Skip to contentSkip to conversation

drm/i915: add missing rpm ref to i915_gem_pwrite_ioctl

PATCH Applied Imre Deak 4 participants 6 messages Nov 12, 2014 – Dec 18, 2014

Subsystems (3)
lore
Share & tools
lore

Imre Deak

author
#0 lore raw

PATCH 1/1 drm/i915: add missing rpm ref to i915_gem_pwrite_ioctl

Without this RPM ref we can hit the device suspended WARN via:
i915_gem_object_pin()->ggtt_bind_vma->gen6_ggtt_insert_entries(). I noticed this on my BYT while keeping the i915 device in runtime suspended state for a while. I chose this place to take the ref to avoid the possible deadlock via the mutex_lock taken both later in this function and in the runtime suspend handler. This can happen if an RPM suspend event is queued and need to be flushed before taking the RPM ref.

---
 drivers/gpu/drm/i915/i915_gem.c | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)
drivers/gpu/drm/i915/i915_gem.c +7 -1
index 86cf428..ec2ee28 100644
@@ -1075,6 +1075,7 @@ int
 i915_gem_pwrite_ioctl(struct drm_device *dev, void *data,
 		      struct drm_file *file)
 {
+	struct drm_i915_private *dev_priv = dev->dev_private;
 	struct drm_i915_gem_pwrite *args = data;
 	struct drm_i915_gem_object *obj;
 	int ret;
@@ -1094,9 +1095,11 @@ i915_gem_pwrite_ioctl(struct drm_device *dev, void *data,
 			return -EFAULT;
 	}
 
+	intel_runtime_pm_get(dev_priv);
+
 	ret = i915_mutex_lock_interruptible(dev);
 	if (ret)
-		return ret;
+		goto put_rpm;
 
 	obj = to_intel_bo(drm_gem_object_lookup(dev, file, args->handle));
 	if (&obj->base == NULL) {
@@ -1148,6 +1151,9 @@ out:
 	drm_gem_object_unreference(&obj->base);
 unlock:
 	mutex_unlock(&dev->struct_mutex);
+put_rpm:
+	intel_runtime_pm_put(dev_priv);
+
 	return ret;
 }
 
Signature · 6 lines
1.8.4

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
#1 lore raw
Quoted 10 lines · Imre Deak

Without this RPM ref we can hit the device suspended WARN via:
i915_gem_object_pin()->ggtt_bind_vma->gen6_ggtt_insert_entries(). I noticed this on my BYT while keeping the i915 device in runtime suspended state for a while. I chose this place to take the ref to avoid the possible deadlock via the mutex_lock taken both later in this function and in the runtime suspend handler. This can happen if an RPM suspend event is queued and need to be flushed before taking the RPM ref.

We should be able to hit this easily with an igt by doing a pwrite to an object which is attached to a cursor. I.e.
1. Set up mode with cursor visible.
2. dpms off
-> ensure rpm is entered
3. pwrite to cursor -> boom.

Can you please add this as a new subtest to pm_rpm.c? I think it should fit fairly easily there.

Thanks, Daniel

Quoted 46 lines
---
 drivers/gpu/drm/i915/i915_gem.c | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)
drivers/gpu/drm/i915/i915_gem.c +7 -1
index 86cf428..ec2ee28 100644
@@ -1075,6 +1075,7 @@ int
 i915_gem_pwrite_ioctl(struct drm_device *dev, void *data,
 		      struct drm_file *file)
 {
+	struct drm_i915_private *dev_priv = dev->dev_private;
 	struct drm_i915_gem_pwrite *args = data;
 	struct drm_i915_gem_object *obj;
 	int ret;
@@ -1094,9 +1095,11 @@ i915_gem_pwrite_ioctl(struct drm_device *dev, void *data,
 			return -EFAULT;
 	}
 
+	intel_runtime_pm_get(dev_priv);
+
 	ret = i915_mutex_lock_interruptible(dev);
 	if (ret)
-		return ret;
+		goto put_rpm;
 
 	obj = to_intel_bo(drm_gem_object_lookup(dev, file, args->handle));
 	if (&obj->base == NULL) {
@@ -1148,6 +1151,9 @@ out:
 	drm_gem_object_unreference(&obj->base);
 unlock:
 	mutex_unlock(&dev->struct_mutex);
+put_rpm:
+	intel_runtime_pm_put(dev_priv);
+
 	return ret;
 }
 
Signature · 6 lines
1.8.4

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 7 lines
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
author
#2 lore raw

PATCH 1/1 tests/pm_rpm: add gem-evict-pwrite subtest

This triggers a device suspended WARN in the kernel in gen6_ggtt_insert_entries() while calling the GEM pwrite ioctl.

The sequence is suggested by Daniel.

---
 lib/igt_aux.h  |  3 +++
 tests/pm_rpm.c | 25 +++++++++++++++++++++++++
 2 files changed, 28 insertions(+)
lib/igt_aux.h +3 -0
index 9b42918..7bca11c 100644
@@ -31,6 +31,9 @@
 #include <intel_bufmgr.h>
 #include <stdbool.h>
 
+extern drm_intel_bo **trash_bos;
+extern int num_trash_bos;
+
 /* auxialiary igt helpers from igt_aux.c */
 /* generally useful helpers */
 void igt_fork_signal_helper(void);
tests/pm_rpm.c +25 -0
index 1e65c04..c120d75 100644
@@ -1319,6 +1319,29 @@ static void gem_idle_subtest(void)
 	gem_quiescent_gpu(drm_fd);
 }
 
+static void gem_evict_pwrite_subtest(void)
+{
+	static drm_intel_bufmgr *bufmgr;
+	uint32_t buf;
+	int i;
+
+	bufmgr = drm_intel_bufmgr_gem_init(drm_fd, 4096);
+	igt_assert(bufmgr);
+	igt_init_aperture_trashers(bufmgr);
+
+	igt_trash_aperture();
+
+	disable_or_dpms_all_screens_and_wait(&ms_data, true);
+	igt_assert(wait_for_suspended());
+
+	buf = 0;
+	for (i = 0; i < num_trash_bos; i++)
+		gem_write(drm_fd, trash_bos[i]->handle, 0, &buf, sizeof(buf));
+
+	igt_cleanup_aperture_trashers();
+	drm_intel_bufmgr_destroy(bufmgr);
+}
+
 /* This also triggered WARNs on dmesg at some point. */
 static void reg_read_ioctl_subtest(void)
 {
@@ -1830,6 +1853,8 @@ int main(int argc, char *argv[])
 		gem_execbuf_subtest();
 	igt_subtest("gem-idle")
 		gem_idle_subtest();
+	igt_subtest("gem-evict-pwrite")
+		gem_evict_pwrite_subtest();
 
 	/* Planes and cursors */
 	igt_subtest("cursor")
Signature · 6 lines
1.8.4

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Tested-by on v1 1/1
#3 lore raw

Re: [PATCH] drm/i915: add missing rpm ref to

In reply to Imre Deak

-------------------------------------Summary-------------------------------------
Platform: baseline_drm_intel_nightly_pass_rate->patch_applied_pass_rate
BYT: pass/total=291/291->290/291
PNV: pass/total=356/356->356/356
ILK: pass/total=372/372->371/372
IVB: pass/total=545/546->543/546
SNB: pass/total=380/380->379/380
HSW: pass/total=579/579->579/579
BDW: pass/total=422/423->422/423
-------------------------------------Detailed-------------------------------------
test_platform: test_suite, test_case, result_with_drm_intel_nightly(count, machine_id...)...->result_with_patch_applied(count, machine_id)...
BYT: Intel_gpu_tools, igt_kms_setmode_invalid-clone-single-crtc, TIMEOUT(9, M36M31)PASS(1, M31) -> TIMEOUT(1, M31)PASS(3, M31)
ILK: Intel_gpu_tools, igt_kms_setmode_invalid-clone-single-crtc, FAIL(2, M26)DMESG_FAIL(1, M26)TIMEOUT(13, M37M6M26)PASS(1, M26) -> TIMEOUT(1, M37)PASS(3, M37)
IVB: Intel_gpu_tools, igt_gem_bad_reloc_negative-reloc, NSPT(5, M21M34)PASS(8, M34M21) -> NSPT(1, M21)PASS(3, M21)
IVB: Intel_gpu_tools, igt_kms_cursor_crc_cursor-128x128-sliding, PASS(4, M21) -> DMESG_WARN(1, M21)PASS(3, M21)
IVB: Intel_gpu_tools, igt_kms_cursor_crc_cursor-64x64-offscreen, PASS(4, M21) -> DMESG_WARN(1, M21)PASS(3, M21)
IVB: Intel_gpu_tools, igt_kms_setmode_invalid-clone-single-crtc, TIMEOUT(12, M34M21)PASS(1, M21) -> TIMEOUT(1, M21)PASS(3, M21)
SNB: Intel_gpu_tools, igt_kms_setmode_invalid-clone-single-crtc, TIMEOUT(18, M35M22)PASS(1, M35) -> TIMEOUT(1, M22)PASS(3, M22)
BDW: Intel_gpu_tools, igt_gem_reset_stats_ban-bsd, DMESG_WARN(1, M28)PASS(15, M42M30M28) -> PASS(4, M42)
BDW: Intel_gpu_tools, igt_kms_setmode_invalid-clone-single-crtc, TIMEOUT(15, M42M30M28)PASS(1, M28) -> TIMEOUT(1, M42)PASS(3, M42)
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

author
#4 lore raw
In reply to Imre Deak
Quoted 8 lines · Imre Deak

Without this RPM ref we can hit the device suspended WARN via:
i915_gem_object_pin()->ggtt_bind_vma->gen6_ggtt_insert_entries(). I noticed this on my BYT while keeping the i915 device in runtime suspended state for a while. I chose this place to take the ref to avoid the possible deadlock via the mutex_lock taken both later in this function and in the runtime suspend handler. This can happen if an RPM suspend event is queued and need to be flushed before taking the RPM ref.

Testcase: igt/pm_rpm/gem-evict-pwrite
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=87363

Quoted 31 lines
---
 drivers/gpu/drm/i915/i915_gem.c | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)
drivers/gpu/drm/i915/i915_gem.c +4 -1
index 86cf428..ec2ee28 100644
@@ -1075,6 +1075,7 @@ int
 i915_gem_pwrite_ioctl(struct drm_device *dev, void *data,
 		      struct drm_file *file)
 {
+	struct drm_i915_private *dev_priv = dev->dev_private;
 	struct drm_i915_gem_pwrite *args = data;
 	struct drm_i915_gem_object *obj;
 	int ret;
@@ -1094,9 +1095,11 @@ i915_gem_pwrite_ioctl(struct drm_device *dev, void *data,
 			return -EFAULT;
 	}
 
+	intel_runtime_pm_get(dev_priv);
+
 	ret = i915_mutex_lock_interruptible(dev);
 	if (ret)
-		return ret;
+		goto put_rpm;
 
 	obj = to_intel_bo(drm_gem_object_lookup(dev, file, args->handle));
 	if (&obj->base == NULL) {

drivers/gpu/drm/i915/i915_gem.c out: · line 1159

Quoted 9 lines
drivers/gpu/drm/i915/i915_gem.c +3 -0
@@ -1148,6 +1151,9 @@ out:
 	drm_gem_object_unreference(&obj->base);
 unlock:
 	mutex_unlock(&dev->struct_mutex);
+put_rpm:
+	intel_runtime_pm_put(dev_priv);
+
 	return ret;
 }
 

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

maintainer Applied
#5 lore raw
Quoted 10 lines · Imre Deak
Quoted 8 lines · Imre Deak

Without this RPM ref we can hit the device suspended WARN via:
i915_gem_object_pin()->ggtt_bind_vma->gen6_ggtt_insert_entries(). I noticed this on my BYT while keeping the i915 device in runtime suspended state for a while. I chose this place to take the ref to avoid the possible deadlock via the mutex_lock taken both later in this function and in the runtime suspend handler. This can happen if an RPM suspend event is queued and need to be flushed before taking the RPM ref.

Quoted 2 lines

Testcase: igt/pm_rpm/gem-evict-pwrite
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=87363

Pushed to drm-intel-next-fixes, thanks for the patch.

BR,
Jani.

Quoted 47 lines
Quoted 32 lines
---
 drivers/gpu/drm/i915/i915_gem.c | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)
drivers/gpu/drm/i915/i915_gem.c +4 -1
index 86cf428..ec2ee28 100644
@@ -1075,6 +1075,7 @@ int
 i915_gem_pwrite_ioctl(struct drm_device *dev, void *data,
 		      struct drm_file *file)
 {
+	struct drm_i915_private *dev_priv = dev->dev_private;
 	struct drm_i915_gem_pwrite *args = data;
 	struct drm_i915_gem_object *obj;
 	int ret;
@@ -1094,9 +1095,11 @@ i915_gem_pwrite_ioctl(struct drm_device *dev, void *data,
 			return -EFAULT;
 	}
 
+	intel_runtime_pm_get(dev_priv);
+
 	ret = i915_mutex_lock_interruptible(dev);
 	if (ret)
-		return ret;
+		goto put_rpm;
 
 	obj = to_intel_bo(drm_gem_object_lookup(dev, file, args->handle));
 	if (&obj->base == NULL) {

drivers/gpu/drm/i915/i915_gem.c out: · line 1159

Quoted 9 lines
drivers/gpu/drm/i915/i915_gem.c +3 -0
@@ -1148,6 +1151,9 @@ out:
 	drm_gem_object_unreference(&obj->base);
 unlock:
 	mutex_unlock(&dev->struct_mutex);
+put_rpm:
+	intel_runtime_pm_put(dev_priv);
+
 	return ret;
 }
 

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Signature · 5 lines
Jani Nikula, Intel Open Source Technology Center
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx