syrjala
0 siblings, 0 replies; 5+ messages in thread
From: ville.syrjala @ 2014-08-06 13:12 UTC (permalink / raw)
To: intel-gfx; +Cc: dri-devel
From: Ville Syrjälä <ville.syrjala@linux.intel.com>
kms_flip_event_leak will issue a page flip and close the file
descriptor before the flip has finished. This may cause the kernel
to leak the page flip event. The test itself won't actually fail but
if the kernel notices the leak and WARNs piglit will report a failure.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
---
tests/Makefile.sources | 1 +
tests/kms_flip_event_leak.c | 132 ++++++++++++++++++++++++++++++++++++++++++++
2 files changed, 133 insertions(+)
create mode 100644 tests/kms_flip_event_leak.c
diff --git a/tests/Makefile.sources b/tests/Makefile.sources
index 0eb9369..698e290 100644
--- a/tests/Makefile.sources
+++ b/tests/Makefile.sources
@@ -66,6 +66,7 @@ TESTS_progs_M = \
kms_cursor_crc \
kms_fbc_crc \
kms_flip \
+ kms_flip_event_leak \
kms_flip_tiling \
kms_mmio_vs_cs_flip \
kms_pipe_crc_basic \
diff --git a/tests/kms_flip_event_leak.c b/tests/kms_flip_event_leak.c
new file mode 100644
index 0000000..9924333
--- /dev/null
+++ b/tests/kms_flip_event_leak.c
@@ -0,0 +1,132 @@
+/*
+ * Copyright © 2014 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ *
+ */
+
+#include <errno.h>
+#include <stdbool.h>
+#include <stdio.h>
+#include <string.h>
+
+#include "drmtest.h"
+#include "igt_debugfs.h"
+#include "igt_kms.h"
+#include "intel_chipset.h"
+#include "intel_batchbuffer.h"
+#include "ioctl_wrappers.h"
+
+typedef struct {
+ int drm_fd;
+ igt_display_t display;
+} data_t;
+
+/*
+ * This test tries to provoke the kernel to leak a pending page flip event
+ * when the fd is closed before the flip has completed. The test itself won't
+ * fail even if the kernel leaks the event, but the resulting dmesg WARN
+ * will cause piglit to report a failure.
+ */
+static bool test(data_t *data, enum pipe pipe, igt_output_t *output)
+{
+ igt_plane_t *primary;
+ drmModeModeInfo *mode;
+ struct igt_fb fb[2];
+ int fd, ret;
+
+ /* select the pipe we want to use */
+ igt_output_set_pipe(output, pipe);
+ igt_display_commit(&data->display);
+
+ if (!output->valid) {
+ igt_output_set_pipe(output, PIPE_ANY);
+ igt_display_commit(&data->display);
+ return false;
+ }
+
+ primary = igt_output_get_plane(output, IGT_PLANE_PRIMARY);
+ mode = igt_output_get_mode(output);
+
+ igt_create_color_fb(data->drm_fd, mode->hdisplay, mode->vdisplay,
+ DRM_FORMAT_XRGB8888,
+ true, /* tiled */
+ 0.0, 0.0, 0.0, &fb[0]);
+
+ igt_plane_set_fb(primary, &fb[0]);
+ igt_display_commit2(&data->display, COMMIT_LEGACY);
+
+ fd = drm_open_any();
+
+ ret = drmDropMaster(data->drm_fd);
+ igt_assert(ret == 0);
+
+ ret = drmSetMaster(fd);
+ igt_assert(ret == 0);
+
+ igt_create_color_fb(fd, mode->hdisplay, mode->vdisplay,
+ DRM_FORMAT_XRGB8888,
+ true, /* tiled */
+ 0.0, 0.0, 0.0, &fb[1]);
+ ret = drmModePageFlip(fd, output->config.crtc->crtc_id,
+ fb[1].fb_id, DRM_MODE_PAGE_FLIP_EVENT,
+ data);
+ igt_assert(ret == 0);
+
+ ret = close(fd);
+ igt_assert(ret == 0);
+
+ ret = drmSetMaster(data->drm_fd);
+ igt_assert(ret == 0);
+
+ igt_plane_set_fb(primary, NULL);
+ igt_output_set_pipe(output, PIPE_ANY);
+ igt_display_commit(&data->display);
+
+ igt_remove_fb(data->drm_fd, &fb[0]);
+
+ return true;
+}
+
+igt_simple_main
+{
+ data_t data = {};
+ igt_output_t *output;
+ int valid_tests = 0;
+ enum pipe pipe;
+
+ igt_skip_on_simulation();
+
+ data.drm_fd = drm_open_any();
+ igt_set_vt_graphics_mode();
+
+ igt_display_init(&data.display, data.drm_fd);
+
+ for (pipe = 0; pipe < 3; pipe++) {
+ for_each_connected_output(&data.display, output) {
+ if (test(&data, pipe, output))
+ valid_tests++;
+ }
+ }
+
+ igt_require_f(valid_tests, "no valid crtc/connector combinations found\n");
+
+ igt_display_fini(&data.display);
+}
--
1.8.5.5
_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/dri-devel
^ permalink raw reply related [flat|nested] 5+ messages in thread
* Re: [PATCH 2/2] drm/i915: Free pending page flip events at .preclose()
2014-08-06 11:02 ` [PATCH 2/2] drm/i915: Free pending page flip events at .preclose() ville.syrjala
2014-08-06 12:52 ` Daniel Vetter
@ 2014-08-06 14:10 ` Daniel Vetter
1 sibling, 0 replies; 5+ messages in thread
From: Daniel Vetter @ 2014-08-06 14:10 UTC (permalink / raw)
To: ville.syrjala; +Cc: intel-gfx, dri-devel
On Wed, Aug 06, 2014 at 02:02:51PM +0300, ville.syrjala@linux.intel.com wrote:
> From: Ville Syrjälä <ville.syrjala@linux.intel.com>
>
> If there are pending page flips when the fd gets closed those page
> flips may have events associated to them. When the page flip eventually
> completes it will queue the event to file_priv->event_list, but that
> may be too late and file_priv->event_list has already been cleaned up.
> Thus we leak a bit of kernel memory in the form of the event structure.
>
> To avoid such problems clear out such pending events from
> intel_crtc->unpin_work at ->preclose(). Any event that already made it
> to file_priv->event_list will get cleaned up by the drm_release_events()
> a bit later.
>
> We can ignore the file_priv->event_space accounting since file_priv is
> going away. This is already how drm core deals with pending vblank
> events, which are maintained by the drm core.
>
> What saves us from a total disaster (ie. dereferencing and alrady
> freed file_priv) is the fact that the fb descruction triggers a modeset
> and there we wait for pending flips.
>
> Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
> ---
> drivers/gpu/drm/i915/i915_dma.c | 3 +++
> drivers/gpu/drm/i915/intel_display.c | 22 ++++++++++++++++++++++
> drivers/gpu/drm/i915/intel_drv.h | 1 +
> 3 files changed, 26 insertions(+)
>
> diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c
> index 2e7f03a..c965698 100644
> --- a/drivers/gpu/drm/i915/i915_dma.c
> +++ b/drivers/gpu/drm/i915/i915_dma.c
> @@ -1981,6 +1981,9 @@ void i915_driver_preclose(struct drm_device *dev, struct drm_file *file)
> i915_gem_context_close(dev, file);
> i915_gem_release(dev, file);
> mutex_unlock(&dev->struct_mutex);
> +
> + if (drm_core_check_feature(dev, DRIVER_MODESET))
> + intel_modeset_preclose(dev, file);
> }
>
> void i915_driver_postclose(struct drm_device *dev, struct drm_file *file)
> diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c
> index 883af0b..4230e4a 100644
> --- a/drivers/gpu/drm/i915/intel_display.c
> +++ b/drivers/gpu/drm/i915/intel_display.c
> @@ -13416,3 +13416,25 @@ intel_display_print_error_state(struct drm_i915_error_state_buf *m,
> err_printf(m, " VSYNC: %08x\n", error->transcoder[i].vsync);
> }
> }
> +
> +void intel_modeset_preclose(struct drm_device *dev, struct drm_file *file)
> +{
> + struct intel_crtc *crtc;
> +
> + for_each_intel_crtc(dev, crtc) {
> + struct intel_unpin_work *work;
> + unsigned long irqflags;
> +
> + spin_lock_irqsave(&dev->event_lock, irqflags);
> +
> + work = crtc->unpin_work;
> +
> + if (work && work->event &&
> + work->event->base.file_priv == file) {
> + kfree(work->event);
> + work->event = NULL;
> + }
> +
> + spin_unlock_irqrestore(&dev->event_lock, irqflags);
> + }
I wonder whether we shouldn't do this in the drm core, with a per-file
event list. Anyway, good for now together with the igt, we can pimp this
later.
Queued for -next, thanks for the patch.
-Daniel
> +}
> diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h
> index 28d185d..8f04ba8 100644
> --- a/drivers/gpu/drm/i915/intel_drv.h
> +++ b/drivers/gpu/drm/i915/intel_drv.h
> @@ -888,6 +888,7 @@ void intel_mode_from_pipe_config(struct drm_display_mode *mode,
> struct intel_crtc_config *pipe_config);
> int intel_format_to_fourcc(int format);
> void intel_crtc_wait_for_pending_flips(struct drm_crtc *crtc);
> +void intel_modeset_preclose(struct drm_device *dev, struct drm_file *file);
>
>
> /* intel_dp.c */
> --
> 1.8.5.5
>
> _______________________________________________
> 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
^ permalink raw reply [flat|nested] 5+ messages in thread
end of thread, other threads:[~2014-08-06 14:10 UTC | newest]
Thread overview: 5+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2014-08-06 11:02 [PATCH 1/2] drm: Warn when leaking flip events on close ville.syrjala
2014-08-06 11:02 ` [PATCH 2/2] drm/i915: Free pending page flip events at .preclose() ville.syrjala
2014-08-06 12:52 ` Daniel Vetter
2014-08-06 13:12 ` [PATCH igt] tests: Add kms_flip_event_leak test ville.syrjala
2014-08-06 14:10 ` [PATCH 2/2] drm/i915: Free pending page flip events at .preclose() Daniel Vetter
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox